[jira] Commented: (HBASE-2219) stop using code mapping for method names in the RPC

["ryan rawson (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-2219?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12833720#action_12833720 ] 

ryan rawson commented on HBASE-2219:
------------------------------------

I did a PE (with no data, thus testing mostly RPC), and I came up with these:

with 2199:

10/02/15 00:41:08 INFO hbase.PerformanceEvaluation: Finished randomRead in 453339ms at offset 0 for 2560000 rows

without:

10/02/15 00:56:44 INFO hbase.PerformanceEvaluation: Finished randomRead in 452285ms at offset 0 for 2560000 rows



Not much of a difference here.

> stop using code mapping for method names in the RPC
> ---------------------------------------------------
>
>                 Key: HBASE-2219
>                 URL: https://issues.apache.org/jira/browse/HBASE-2219
>             Project: Hadoop HBase
>          Issue Type: Bug
>          Components: ipc
>    Affects Versions: 0.20.3
>            Reporter: ryan rawson
>            Priority: Blocker
>             Fix For: 0.21.0
>
>         Attachments: HBASE-2219-2.patch, HBASE-2219.patch
>
>
> since we use a sorted mapping of method names -> codes and send that over the wire, even trivial changes, such as adding a new call, become wire-incompatible.  This means many features which could easily have gone into a minor update must wait for a major update.  Eg: 2066, 1845, etc.
> This will increase on-wire overhead, but the compatibility is worth it I think.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.