[
https://issues.apache.org/jira/browse/IGNITE-7182?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ivan Rakov updated IGNITE-7182:
-------------------------------
Description:
Tests show that GridCacheDatabaseSharedManager#splitAndSortCpPagesIfNeeded call
can last several seconds on nodes with big amount of memory (>10GB). We should
optimize sorting algorithm, possibly making it multithreaded.
Another option to make pages write throttling more smooth is to get rid of this
heuristic:
{noformat}
// Starting with 0.05 to avoid throttle right after checkpoint
start
// 7/12 is maximum ratio of dirty pages
dirtyRatioThreshold = (dirtyRatioThreshold * 0.95 + 0.05) * 7 /
12;
{noformat}
We should replace "magic" lower bound 0.05 * 7 / 12 with the real percentage of
dirty pages at the moment of
GridCacheDatabaseSharedManager.Checkpointer#markCheckpointBegin call return.
was:
Tests show that GridCacheDatabaseSharedManager#splitAndSortCpPagesIfNeeded call
can last several seconds on nodes with big amount af memory (>10GB). We should
optimize sorting algorithm, possibly making it multithreaded.
Another option to make pages write throttling more smooth is to get rid of this
heuristic:
{noformat}
// Starting with 0.05 to avoid throttle right after checkpoint
start
// 7/12 is maximum ratio of dirty pages
dirtyRatioThreshold = (dirtyRatioThreshold * 0.95 + 0.05) * 7 /
12;
{noformat}
We should replace "magic" lower bound 0.05 * 7 / 12 with the real percentage of
dirty pages at the moment of
GridCacheDatabaseSharedManager.Checkpointer#markCheckpointBegin call return.
> Slow sorting of pages collection on checkpoint begin can cause zero dropdown
> even with throttling enabled
> ---------------------------------------------------------------------------------------------------------
>
> Key: IGNITE-7182
> URL: https://issues.apache.org/jira/browse/IGNITE-7182
> Project: Ignite
> Issue Type: Improvement
> Components: persistence
> Affects Versions: 2.3
> Reporter: Ivan Rakov
> Assignee: Dmitriy Pavlov
> Fix For: 2.4
>
>
> Tests show that GridCacheDatabaseSharedManager#splitAndSortCpPagesIfNeeded
> call can last several seconds on nodes with big amount of memory (>10GB). We
> should optimize sorting algorithm, possibly making it multithreaded.
> Another option to make pages write throttling more smooth is to get rid of
> this heuristic:
> {noformat}
> // Starting with 0.05 to avoid throttle right after
> checkpoint start
> // 7/12 is maximum ratio of dirty pages
> dirtyRatioThreshold = (dirtyRatioThreshold * 0.95 + 0.05) * 7
> / 12;
> {noformat}
> We should replace "magic" lower bound 0.05 * 7 / 12 with the real percentage
> of dirty pages at the moment of
> GridCacheDatabaseSharedManager.Checkpointer#markCheckpointBegin call return.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
