[
https://issues.apache.org/jira/browse/IGNITE-12087?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alexey Goncharuk updated IGNITE-12087:
--------------------------------------
Description:
After IGNITE-5227 have been fixed I found significant performance drop in
putAll operation.
Insertion of 30_000 entries before IGNITE-5227 took ~1 second.
After IGNITE-5227 - 130 seconds (~100x slower).
I checked a different batch size:
10_000 - 10 seconds
20_000 - 48 seconds
30_000 - 130 seconds
and I was not able to wait for the result of 100_000 entries.
Reproducer:
{code:java}
public class CheckPutAll extends GridCommonAbstractTest {
@Override protected IgniteConfiguration getConfiguration(String
igniteInstanceName) throws Exception {
IgniteConfiguration cfg = super.getConfiguration(igniteInstanceName);
CacheConfiguration ccfg = new CacheConfiguration(DEFAULT_CACHE_NAME);
ccfg.setAtomicityMode(TRANSACTIONAL);
cfg.setCacheConfiguration(ccfg);
return cfg;
}
@Test
public void check() throws Exception {
int cnt = 30_000;
Map<Integer, Integer> data = new HashMap<>(U.capacity(cnt));
for (int i = 0; i < cnt; i++)
data.put(i, i);
Ignite node0 = startGrid(0);
IgniteCache<Integer, Integer> cache0 = node0.cache(DEFAULT_CACHE_NAME);
cache0.putAll(data);
}
}{code}
was:
After IGNITE-5227 have been fixed I found significant performance drop in
putAll operation.
Insertion of 30_000 entries before IGNITE-5227 took ~1 second.
After IGNITE-5227 - 130 seconds (~100x slower).
I checked a different batch size:
10_000 - 10 seconds
20_000 - 48 seconds
30_000 - 130 seconds
and I was not able to wait for the result of 100_000 entries.
Reproducer
{code:java}
public class CheckPutAll extends GridCommonAbstractTest {
@Override protected IgniteConfiguration getConfiguration(String
igniteInstanceName) throws Exception {
IgniteConfiguration cfg = super.getConfiguration(igniteInstanceName);
CacheConfiguration ccfg = new CacheConfiguration(DEFAULT_CACHE_NAME);
ccfg.setAtomicityMode(TRANSACTIONAL);
cfg.setCacheConfiguration(ccfg);
return cfg;
}
@Test
public void check() throws Exception {
int cnt = 30_000;
Map<Integer, Integer> data = new HashMap<>(U.capacity(cnt));
for (int i = 0; i < cnt; i++)
data.put(i, i);
Ignite node0 = startGrid(0);
IgniteCache<Integer, Integer> cache0 = node0.cache(DEFAULT_CACHE_NAME);
cache0.putAll(data);
}
}{code}
> Transactional putAll - significant performance drop on big batches of entries.
> ------------------------------------------------------------------------------
>
> Key: IGNITE-12087
> URL: https://issues.apache.org/jira/browse/IGNITE-12087
> Project: Ignite
> Issue Type: Bug
> Components: cache
> Reporter: Pavel Pereslegin
> Assignee: Eduard Shangareev
> Priority: Major
> Fix For: 2.8
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> After IGNITE-5227 have been fixed I found significant performance drop in
> putAll operation.
> Insertion of 30_000 entries before IGNITE-5227 took ~1 second.
> After IGNITE-5227 - 130 seconds (~100x slower).
> I checked a different batch size:
> 10_000 - 10 seconds
> 20_000 - 48 seconds
> 30_000 - 130 seconds
> and I was not able to wait for the result of 100_000 entries.
> Reproducer:
> {code:java}
> public class CheckPutAll extends GridCommonAbstractTest {
> @Override protected IgniteConfiguration getConfiguration(String
> igniteInstanceName) throws Exception {
> IgniteConfiguration cfg = super.getConfiguration(igniteInstanceName);
> CacheConfiguration ccfg = new CacheConfiguration(DEFAULT_CACHE_NAME);
> ccfg.setAtomicityMode(TRANSACTIONAL);
> cfg.setCacheConfiguration(ccfg);
> return cfg;
> }
> @Test
> public void check() throws Exception {
> int cnt = 30_000;
> Map<Integer, Integer> data = new HashMap<>(U.capacity(cnt));
> for (int i = 0; i < cnt; i++)
> data.put(i, i);
> Ignite node0 = startGrid(0);
> IgniteCache<Integer, Integer> cache0 =
> node0.cache(DEFAULT_CACHE_NAME);
> cache0.putAll(data);
> }
> }{code}
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
