[
https://issues.apache.org/jira/browse/LUCENE-6614?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14602844#comment-14602844
]
Dawid Weiss commented on LUCENE-6614:
-------------------------------------
> you are a seriously advanced user
The only reason I know it's working at sub-par performance level is because I
am an advanced user... I know the source code. :)
I disagree with your opinion, but I respect it since I understand those
low-level methods to detect various things would proliferate. I still have this
vision of the CMS just being smart enough to figure out how many concurrent
merges it can run without blocking I/O. This really shouldn't be that difficult
since it's essentially a maximization problem of {{f(count-of-merge-threads,
max-throughput-per-thread) = total merge throughput}}. The problem is in how to
pick a strategy that is robust enough to handle and recover from intermediate
external system loads, disk/ system buffer spills, etc. An interesting research
problem, really.
> IOUtils.spins doesn't work for drives NVMe drives
> -------------------------------------------------
>
> Key: LUCENE-6614
> URL: https://issues.apache.org/jira/browse/LUCENE-6614
> Project: Lucene - Core
> Issue Type: Bug
> Reporter: Michael McCandless
> Assignee: Uwe Schindler
> Priority: Minor
> Fix For: 5.3, Trunk
>
> Attachments: LUCENE-6614.patch
>
>
> NVMe is the faster (than AHCI) protocol for newer SSDs that plug into the
> PCIE bus.
> I just built a new beast box with one of these drives, and the partition is
> named {{/dev/nvme0n1p1}} while the device is {{/dev/nvme0n1}} by Linux - this
> also appears in /sys/block with rotational=0.
> I think [~steve_rowe] also has an NVME drive ...
> [~thetaphi] (who got the box working for me: thank you!!!) has ideas on how
> to fix it!
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
