[
https://issues.apache.org/jira/browse/TS-87?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12834102#action_12834102
]
Manjesh Nilange commented on TS-87:
-----------------------------------
Leif, I tested the patch with a similar remap.config for around 3 hrs with 10M
requests with 100 concurrent clients and didn't see a crash. The aspect in
which my remap.config differed is the "to field" of the lines. I run an Apache
on my box and use that as the OS and I used Apache's URL with all caching
turned off. I don't know if turning caching off made the difference. Can you
either try with no caching or give me a core dump or something?
Btw, the crash that we followed up in Y! was triggered by regex mappings. So
that shouldn't have anything to do with this.
> Performance improvement: Avoid linear search in remap code
> ----------------------------------------------------------
>
> Key: TS-87
> URL: https://issues.apache.org/jira/browse/TS-87
> Project: Traffic Server
> Issue Type: Improvement
> Components: Core
> Reporter: Manjesh Nilange
> Assignee: Leif Hedstrom
> Priority: Minor
> Fix For: 2.0.0a
>
> Attachments: trie.patch
>
>
> Currently, the remap code stores all the remap rules in a hash table keyed by
> the host field of the "from URL" of the rule. Entries with the same host
> field are chained in a linked list. When looking for a mapping, the code has
> to do a linear traversal of this list, based on the path of the incoming
> request to find a matching rule. This performance should be improved as
> overhead can be substantial in cases where there are hundreds of the entries
> for the same host.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
