| Summary: | Serve up p2 metadata at high (or, normal) priority | ||||||
|---|---|---|---|---|---|---|---|
| Product: | Community | Reporter: | David Williams <david_williams> | ||||
| Component: | Servers | Assignee: | Eclipse Webmaster <webmaster> | ||||
| Status: | RESOLVED WONTFIX | QA Contact: | |||||
| Severity: | normal | ||||||
| Priority: | P3 | CC: | david_williams, denis.roy, igor, jan.sievers, john.arthorne, mknauer, mober.at+eclipse, nicolas.bros, pascal, sbouchet, steffen.pingel, t-oberlies, thomas, webmaster | ||||
| Version: | unspecified | ||||||
| Target Milestone: | --- | ||||||
| Hardware: | PC | ||||||
| OS: | Windows XP | ||||||
| Whiteboard: | |||||||
| Attachments: |
|
||||||
|
Description
David Williams
Created attachment 203926 [details]
List of downloads for Sept. 22
I've compiled the top-250 files that were downloaded from download.eclipse.org for the 24-hour period of Sept. 22.
At first glance, it appears the vast majority of files we're serving are the ones you've listed.
Because that file list represents lots of data traffic, if I place it into high priority, I fear it will have an impact on current high-priority traffic, such as CVS/SVN/Git, Bugzilla and www.eclipse.org.
(In reply to comment #1) > Created attachment 203926 [details] > List of downloads for Sept. 22 > a day in the life of ... eh? Very interesting numbers. On several levels. If it helps, some of the "meta data files" are relatively small in size, and would always be expected to be, since they just contain content that points to other files (essentially). Maybe they could be "high"? p2.index compositeContent.xml compositeContent.jar compositeArtifacts.xml compositeArtifacts.jar These other 4, though, while sometimes small, are sometimes large, as you can see in the data ... think they could be a "medium" priority instead of low? content.xml content.jar artifacts.xml artifacts.jar It may take some "experimentation" to see how high we change things without overly impacting other services ... unfortunately, I don't really know how to (easily) measure p2 improvement (much less, impact on other services). Since bumping up the bandwidth substantially, there's been no need to do this. |