Marcel Kornacker has posted comments on this change.

Change subject: IMPALA-4172: Switch to BlockLocation methods for disk IDs
......................................................................


Patch Set 1:

(1 comment)

http://gerrit.cloudera.org:8080/#/c/4914/1/fe/src/main/java/org/apache/impala/catalog/HdfsTable.java
File fe/src/main/java/org/apache/impala/catalog/HdfsTable.java:

Line 379:   private static int getDiskId(String storageId) {
> In the BE we mod the disk id to determine which disk queue to put reads on,
it's very unusual for a server to have more than 128 disks (i don't think we've 
ever seen one), so from that perspective i'm not worried about using fewer bits 
to represent the disk "ordinal".

keep in mind that we really don't care about distinguishing disks globally. 
using a designated "out of range" value (such as 127) if a node should really 
exceed 127 disks, and then placing scan ranges with that value on random disk 
queues should degrade performance gracefully in cases where the number of disks 
is just a bit above that threshold.


-- 
To view, visit http://gerrit.cloudera.org:8080/4914
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: Ibbff94cef9a9db7b3945f8e7b0286866d2cc3b61
Gerrit-PatchSet: 1
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Laszlo Gaal <[email protected]>
Gerrit-Reviewer: Alex Behm <[email protected]>
Gerrit-Reviewer: Laszlo Gaal <[email protected]>
Gerrit-Reviewer: Marcel Kornacker <[email protected]>
Gerrit-Reviewer: Tim Armstrong <[email protected]>
Gerrit-HasComments: Yes

Reply via email to