Steve Sanbeg wrote:
On Tue, 24 Apr 2007 17:22:57 -0600, Jeff V. Merkey wrote:
Since I have open sourced the wikix program, now anyone wanting the images can download them directly from Wikipedia. I am in process of restructuring the WikiGadugi site, so anyone wanting the bittorrent downloads need to finish up this week, as I will discontinue them shortly since folks now have the ability to download them directly. The wikix program is not very intensive on the main Wikimedia servers. The program is setup to behave as several workstations, and it really does not take that long to get the images.
I was under the impression that bulk downloads needed to be throttled, and that it would take a lot longer than that to download everything. Does this just grab the images as fast as it can get them? Is that allowed?
It's faster to get them from Wikipedia. The bittorrent downloads take about 1 1.2 weeks to download the archive. Using wikix directly only takes 1 1/2 days given the current size of the image set for commons.
Getting them from Wikipedia is faster due to the squid caching both locally and internet wide. My analysis of the data sets from Wikipedia indicates that 60% of the images are cached either locally on squid or at other remote cache servers.
Since they are cached in a distributed manner, the program will only access wikipedia intermittently. Copyvio is the bigger issue than performance. My image mirroring has had almost no noticable impact on Wikipedia with wikix. The program behaves like 16 workstations, so Wikipedia seems to be able to handle it with little additional overhead. Given the number of squid servers Brion has active, I think the impact is minimal in comparison to the massive amounts of access the site gets daily.
Jeff
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org http://lists.wikimedia.org/mailman/listinfo/wikitech-l