[
https://issues.apache.org/jira/browse/HDFS-7276?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14190742#comment-14190742
]
Colin Patrick McCabe commented on HDFS-7276:
--------------------------------------------
So, if we're managing powers of 2, we should have 16 managers at most for
arrays up to 2**16. Actually less than that if we decide not to handle
degenerate cases like n = 1, n =2 , n =4... we can simply make 8 or 16 the
smallest array we give out... I don't even think Packet objects can be that
small once you include the header, although I haven't checked the exact minimum.
Can't we simply create those 12 or so managers immediately, and avoid all the
"create a new manager if statistics say so" logic?
{code}
174 wait(1000);
{code}
Why does this wait have a timeout? There isn't anything that we're polling
here.
{code}
... synchronized int recycle(byte[] array) {
...
notifyAll();
{code}
Shouldn't this simply be "notify"? There is no point in waking up all waiters,
because only one of them is going to get the buffer we're recycling.
> Limit the number of byte arrays used by DFSOutputStream
> -------------------------------------------------------
>
> Key: HDFS-7276
> URL: https://issues.apache.org/jira/browse/HDFS-7276
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: hdfs-client
> Reporter: Tsz Wo Nicholas Sze
> Assignee: Tsz Wo Nicholas Sze
> Attachments: h7276_20141021.patch, h7276_20141022.patch,
> h7276_20141023.patch, h7276_20141024.patch, h7276_20141027.patch,
> h7276_20141027b.patch, h7276_20141028.patch, h7276_20141029.patch,
> h7276_20141029b.patch
>
>
> When there are a lot of DFSOutputStream's writing concurrently, the number of
> outstanding packets could be large. The byte arrays created by those packets
> could occupy a lot of memory.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
