[
https://issues.apache.org/jira/browse/TS-1326?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13875929#comment-13875929
]
Leif Hedstrom commented on TS-1326:
-----------------------------------
I've had a use case where a large box, 8 disks and 64GB of RAM cache, which
yielded 8GB of RAM per stripe (disk). However, this prevented a very large
object (12GB) from being cached in RAM, and therefore, killed the single disk
which held this popular, large object. Our solution was to RAID stripe the 8
disks on the box, presenting one big disk to ATS, such that the cache gets the
full 64GB.
I can see a number of options here, but one would be to coalesce the RAM cache
into one single entity (and not divided). Another would be to be able to
control the RAM size per volume / stripe.
> way to control the ram space usage
> ----------------------------------
>
> Key: TS-1326
> URL: https://issues.apache.org/jira/browse/TS-1326
> Project: Traffic Server
> Issue Type: Improvement
> Components: Cache
> Reporter: Zhao Yongming
> Priority: Minor
> Fix For: 5.3.0
>
>
> we have the ram cache binding to volume and the space is used for all
> contents in the volume, with LRU.
> in some situation, you'd need to give one site for more or less ram cache
> because you know they demand it.
> we should take the ram space as a control resource, and control it
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
