2
0
mirror of https://github.com/openvswitch/ovs synced 2025-08-31 06:15:47 +00:00
Files
ovs/ovsdb/monitor.c

1603 lines
50 KiB
C
Raw Normal View History

/*
* Copyright (c) 2015 Nicira, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at:
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <config.h>
#include <errno.h>
#include "bitmap.h"
#include "column.h"
#include "openvswitch/dynamic-string.h"
#include "openvswitch/json.h"
#include "jsonrpc.h"
#include "ovsdb-error.h"
#include "ovsdb-parser.h"
#include "ovsdb.h"
#include "row.h"
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
#include "condition.h"
#include "simap.h"
#include "hash.h"
#include "table.h"
#include "hash.h"
#include "timeval.h"
#include "transaction.h"
#include "jsonrpc-server.h"
#include "monitor.h"
#include "util.h"
#include "openvswitch/vlog.h"
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
VLOG_DEFINE_THIS_MODULE(ovsdb_monitor);
static const struct ovsdb_replica_class ovsdb_jsonrpc_replica_class;
static struct hmap ovsdb_monitors = HMAP_INITIALIZER(&ovsdb_monitors);
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
/* Keep state of session's conditions */
struct ovsdb_monitor_session_condition {
bool conditional;
size_t n_true_cnd;
struct shash tables; /* Contains
* "struct ovsdb_monitor_table_condition *"s. */
};
/* Monitored table session's conditions */
struct ovsdb_monitor_table_condition {
const struct ovsdb_table *table;
struct ovsdb_monitor_table *mt;
struct ovsdb_condition old_condition;
struct ovsdb_condition new_condition;
};
/* Backend monitor.
*
* ovsdb_monitor keep track of the ovsdb changes.
*/
/* A collection of tables being monitored. */
struct ovsdb_monitor {
struct ovsdb_replica replica;
struct shash tables; /* Holds "struct ovsdb_monitor_table"s. */
struct ovs_list jsonrpc_monitors; /* Contains "jsonrpc_monitor_node"s. */
struct ovsdb *db;
uint64_t n_transactions; /* Count number of committed transactions. */
struct hmap_node hmap_node; /* Elements within ovsdb_monitors. */
struct hmap json_cache; /* Contains "ovsdb_monitor_json_cache_node"s.*/
};
/* A json object of updates between 'from_txn' and 'dbmon->n_transactions'
* inclusive. */
struct ovsdb_monitor_json_cache_node {
struct hmap_node hmap_node; /* Elements in json cache. */
enum ovsdb_monitor_version version;
uint64_t from_txn;
struct json *json; /* Null, or a cloned of json */
};
struct jsonrpc_monitor_node {
struct ovsdb_jsonrpc_monitor *jsonrpc_monitor;
struct ovs_list node;
};
/* A particular column being monitored. */
struct ovsdb_monitor_column {
const struct ovsdb_column *column;
enum ovsdb_monitor_selection select;
bool monitored;
};
/* A row that has changed in a monitored table. */
struct ovsdb_monitor_row {
struct hmap_node hmap_node; /* In ovsdb_jsonrpc_monitor_table.changes. */
struct uuid uuid; /* UUID of row that changed. */
struct ovsdb_datum *old; /* Old data, NULL for an inserted row. */
struct ovsdb_datum *new; /* New data, NULL for a deleted row. */
};
/* Contains 'struct ovsdb_monitor_row's for rows that have been
* updated but not yet flushed to all the jsonrpc connection.
*
* 'n_refs' represent the number of jsonrpc connections that have
* not received updates. Generate the update for the last jsonprc
* connection will also destroy the whole "struct ovsdb_monitor_changes"
* object.
*
* 'transaction' stores the first update's transaction id.
* */
struct ovsdb_monitor_changes {
struct ovsdb_monitor_table *mt;
struct hmap rows;
int n_refs;
uint64_t transaction;
struct hmap_node hmap_node; /* Element in ovsdb_monitor_tables' changes
hmap. */
};
/* A particular table being monitored. */
struct ovsdb_monitor_table {
const struct ovsdb_table *table;
/* This is the union (bitwise-OR) of the 'select' values in all of the
* members of 'columns' below. */
enum ovsdb_monitor_selection select;
/* Columns being monitored. */
struct ovsdb_monitor_column *columns;
size_t n_columns;
size_t n_monitored_columns;
size_t allocated_columns;
/* Columns in ovsdb_monitor_row have different indexes then in
* ovsdb_row. This field maps between column->index to the index in the
* ovsdb_monitor_row. It is used for condition evaluation. */
unsigned int *columns_index_map;
/* Contains 'ovsdb_monitor_changes' indexed by 'transaction'. */
struct hmap changes;
};
enum ovsdb_monitor_row_type {
OVSDB_ROW,
OVSDB_MONITOR_ROW
};
typedef struct json *
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
(*compose_row_update_cb_func)
(const struct ovsdb_monitor_table *mt,
const struct ovsdb_monitor_session_condition * condition,
enum ovsdb_monitor_row_type row_type,
const void *,
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
bool initial, unsigned long int *changed);
static void ovsdb_monitor_destroy(struct ovsdb_monitor *dbmon);
static struct ovsdb_monitor_changes * ovsdb_monitor_table_add_changes(
struct ovsdb_monitor_table *mt, uint64_t next_txn);
static struct ovsdb_monitor_changes *ovsdb_monitor_table_find_changes(
struct ovsdb_monitor_table *mt, uint64_t unflushed);
static void ovsdb_monitor_changes_destroy(
struct ovsdb_monitor_changes *changes);
static void ovsdb_monitor_table_track_changes(struct ovsdb_monitor_table *mt,
uint64_t unflushed);
static uint32_t
json_cache_hash(enum ovsdb_monitor_version version, uint64_t from_txn)
{
uint32_t hash;
hash = hash_uint64(version);
hash = hash_uint64_basis(from_txn, hash);
return hash;
}
static struct ovsdb_monitor_json_cache_node *
ovsdb_monitor_json_cache_search(const struct ovsdb_monitor *dbmon,
enum ovsdb_monitor_version version,
uint64_t from_txn)
{
struct ovsdb_monitor_json_cache_node *node;
uint32_t hash = json_cache_hash(version, from_txn);
HMAP_FOR_EACH_WITH_HASH(node, hmap_node, hash, &dbmon->json_cache) {
if (node->from_txn == from_txn && node->version == version) {
return node;
}
}
return NULL;
}
static void
ovsdb_monitor_json_cache_insert(struct ovsdb_monitor *dbmon,
enum ovsdb_monitor_version version,
uint64_t from_txn, struct json *json)
{
struct ovsdb_monitor_json_cache_node *node;
uint32_t hash = json_cache_hash(version, from_txn);
node = xmalloc(sizeof *node);
node->version = version;
node->from_txn = from_txn;
node->json = json ? json_clone(json) : NULL;
hmap_insert(&dbmon->json_cache, &node->hmap_node, hash);
}
static void
ovsdb_monitor_json_cache_flush(struct ovsdb_monitor *dbmon)
{
struct ovsdb_monitor_json_cache_node *node;
HMAP_FOR_EACH_POP(node, hmap_node, &dbmon->json_cache) {
json_destroy(node->json);
free(node);
}
}
static int
compare_ovsdb_monitor_column(const void *a_, const void *b_)
{
const struct ovsdb_monitor_column *a = a_;
const struct ovsdb_monitor_column *b = b_;
/* put all monitored columns at the begining */
if (a->monitored != b->monitored) {
return a->monitored ? -1 : 1;
}
return a->column < b->column ? -1 : a->column > b->column;
}
static struct ovsdb_monitor *
ovsdb_monitor_cast(struct ovsdb_replica *replica)
{
ovs_assert(replica->class == &ovsdb_jsonrpc_replica_class);
return CONTAINER_OF(replica, struct ovsdb_monitor, replica);
}
/* Finds and returns the ovsdb_monitor_row in 'mt->changes->rows' for the
* given 'uuid', or NULL if there is no such row. */
static struct ovsdb_monitor_row *
ovsdb_monitor_changes_row_find(const struct ovsdb_monitor_changes *changes,
const struct uuid *uuid)
{
struct ovsdb_monitor_row *row;
HMAP_FOR_EACH_WITH_HASH (row, hmap_node, uuid_hash(uuid),
&changes->rows) {
if (uuid_equals(uuid, &row->uuid)) {
return row;
}
}
return NULL;
}
/* Allocates an array of 'mt->n_columns' ovsdb_datums and initializes them as
* copies of the data in 'row' drawn from the columns represented by
* mt->columns[]. Returns the array.
*
* If 'row' is NULL, returns NULL. */
static struct ovsdb_datum *
clone_monitor_row_data(const struct ovsdb_monitor_table *mt,
const struct ovsdb_row *row)
{
struct ovsdb_datum *data;
size_t i;
if (!row) {
return NULL;
}
data = xmalloc(mt->n_columns * sizeof *data);
for (i = 0; i < mt->n_columns; i++) {
const struct ovsdb_column *c = mt->columns[i].column;
const struct ovsdb_datum *src = &row->fields[c->index];
struct ovsdb_datum *dst = &data[i];
const struct ovsdb_type *type = &c->type;
ovsdb_datum_clone(dst, src, type);
}
return data;
}
/* Replaces the mt->n_columns ovsdb_datums in row[] by copies of the data from
* in 'row' drawn from the columns represented by mt->columns[]. */
static void
update_monitor_row_data(const struct ovsdb_monitor_table *mt,
const struct ovsdb_row *row,
struct ovsdb_datum *data)
{
size_t i;
for (i = 0; i < mt->n_columns; i++) {
const struct ovsdb_column *c = mt->columns[i].column;
const struct ovsdb_datum *src = &row->fields[c->index];
struct ovsdb_datum *dst = &data[i];
const struct ovsdb_type *type = &c->type;
if (!ovsdb_datum_equals(src, dst, type)) {
ovsdb_datum_destroy(dst, type);
ovsdb_datum_clone(dst, src, type);
}
}
}
/* Frees all of the mt->n_columns ovsdb_datums in data[], using the types taken
* from mt->columns[], plus 'data' itself. */
static void
free_monitor_row_data(const struct ovsdb_monitor_table *mt,
struct ovsdb_datum *data)
{
if (data) {
size_t i;
for (i = 0; i < mt->n_columns; i++) {
const struct ovsdb_column *c = mt->columns[i].column;
ovsdb_datum_destroy(&data[i], &c->type);
}
free(data);
}
}
/* Frees 'row', which must have been created from 'mt'. */
static void
ovsdb_monitor_row_destroy(const struct ovsdb_monitor_table *mt,
struct ovsdb_monitor_row *row)
{
if (row) {
free_monitor_row_data(mt, row->old);
free_monitor_row_data(mt, row->new);
free(row);
}
}
static void
ovsdb_monitor_columns_sort(struct ovsdb_monitor *dbmon)
{
int i;
struct shash_node *node;
SHASH_FOR_EACH (node, &dbmon->tables) {
struct ovsdb_monitor_table *mt = node->data;
qsort(mt->columns, mt->n_columns, sizeof *mt->columns,
compare_ovsdb_monitor_column);
for (i = 0; i < mt->n_columns; i++) {
/* re-set index map due to sort */
mt->columns_index_map[mt->columns[i].column->index] = i;
}
}
}
void
ovsdb_monitor_add_jsonrpc_monitor(struct ovsdb_monitor *dbmon,
struct ovsdb_jsonrpc_monitor *jsonrpc_monitor)
{
struct jsonrpc_monitor_node *jm;
jm = xzalloc(sizeof *jm);
jm->jsonrpc_monitor = jsonrpc_monitor;
ovs_list_push_back(&dbmon->jsonrpc_monitors, &jm->node);
}
struct ovsdb_monitor *
ovsdb_monitor_create(struct ovsdb *db,
struct ovsdb_jsonrpc_monitor *jsonrpc_monitor)
{
struct ovsdb_monitor *dbmon;
dbmon = xzalloc(sizeof *dbmon);
ovsdb_replica_init(&dbmon->replica, &ovsdb_jsonrpc_replica_class);
ovsdb_add_replica(db, &dbmon->replica);
ovs_list_init(&dbmon->jsonrpc_monitors);
dbmon->db = db;
dbmon->n_transactions = 0;
shash_init(&dbmon->tables);
hmap_node_nullify(&dbmon->hmap_node);
hmap_init(&dbmon->json_cache);
ovsdb_monitor_add_jsonrpc_monitor(dbmon, jsonrpc_monitor);
return dbmon;
}
void
ovsdb_monitor_add_table(struct ovsdb_monitor *m,
const struct ovsdb_table *table)
{
struct ovsdb_monitor_table *mt;
int i;
size_t n_columns = shash_count(&table->schema->columns);
mt = xzalloc(sizeof *mt);
mt->table = table;
shash_add(&m->tables, table->schema->name, mt);
hmap_init(&mt->changes);
mt->columns_index_map =
xmalloc(sizeof *mt->columns_index_map * n_columns);
for (i = 0; i < n_columns; i++) {
mt->columns_index_map[i] = -1;
}
}
const char *
ovsdb_monitor_add_column(struct ovsdb_monitor *dbmon,
const struct ovsdb_table *table,
const struct ovsdb_column *column,
enum ovsdb_monitor_selection select,
bool monitored)
{
struct ovsdb_monitor_table *mt;
struct ovsdb_monitor_column *c;
mt = shash_find_data(&dbmon->tables, table->schema->name);
/* Check for column duplication. Return duplicated column name. */
if (mt->columns_index_map[column->index] != -1) {
return column->name;
}
if (mt->n_columns >= mt->allocated_columns) {
mt->columns = x2nrealloc(mt->columns, &mt->allocated_columns,
sizeof *mt->columns);
}
mt->select |= select;
mt->columns_index_map[column->index] = mt->n_columns;
c = &mt->columns[mt->n_columns++];
c->column = column;
c->select = select;
c->monitored = monitored;
if (monitored) {
mt->n_monitored_columns++;
}
return NULL;
}
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
static void
ovsdb_monitor_condition_add_columns(struct ovsdb_monitor *dbmon,
const struct ovsdb_table *table,
struct ovsdb_condition *condition)
{
size_t n_columns;
int i;
const struct ovsdb_column **columns =
ovsdb_condition_get_columns(condition, &n_columns);
for (i = 0; i < n_columns; i++) {
ovsdb_monitor_add_column(dbmon, table, columns[i],
OJMS_NONE, false);
}
free(columns);
}
/* Bind this session's condition to ovsdb_monitor */
void
ovsdb_monitor_condition_bind(struct ovsdb_monitor *dbmon,
struct ovsdb_monitor_session_condition *cond)
{
struct shash_node *node;
SHASH_FOR_EACH(node, &cond->tables) {
struct ovsdb_monitor_table_condition *mtc = node->data;
struct ovsdb_monitor_table *mt =
shash_find_data(&dbmon->tables, mtc->table->schema->name);
mtc->mt = mt;
ovsdb_monitor_condition_add_columns(dbmon, mtc->table,
&mtc->new_condition);
}
}
bool
ovsdb_monitor_table_exists(struct ovsdb_monitor *m,
const struct ovsdb_table *table)
{
return shash_find_data(&m->tables, table->schema->name);
}
static struct ovsdb_monitor_changes *
ovsdb_monitor_table_add_changes(struct ovsdb_monitor_table *mt,
uint64_t next_txn)
{
struct ovsdb_monitor_changes *changes;
changes = xzalloc(sizeof *changes);
changes->transaction = next_txn;
changes->mt = mt;
changes->n_refs = 1;
hmap_init(&changes->rows);
hmap_insert(&mt->changes, &changes->hmap_node, hash_uint64(next_txn));
return changes;
};
static struct ovsdb_monitor_changes *
ovsdb_monitor_table_find_changes(struct ovsdb_monitor_table *mt,
uint64_t transaction)
{
struct ovsdb_monitor_changes *changes;
size_t hash = hash_uint64(transaction);
HMAP_FOR_EACH_WITH_HASH(changes, hmap_node, hash, &mt->changes) {
if (changes->transaction == transaction) {
return changes;
}
}
return NULL;
}
/* Stop currently tracking changes to table 'mt' since 'transaction'. */
static void
ovsdb_monitor_table_untrack_changes(struct ovsdb_monitor_table *mt,
uint64_t transaction)
{
struct ovsdb_monitor_changes *changes =
ovsdb_monitor_table_find_changes(mt, transaction);
if (changes) {
if (--changes->n_refs == 0) {
hmap_remove(&mt->changes, &changes->hmap_node);
ovsdb_monitor_changes_destroy(changes);
}
}
}
/* Start tracking changes to table 'mt' begins from 'transaction' inclusive.
*/
static void
ovsdb_monitor_table_track_changes(struct ovsdb_monitor_table *mt,
uint64_t transaction)
{
struct ovsdb_monitor_changes *changes;
changes = ovsdb_monitor_table_find_changes(mt, transaction);
if (changes) {
changes->n_refs++;
} else {
ovsdb_monitor_table_add_changes(mt, transaction);
}
}
static void
ovsdb_monitor_changes_destroy(struct ovsdb_monitor_changes *changes)
{
struct ovsdb_monitor_row *row, *next;
HMAP_FOR_EACH_SAFE (row, next, hmap_node, &changes->rows) {
hmap_remove(&changes->rows, &row->hmap_node);
ovsdb_monitor_row_destroy(changes->mt, row);
}
hmap_destroy(&changes->rows);
free(changes);
}
static enum ovsdb_monitor_selection
ovsdb_monitor_row_update_type(bool initial, const bool old, const bool new)
{
return initial ? OJMS_INITIAL
: !old ? OJMS_INSERT
: !new ? OJMS_DELETE
: OJMS_MODIFY;
}
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
/* Set conditional monitoring mode only if we have non-empty condition in one
* of the tables at least */
static inline void
ovsdb_monitor_session_condition_set_mode(
struct ovsdb_monitor_session_condition *cond)
{
cond->conditional = shash_count(&cond->tables) !=
cond->n_true_cnd;
}
/* Returnes an empty allocated session's condition state holder */
struct ovsdb_monitor_session_condition *
ovsdb_monitor_session_condition_create(void)
{
struct ovsdb_monitor_session_condition *condition =
xzalloc(sizeof *condition);
condition->conditional = false;
shash_init(&condition->tables);
return condition;
}
void
ovsdb_monitor_session_condition_destroy(
struct ovsdb_monitor_session_condition *condition)
{
struct shash_node *node, *next;
if (!condition) {
return;
}
SHASH_FOR_EACH_SAFE (node, next, &condition->tables) {
struct ovsdb_monitor_table_condition *mtc = node->data;
ovsdb_condition_destroy(&mtc->new_condition);
ovsdb_condition_destroy(&mtc->old_condition);
shash_delete(&condition->tables, node);
free(mtc);
}
free(condition);
}
struct ovsdb_error *
ovsdb_monitor_table_condition_create(
struct ovsdb_monitor_session_condition *condition,
const struct ovsdb_table *table,
const struct json *json_cnd)
{
struct ovsdb_monitor_table_condition *mtc;
struct ovsdb_error *error;
mtc = xzalloc(sizeof *mtc);
mtc->table = table;
ovsdb_condition_init(&mtc->old_condition);
ovsdb_condition_init(&mtc->new_condition);
if (json_cnd) {
error = ovsdb_condition_from_json(table->schema,
json_cnd,
NULL,
&mtc->old_condition);
if (error) {
free(mtc);
return error;
}
}
shash_add(&condition->tables, table->schema->name, mtc);
/* On session startup old == new condition */
ovsdb_condition_clone(&mtc->new_condition, &mtc->old_condition);
if (ovsdb_condition_is_true(&mtc->old_condition)) {
condition->n_true_cnd++;
ovsdb_monitor_session_condition_set_mode(condition);
}
return NULL;
}
static bool
ovsdb_monitor_get_table_conditions(
const struct ovsdb_monitor_table *mt,
const struct ovsdb_monitor_session_condition *condition,
struct ovsdb_condition **old_condition,
struct ovsdb_condition **new_condition)
{
if (!condition) {
return false;
}
struct ovsdb_monitor_table_condition *mtc =
shash_find_data(&condition->tables, mt->table->schema->name);
if (!mtc) {
return false;
}
*old_condition = &mtc->old_condition;
*new_condition = &mtc->new_condition;
return true;
}
struct ovsdb_error *
ovsdb_monitor_table_condition_update(
struct ovsdb_monitor *dbmon,
struct ovsdb_monitor_session_condition *condition,
const struct ovsdb_table *table,
const struct json *cond_json)
{
struct ovsdb_monitor_table_condition *mtc =
shash_find_data(&condition->tables, table->schema->name);
struct ovsdb_error *error;
struct ovsdb_condition cond = OVSDB_CONDITION_INITIALIZER(&cond);
if (!condition) {
return NULL;
}
error = ovsdb_condition_from_json(table->schema, cond_json,
NULL, &cond);
if (error) {
return error;
}
ovsdb_condition_destroy(&mtc->new_condition);
ovsdb_condition_clone(&mtc->new_condition, &cond);
ovsdb_condition_destroy(&cond);
ovsdb_monitor_condition_add_columns(dbmon,
table,
&mtc->new_condition);
return NULL;
}
static void
ovsdb_monitor_table_condition_updated(struct ovsdb_monitor_table *mt,
struct ovsdb_monitor_session_condition *condition)
{
struct ovsdb_monitor_table_condition *mtc =
shash_find_data(&condition->tables, mt->table->schema->name);
if (mtc) {
/* If conditional monitoring - set old condition to new condition */
if (ovsdb_condition_cmp_3way(&mtc->old_condition,
&mtc->new_condition)) {
if (ovsdb_condition_is_true(&mtc->new_condition)) {
if (!ovsdb_condition_is_true(&mtc->old_condition)) {
condition->n_true_cnd++;
}
} else {
if (ovsdb_condition_is_true(&mtc->old_condition)) {
condition->n_true_cnd--;
}
}
ovsdb_condition_destroy(&mtc->old_condition);
ovsdb_condition_clone(&mtc->old_condition, &mtc->new_condition);
ovsdb_monitor_session_condition_set_mode(condition);
}
}
}
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
static enum ovsdb_monitor_selection
ovsdb_monitor_row_update_type_condition(
const struct ovsdb_monitor_table *mt,
const struct ovsdb_monitor_session_condition *condition,
bool initial,
enum ovsdb_monitor_row_type row_type,
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
const struct ovsdb_datum *old,
const struct ovsdb_datum *new)
{
struct ovsdb_condition *old_condition, *new_condition;
enum ovsdb_monitor_selection type =
ovsdb_monitor_row_update_type(initial, old, new);
if (ovsdb_monitor_get_table_conditions(mt,
condition,
&old_condition,
&new_condition)) {
bool old_cond = !old ? false
: ovsdb_condition_empty_or_match_any(old,
old_condition,
row_type == OVSDB_MONITOR_ROW ?
mt->columns_index_map :
NULL);
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
bool new_cond = !new ? false
: ovsdb_condition_empty_or_match_any(new,
new_condition,
row_type == OVSDB_MONITOR_ROW ?
mt->columns_index_map :
NULL);
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
if (!old_cond && !new_cond) {
type = OJMS_NONE;
}
switch (type) {
case OJMS_INITIAL:
case OJMS_INSERT:
if (!new_cond) {
type = OJMS_NONE;
}
break;
case OJMS_MODIFY:
type = !old_cond ? OJMS_INSERT : !new_cond
? OJMS_DELETE : OJMS_MODIFY;
break;
case OJMS_DELETE:
if (!old_cond) {
type = OJMS_NONE;
}
break;
case OJMS_NONE:
break;
}
}
return type;
}
static bool
ovsdb_monitor_row_skip_update(const struct ovsdb_monitor_table *mt,
enum ovsdb_monitor_row_type row_type,
const struct ovsdb_datum *old,
const struct ovsdb_datum *new,
enum ovsdb_monitor_selection type,
unsigned long int *changed)
{
if (!(mt->select & type)) {
return true;
}
if (type == OJMS_MODIFY) {
size_t i, n_changes;
n_changes = 0;
memset(changed, 0, bitmap_n_bytes(mt->n_columns));
for (i = 0; i < mt->n_columns; i++) {
const struct ovsdb_column *c = mt->columns[i].column;
size_t index = row_type == OVSDB_ROW ? c->index : i;
if (!ovsdb_datum_equals(&old[index], &new[index], &c->type)) {
bitmap_set1(changed, i);
n_changes++;
}
}
if (!n_changes) {
/* No actual changes: presumably a row changed and then
* changed back later. */
return true;
}
}
return false;
}
/* Returns JSON for a <row-update> (as described in RFC 7047) for 'row' within
* 'mt', or NULL if no row update should be sent.
*
* The caller should specify 'initial' as true if the returned JSON is going to
* be used as part of the initial reply to a "monitor" request, false if it is
* going to be used as part of an "update" notification.
*
* 'changed' must be a scratch buffer for internal use that is at least
* bitmap_n_bytes(mt->n_columns) bytes long. */
static struct json *
ovsdb_monitor_compose_row_update(
const struct ovsdb_monitor_table *mt,
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
const struct ovsdb_monitor_session_condition *condition OVS_UNUSED,
enum ovsdb_monitor_row_type row_type OVS_UNUSED,
const void *_row,
bool initial, unsigned long int *changed)
{
const struct ovsdb_monitor_row *row = _row;
enum ovsdb_monitor_selection type;
struct json *old_json, *new_json;
struct json *row_json;
size_t i;
ovs_assert(row_type == OVSDB_MONITOR_ROW);
type = ovsdb_monitor_row_update_type(initial, row->old, row->new);
if (ovsdb_monitor_row_skip_update(mt, row_type, row->old,
row->new, type, changed)) {
return NULL;
}
row_json = json_object_create();
old_json = new_json = NULL;
if (type & (OJMS_DELETE | OJMS_MODIFY)) {
old_json = json_object_create();
json_object_put(row_json, "old", old_json);
}
if (type & (OJMS_INITIAL | OJMS_INSERT | OJMS_MODIFY)) {
new_json = json_object_create();
json_object_put(row_json, "new", new_json);
}
for (i = 0; i < mt->n_monitored_columns; i++) {
const struct ovsdb_monitor_column *c = &mt->columns[i];
if (!c->monitored || !(type & c->select)) {
/* We don't care about this type of change for this
* particular column (but we will care about it for some
* other column). */
continue;
}
if ((type == OJMS_MODIFY && bitmap_is_set(changed, i))
|| type == OJMS_DELETE) {
json_object_put(old_json, c->column->name,
ovsdb_datum_to_json(&row->old[i],
&c->column->type));
}
if (type & (OJMS_INITIAL | OJMS_INSERT | OJMS_MODIFY)) {
json_object_put(new_json, c->column->name,
ovsdb_datum_to_json(&row->new[i],
&c->column->type));
}
}
return row_json;
}
/* Returns JSON for a <row-update2> (as described in ovsdb-server(1) mapage)
* for 'row' within * 'mt', or NULL if no row update should be sent.
*
* The caller should specify 'initial' as true if the returned JSON is
* going to be used as part of the initial reply to a "monitor_cond" request,
* false if it is going to be used as part of an "update2" notification.
*
* 'changed' must be a scratch buffer for internal use that is at least
* bitmap_n_bytes(mt->n_columns) bytes long. */
static struct json *
ovsdb_monitor_compose_row_update2(
const struct ovsdb_monitor_table *mt,
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
const struct ovsdb_monitor_session_condition *condition,
enum ovsdb_monitor_row_type row_type,
const void *_row,
bool initial, unsigned long int *changed)
{
enum ovsdb_monitor_selection type;
struct json *row_update2, *diff_json;
const struct ovsdb_datum *old, *new;
size_t i;
if (row_type == OVSDB_MONITOR_ROW) {
old = ((const struct ovsdb_monitor_row *)_row)->old;;
new = ((const struct ovsdb_monitor_row *)_row)->new;
} else {
old = new = ((const struct ovsdb_row *)_row)->fields;
}
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
type = ovsdb_monitor_row_update_type_condition(mt, condition, initial,
row_type, old, new);
if (ovsdb_monitor_row_skip_update(mt, row_type, old, new, type, changed)) {
return NULL;
}
row_update2 = json_object_create();
if (type == OJMS_DELETE) {
json_object_put(row_update2, "delete", json_null_create());
} else {
diff_json = json_object_create();
const char *op;
for (i = 0; i < mt->n_monitored_columns; i++) {
const struct ovsdb_monitor_column *c = &mt->columns[i];
size_t index = row_type == OVSDB_ROW ? c->column->index : i;
if (!c->monitored || !(type & c->select)) {
/* We don't care about this type of change for this
* particular column (but we will care about it for some
* other column). */
continue;
}
if (type == OJMS_MODIFY) {
struct ovsdb_datum diff;
if (!bitmap_is_set(changed, i)) {
continue;
}
ovsdb_datum_diff(&diff ,&old[index], &new[index],
&c->column->type);
json_object_put(diff_json, c->column->name,
ovsdb_datum_to_json(&diff, &c->column->type));
ovsdb_datum_destroy(&diff, &c->column->type);
} else {
if (!ovsdb_datum_is_default(&new[index], &c->column->type)) {
json_object_put(diff_json, c->column->name,
ovsdb_datum_to_json(&new[index],
&c->column->type));
}
}
}
op = type == OJMS_INITIAL ? "initial"
: type == OJMS_MODIFY ? "modify" : "insert";
json_object_put(row_update2, op, diff_json);
}
return row_update2;
}
static size_t
ovsdb_monitor_max_columns(struct ovsdb_monitor *dbmon)
{
struct shash_node *node;
size_t max_columns = 0;
SHASH_FOR_EACH (node, &dbmon->tables) {
struct ovsdb_monitor_table *mt = node->data;
max_columns = MAX(max_columns, mt->n_columns);
}
return max_columns;
}
static void
ovsdb_monitor_add_json_row(struct json **json, const char *table_name,
struct json **table_json, struct json *row_json,
const struct uuid *row_uuid)
{
char uuid[UUID_LEN + 1];
/* Create JSON object for transaction overall. */
if (!*json) {
*json = json_object_create();
}
/* Create JSON object for transaction on this table. */
if (!*table_json) {
*table_json = json_object_create();
json_object_put(*json, table_name, *table_json);
}
/* Add JSON row to JSON table. */
snprintf(uuid, sizeof uuid, UUID_FMT, UUID_ARGS(row_uuid));
json_object_put(*table_json, uuid, row_json);
}
/* Constructs and returns JSON for a <table-updates> object (as described in
* RFC 7047) for all the outstanding changes within 'monitor', starting from
* 'transaction'. */
static struct json*
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
ovsdb_monitor_compose_update(
struct ovsdb_monitor *dbmon,
bool initial, uint64_t transaction,
const struct ovsdb_monitor_session_condition *condition,
compose_row_update_cb_func row_update)
{
struct shash_node *node;
struct json *json;
size_t max_columns = ovsdb_monitor_max_columns(dbmon);
unsigned long int *changed = xmalloc(bitmap_n_bytes(max_columns));
json = NULL;
SHASH_FOR_EACH (node, &dbmon->tables) {
struct ovsdb_monitor_table *mt = node->data;
struct ovsdb_monitor_row *row, *next;
struct ovsdb_monitor_changes *changes;
struct json *table_json = NULL;
changes = ovsdb_monitor_table_find_changes(mt, transaction);
if (!changes) {
continue;
}
HMAP_FOR_EACH_SAFE (row, next, hmap_node, &changes->rows) {
struct json *row_json;
row_json = (*row_update)(mt, condition, OVSDB_MONITOR_ROW, row,
initial, changed);
if (row_json) {
ovsdb_monitor_add_json_row(&json, mt->table->schema->name,
&table_json, row_json,
&row->uuid);
}
}
}
free(changed);
return json;
}
static struct json*
ovsdb_monitor_compose_cond_change_update(
struct ovsdb_monitor *dbmon,
struct ovsdb_monitor_session_condition *condition)
{
struct shash_node *node;
struct json *json = NULL;
size_t max_columns = ovsdb_monitor_max_columns(dbmon);
unsigned long int *changed = xmalloc(bitmap_n_bytes(max_columns));
SHASH_FOR_EACH (node, &dbmon->tables) {
struct ovsdb_monitor_table *mt = node->data;
struct ovsdb_row *row;
struct json *table_json = NULL;
struct ovsdb_condition *old_condition, *new_condition;
if (!ovsdb_monitor_get_table_conditions(mt,
condition,
&old_condition,
&new_condition) ||
!ovsdb_condition_cmp_3way(old_condition, new_condition)) {
/* Nothing to update on this table */
continue;
}
/* Iterate over all rows in table */
HMAP_FOR_EACH (row, hmap_node, &mt->table->rows) {
struct json *row_json;
row_json = ovsdb_monitor_compose_row_update2(mt, condition,
OVSDB_ROW, row,
false, changed);
if (row_json) {
ovsdb_monitor_add_json_row(&json, mt->table->schema->name,
&table_json, row_json,
ovsdb_row_get_uuid(row));
}
}
ovsdb_monitor_table_condition_updated(mt, condition);
}
free(changed);
return json;
}
/* Returns JSON for a <table-updates> object (as described in RFC 7047)
* for all the outstanding changes within 'monitor' that starts from
* '*unflushed'.
* If cond_updated is true all rows in the db that match conditions will be
* sent.
*
* The caller should specify 'initial' as true if the returned JSON is going to
* be used as part of the initial reply to a "monitor" request, false if it is
* going to be used as part of an "update" notification. */
struct json *
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
ovsdb_monitor_get_update(
struct ovsdb_monitor *dbmon,
bool initial, bool cond_updated,
uint64_t *unflushed_,
struct ovsdb_monitor_session_condition *condition,
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
enum ovsdb_monitor_version version)
{
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
struct ovsdb_monitor_json_cache_node *cache_node = NULL;
struct shash_node *node;
struct json *json;
const uint64_t unflushed = *unflushed_;
const uint64_t next_unflushed = dbmon->n_transactions + 1;
ovs_assert(cond_updated ? unflushed == next_unflushed : true);
/* Return a clone of cached json if one exists. Otherwise,
* generate a new one and add it to the cache. */
if (!condition || (!condition->conditional && !cond_updated)) {
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
cache_node = ovsdb_monitor_json_cache_search(dbmon, version,
unflushed);
}
if (cache_node) {
json = cache_node->json ? json_clone(cache_node->json) : NULL;
} else {
if (version == OVSDB_MONITOR_V1) {
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
json =
ovsdb_monitor_compose_update(dbmon, initial, unflushed,
condition,
ovsdb_monitor_compose_row_update);
} else {
ovs_assert(version == OVSDB_MONITOR_V2);
if (!cond_updated) {
json = ovsdb_monitor_compose_update(dbmon, initial, unflushed,
condition,
ovsdb_monitor_compose_row_update2);
if (!condition || !condition->conditional) {
ovsdb_monitor_json_cache_insert(dbmon, version, unflushed,
json);
}
} else {
/* Compose update on whole db due to condition update.
Session must be flushed (change list is empty)*/
json =
ovsdb_monitor_compose_cond_change_update(dbmon, condition);
}
}
}
/* Maintain transaction id of 'changes'. */
SHASH_FOR_EACH (node, &dbmon->tables) {
struct ovsdb_monitor_table *mt = node->data;
ovsdb_monitor_table_untrack_changes(mt, unflushed);
ovsdb_monitor_table_track_changes(mt, next_unflushed);
}
*unflushed_ = next_unflushed;
return json;
}
bool
ovsdb_monitor_needs_flush(struct ovsdb_monitor *dbmon,
uint64_t next_transaction)
{
ovs_assert(next_transaction <= dbmon->n_transactions + 1);
return (next_transaction <= dbmon->n_transactions);
}
void
ovsdb_monitor_table_add_select(struct ovsdb_monitor *dbmon,
const struct ovsdb_table *table,
enum ovsdb_monitor_selection select)
{
struct ovsdb_monitor_table * mt;
mt = shash_find_data(&dbmon->tables, table->schema->name);
mt->select |= select;
}
/*
* If a row's change type (insert, delete or modify) matches that of
* the monitor, they should be sent to the monitor's clients as updates.
* Of cause, the monitor should also internally update with this change.
*
* When a change type does not require client side update, the monitor
* may still need to keep track of certain changes in order to generate
* correct future updates. For example, the monitor internal state should
* be updated whenever a new row is inserted, in order to generate the
* correct initial state, regardless if a insert change type is being
* monitored.
*
* On the other hand, if a transaction only contains changes to columns
* that are not monitored, this transaction can be safely ignored by the
* monitor.
*
* Thus, the order of the declaration is important:
* 'OVSDB_CHANGES_REQUIRE_EXTERNAL_UPDATE' always implies
* 'OVSDB_CHANGES_REQUIRE_INTERNAL_UPDATE', but not vice versa. */
enum ovsdb_monitor_changes_efficacy {
OVSDB_CHANGES_NO_EFFECT, /* Monitor does not care about this
change. */
OVSDB_CHANGES_REQUIRE_INTERNAL_UPDATE, /* Monitor internal updates. */
OVSDB_CHANGES_REQUIRE_EXTERNAL_UPDATE, /* Client needs to be updated. */
};
struct ovsdb_monitor_aux {
const struct ovsdb_monitor *monitor;
struct ovsdb_monitor_table *mt;
enum ovsdb_monitor_changes_efficacy efficacy;
};
static void
ovsdb_monitor_init_aux(struct ovsdb_monitor_aux *aux,
const struct ovsdb_monitor *m)
{
aux->monitor = m;
aux->mt = NULL;
aux->efficacy = OVSDB_CHANGES_NO_EFFECT;
}
static void
ovsdb_monitor_changes_update(const struct ovsdb_row *old,
const struct ovsdb_row *new,
const struct ovsdb_monitor_table *mt,
struct ovsdb_monitor_changes *changes)
{
const struct uuid *uuid = ovsdb_row_get_uuid(new ? new : old);
struct ovsdb_monitor_row *change;
change = ovsdb_monitor_changes_row_find(changes, uuid);
if (!change) {
change = xzalloc(sizeof *change);
hmap_insert(&changes->rows, &change->hmap_node, uuid_hash(uuid));
change->uuid = *uuid;
change->old = clone_monitor_row_data(mt, old);
change->new = clone_monitor_row_data(mt, new);
} else {
if (new) {
update_monitor_row_data(mt, new, change->new);
} else {
free_monitor_row_data(mt, change->new);
change->new = NULL;
if (!change->old) {
/* This row was added then deleted. Forget about it. */
hmap_remove(&changes->rows, &change->hmap_node);
free(change);
}
}
}
}
static bool
ovsdb_monitor_columns_changed(const struct ovsdb_monitor_table *mt,
const unsigned long int *changed)
{
size_t i;
for (i = 0; i < mt->n_columns; i++) {
size_t column_index = mt->columns[i].column->index;
if (bitmap_is_set(changed, column_index)) {
return true;
}
}
return false;
}
/* Return the efficacy of a row's change to a monitor table.
*
* Please see the block comment above 'ovsdb_monitor_changes_efficacy'
* definition form more information. */
static enum ovsdb_monitor_changes_efficacy
ovsdb_monitor_changes_classify(enum ovsdb_monitor_selection type,
const struct ovsdb_monitor_table *mt,
const unsigned long int *changed)
{
if (type == OJMS_MODIFY &&
!ovsdb_monitor_columns_changed(mt, changed)) {
return OVSDB_CHANGES_NO_EFFECT;
}
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
if (type == OJMS_MODIFY) {
/* Condition might turn a modify operation to insert or delete */
type |= OJMS_INSERT | OJMS_DELETE;
}
return (mt->select & type)
? OVSDB_CHANGES_REQUIRE_EXTERNAL_UPDATE
: OVSDB_CHANGES_REQUIRE_INTERNAL_UPDATE;
}
static bool
ovsdb_monitor_change_cb(const struct ovsdb_row *old,
const struct ovsdb_row *new,
const unsigned long int *changed,
void *aux_)
{
struct ovsdb_monitor_aux *aux = aux_;
const struct ovsdb_monitor *m = aux->monitor;
struct ovsdb_table *table = new ? new->table : old->table;
struct ovsdb_monitor_table *mt;
struct ovsdb_monitor_changes *changes;
if (!aux->mt || table != aux->mt->table) {
aux->mt = shash_find_data(&m->tables, table->schema->name);
if (!aux->mt) {
/* We don't care about rows in this table at all. Tell the caller
* to skip it. */
return false;
}
}
mt = aux->mt;
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
enum ovsdb_monitor_selection type =
ovsdb_monitor_row_update_type(false, old, new);
enum ovsdb_monitor_changes_efficacy efficacy =
ovsdb_monitor_changes_classify(type, mt, changed);
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
HMAP_FOR_EACH(changes, hmap_node, &mt->changes) {
if (efficacy > OVSDB_CHANGES_NO_EFFECT) {
ovsdb_monitor_changes_update(old, new, mt, changes);
}
ovsdb: generate update notifications for monitor_cond session Hold session's conditions in ovsdb_monitor_session_condition. Pass it to ovsdb_monitor for generating "update2" notifications. Add functions that can generate "update2" notification for a "monitor_cond" session. JSON cache is enabled only for session's with true condition only. "monitor_cond" and "monitor_cond_change" are RFC 7047 extensions described by ovsdb-server(1) manpage. Performance evaluation: OVN is the main candidate for conditional monitoring usage. It is clear that conditional monitoring reduces computation on the ovn-controller (client) side due to the reduced size of flow tables and update messages. Performance evaluation shows up to 75% computation reduction. However, performance evaluation shows also a reduction in computation on the SB ovsdb-server side proportional to the degree that each logical network is spread over physical hosts in the DC. Evaluation shows that in a realistic scenarios there is a computation reduction also in the server side. Evaluation on simulated environment of 50 hosts and 1000 logical ports shows the following results (cycles #): LN spread over # hosts| master | patch | change ------------------------------------------------------------- 1 | 24597200127 | 24339235374 | 1.0% 6 | 23788521572 | 19145229352 | 19.5% 12 | 23886405758 | 17913143176 | 25.0% 18 | 25812686279 | 23675094540 | 8.2% 24 | 28414671499 | 24770202308 | 12.8% 30 | 31487218890 | 28397543436 | 9.8% 36 | 36116993930 | 34105388739 | 5.5% 42 | 37898342465 | 38647139083 | -1.9% 48 | 41637996229 | 41846616306 | -0.5% 50 | 41679995357 | 43455565977 | -4.2% Signed-off-by: Liran Schour <lirans@il.ibm.com> Signed-off-by: Ben Pfaff <blp@ovn.org>
2016-07-18 11:45:52 +03:00
}
if (aux->efficacy < efficacy) {
aux->efficacy = efficacy;
}
return true;
}
void
ovsdb_monitor_get_initial(const struct ovsdb_monitor *dbmon)
{
struct shash_node *node;
SHASH_FOR_EACH (node, &dbmon->tables) {
struct ovsdb_monitor_table *mt = node->data;
if (mt->select & OJMS_INITIAL) {
struct ovsdb_row *row;
struct ovsdb_monitor_changes *changes;
changes = ovsdb_monitor_table_find_changes(mt, 0);
if (!changes) {
changes = ovsdb_monitor_table_add_changes(mt, 0);
HMAP_FOR_EACH (row, hmap_node, &mt->table->rows) {
ovsdb_monitor_changes_update(NULL, row, mt, changes);
}
} else {
changes->n_refs++;
}
}
}
}
void
ovsdb_monitor_remove_jsonrpc_monitor(struct ovsdb_monitor *dbmon,
struct ovsdb_jsonrpc_monitor *jsonrpc_monitor,
uint64_t unflushed)
{
struct jsonrpc_monitor_node *jm;
if (ovs_list_is_empty(&dbmon->jsonrpc_monitors)) {
ovsdb_monitor_destroy(dbmon);
return;
}
/* Find and remove the jsonrpc monitor from the list. */
LIST_FOR_EACH(jm, node, &dbmon->jsonrpc_monitors) {
if (jm->jsonrpc_monitor == jsonrpc_monitor) {
/* Release the tracked changes. */
struct shash_node *node;
SHASH_FOR_EACH (node, &dbmon->tables) {
struct ovsdb_monitor_table *mt = node->data;
ovsdb_monitor_table_untrack_changes(mt, unflushed);
}
ovs_list_remove(&jm->node);
free(jm);
/* Destroy ovsdb monitor if this is the last user. */
if (ovs_list_is_empty(&dbmon->jsonrpc_monitors)) {
ovsdb_monitor_destroy(dbmon);
}
return;
};
}
/* Should never reach here. jsonrpc_monitor should be on the list. */
OVS_NOT_REACHED();
}
static bool
ovsdb_monitor_table_equal(const struct ovsdb_monitor_table *a,
const struct ovsdb_monitor_table *b)
{
size_t i;
ovs_assert(b->n_columns == b->n_monitored_columns);
if ((a->table != b->table) ||
(a->select != b->select) ||
(a->n_monitored_columns != b->n_monitored_columns)) {
return false;
}
/* Compare only monitored columns that must be sorted already */
for (i = 0; i < a->n_monitored_columns; i++) {
if ((a->columns[i].column != b->columns[i].column) ||
(a->columns[i].select != b->columns[i].select)) {
return false;
}
}
return true;
}
static bool
ovsdb_monitor_equal(const struct ovsdb_monitor *a,
const struct ovsdb_monitor *b)
{
struct shash_node *node;
if (shash_count(&a->tables) != shash_count(&b->tables)) {
return false;
}
SHASH_FOR_EACH(node, &a->tables) {
const struct ovsdb_monitor_table *mta = node->data;
const struct ovsdb_monitor_table *mtb;
mtb = shash_find_data(&b->tables, node->name);
if (!mtb) {
return false;
}
if (!ovsdb_monitor_table_equal(mta, mtb)) {
return false;
}
}
return true;
}
static size_t
ovsdb_monitor_hash(const struct ovsdb_monitor *dbmon, size_t basis)
{
const struct shash_node **nodes;
size_t i, j, n;
nodes = shash_sort(&dbmon->tables);
n = shash_count(&dbmon->tables);
for (i = 0; i < n; i++) {
struct ovsdb_monitor_table *mt = nodes[i]->data;
basis = hash_pointer(mt->table, basis);
basis = hash_3words(mt->select, mt->n_columns, basis);
for (j = 0; j < mt->n_columns; j++) {
basis = hash_pointer(mt->columns[j].column, basis);
basis = hash_2words(mt->columns[j].select, basis);
}
}
free(nodes);
return basis;
}
struct ovsdb_monitor *
ovsdb_monitor_add(struct ovsdb_monitor *new_dbmon)
{
struct ovsdb_monitor *dbmon;
size_t hash;
/* New_dbmon should be associated with only one jsonrpc
* connections. */
ovs_assert(ovs_list_is_singleton(&new_dbmon->jsonrpc_monitors));
ovsdb_monitor_columns_sort(new_dbmon);
hash = ovsdb_monitor_hash(new_dbmon, 0);
HMAP_FOR_EACH_WITH_HASH(dbmon, hmap_node, hash, &ovsdb_monitors) {
if (ovsdb_monitor_equal(dbmon, new_dbmon)) {
return dbmon;
}
}
hmap_insert(&ovsdb_monitors, &new_dbmon->hmap_node, hash);
return new_dbmon;
}
static void
ovsdb_monitor_destroy(struct ovsdb_monitor *dbmon)
{
struct shash_node *node;
ovs_list_remove(&dbmon->replica.node);
if (!hmap_node_is_null(&dbmon->hmap_node)) {
hmap_remove(&ovsdb_monitors, &dbmon->hmap_node);
}
ovsdb_monitor_json_cache_flush(dbmon);
hmap_destroy(&dbmon->json_cache);
SHASH_FOR_EACH (node, &dbmon->tables) {
struct ovsdb_monitor_table *mt = node->data;
struct ovsdb_monitor_changes *changes, *next;
HMAP_FOR_EACH_SAFE (changes, next, hmap_node, &mt->changes) {
hmap_remove(&mt->changes, &changes->hmap_node);
ovsdb_monitor_changes_destroy(changes);
}
hmap_destroy(&mt->changes);
free(mt->columns);
free(mt->columns_index_map);
free(mt);
}
shash_destroy(&dbmon->tables);
free(dbmon);
}
static struct ovsdb_error *
ovsdb_monitor_commit(struct ovsdb_replica *replica,
const struct ovsdb_txn *txn,
bool durable OVS_UNUSED)
{
struct ovsdb_monitor *m = ovsdb_monitor_cast(replica);
struct ovsdb_monitor_aux aux;
ovsdb_monitor_init_aux(&aux, m);
/* Update ovsdb_monitor's transaction number for
* each transaction, before calling ovsdb_monitor_change_cb(). */
m->n_transactions++;
ovsdb_txn_for_each_change(txn, ovsdb_monitor_change_cb, &aux);
switch(aux.efficacy) {
case OVSDB_CHANGES_NO_EFFECT:
/* The transaction is ignored by the monitor.
* Roll back the 'n_transactions' as if the transaction
* has never happened. */
m->n_transactions--;
break;
case OVSDB_CHANGES_REQUIRE_INTERNAL_UPDATE:
/* Nothing. */
break;
case OVSDB_CHANGES_REQUIRE_EXTERNAL_UPDATE:
ovsdb_monitor_json_cache_flush(m);
break;
}
return NULL;
}
static void
ovsdb_monitor_destroy_callback(struct ovsdb_replica *replica)
{
struct ovsdb_monitor *dbmon = ovsdb_monitor_cast(replica);
struct jsonrpc_monitor_node *jm, *next;
/* Delete all front end monitors. Removing the last front
* end monitor will also destroy the corresponding 'ovsdb_monitor'.
* ovsdb monitor will also be destroied. */
LIST_FOR_EACH_SAFE(jm, next, node, &dbmon->jsonrpc_monitors) {
ovsdb_jsonrpc_monitor_destroy(jm->jsonrpc_monitor);
}
}
/* Add some memory usage statics for monitors into 'usage', for use with
* memory_report(). */
void
ovsdb_monitor_get_memory_usage(struct simap *usage)
{
struct ovsdb_monitor *dbmon;
simap_put(usage, "monitors", hmap_count(&ovsdb_monitors));
HMAP_FOR_EACH(dbmon, hmap_node, &ovsdb_monitors) {
simap_increase(usage, "json-caches", hmap_count(&dbmon->json_cache));
}
}
static const struct ovsdb_replica_class ovsdb_jsonrpc_replica_class = {
ovsdb_monitor_commit,
ovsdb_monitor_destroy_callback,
};