[
https://issues.apache.org/jira/browse/HBASE-19357?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16281944#comment-16281944
]
Anoop Sam John commented on HBASE-19357:
----------------------------------------
Did the tests comparing On heap LRU vs file mode BC.
Again on a single server. Total 25 GB data. On heap test having 42 GB xmx and
60% and cache. File mode having the 25 GB cache in file (on disk). The total
RAM in server is 140 GB means the whole date can be accommodated in OS cache.
File mode cache with META and NS table data still on on heap LRU (with out this
patch) is have a lesser QPS (as expected) compared to LRU cache. This is ~19%
less avg throughput. All data in file BC (META and NS table data also as in
this patch) is almost having almost same QPS . Just 0.9% less throughput
compared to META and NS data in LRU cache case. This might be a change from run
to run. Not just OS cache the factor.. The META entries are cached at
connection level any way. In a real case also chances of hitting the META again
and again is very less. So I am no longer having the concern of META data NOT
in LRU cache even when BC file mode is in use. Sounds good abt the conclusion
boss?
> Bucket cache no longer L2 for LRU cache
> ---------------------------------------
>
> Key: HBASE-19357
> URL: https://issues.apache.org/jira/browse/HBASE-19357
> Project: HBase
> Issue Type: Sub-task
> Reporter: Anoop Sam John
> Assignee: Anoop Sam John
> Fix For: 2.0.0-beta-1
>
> Attachments: HBASE-19357.patch, HBASE-19357.patch,
> HBASE-19357_V2.patch, HBASE-19357_V3.patch, HBASE-19357_V3.patch
>
>
> When Bucket cache is used, by default we dont configure it as an L2 cache
> alone. The default setting is combined mode ON where the data blocks to
> Bucket cache and index/bloom blocks go to LRU cache. But there is a way to
> turn this off and make LRU as L1 and Bucket cache as a victim handler for L1.
> It will be just L2.
> After the off heap read path optimization Bucket cache is no longer slower
> compared to L1. We have test results on data sizes from 12 GB. The Alibaba
> use case was also with 12 GB and they have observed a ~30% QPS improve over
> the LRU cache.
> This issue is to remove the option for combined mode = false. So when Bucket
> cache is in use, data blocks will go to it only and LRU will get only index
> /meta/bloom blocks. Bucket cache will no longer be configured as a victim
> handler for LRU.
> Note : WHen external cache is in use, there only the L1 L2 thing comes. LRU
> will be L1 and external cache act as its L2. That make full sense.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
