hadoop-common-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jason Kace (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HADOOP-13144) Enhancing IPC client throughput via multiple connections per user
Date Fri, 13 May 2016 22:33:12 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-13144?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15283247#comment-15283247
] 

Jason Kace commented on HADOOP-13144:
-------------------------------------

One solution is to allow each user + remote address to utilize multiple connection threads.
 This will require reconfiguration of the Client class to permit a pool of connections per
ConnectionId.

An alternative solution is to create multiple ConnectionIds per user + remote address.  The
current {{ConnectionId}} class does not support multiple hashCodes per user + remote address.
  The {{ConnectionId}} can either be modified or made visible (attached solution) for inheritance.

Our use case for this feature requires a single user to be able to issue a large number of
RPC requests to a single NN via the IPC client.  Better throughput is required in the existing
IPC client to allow up to 100k requests/second from the same user to the same remote address.


> Enhancing IPC client throughput via multiple connections per user
> -----------------------------------------------------------------
>
>                 Key: HADOOP-13144
>                 URL: https://issues.apache.org/jira/browse/HADOOP-13144
>             Project: Hadoop Common
>          Issue Type: Improvement
>          Components: ipc
>            Reporter: Jason Kace
>            Priority: Minor
>             Fix For: 2.8.0
>
>
> The generic IPC client ({{org.apache.hadoop.ipc.Client}}) utilizes a single connection
thread for each {{ConnectionId}}.  The {{ConnectionId}} is unique to the connection's remote
address, ticket and protocol.  Each ConnectionId is 1:1 mapped to a connection thread by the
client via a map cache.
> The result is to serialize all IPC read/write activity through a single thread for a
each user/ticket + address.  If a single user makes repeated calls (1k-100k/sec) to the same
destination, the IPC client becomes a bottleneck.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-issues-help@hadoop.apache.org


Mime
View raw message