[
https://issues.apache.org/jira/browse/HBASE-9869?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13822368#comment-13822368
]
Nicolas Liochon commented on HBASE-9869:
----------------------------------------
bq. See any difference? Whats the math like? How large is the Map if 1M regions
in it? We remove when region is bad but ones we don't access could stick around
for ever. I suppose that serves us right if we don't access them. Folks like
Lars are sensitive about clients being well-behaved especially when embedded in
an apps server. I looked around for a simple LRU – the guava one – but we need
SortedMap.
My feeling is that if you have a table with 1 million regions you need to pay
the price for this: i.e. have enough memory on the client.
Let me do the math & test it however.
We can as well do something in the middle: do a LRU on the first map, the one
on the tableName.
> Optimize HConnectionManager#getCachedLocation
> ---------------------------------------------
>
> Key: HBASE-9869
> URL: https://issues.apache.org/jira/browse/HBASE-9869
> Project: HBase
> Issue Type: Bug
> Components: Client
> Affects Versions: 0.98.0, 0.96.0
> Reporter: Nicolas Liochon
> Assignee: Nicolas Liochon
> Fix For: 0.98.0, 0.96.1
>
> Attachments: 9869.v1.patch, 9869.v1.patch, 9869.v2.patch
>
>
> It javadoc says: "TODO: This method during writing consumes 15% of CPU doing
> lookup". This is still true, says Yourkit. With 0.96, we also spend more time
> in these methods. We retry more, and the AsyncProcess calls it in parallel.
> I don't have the patch for this yet, but I will spend some time on it.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
