Submissions/Speeding up the Wikimedia websites without breaking them
|
This is an open submission for Wikimania 2011. |
- Review no.
132
- Title of the submission
Speeding up Wikimedia's websites without breaking them
- Type of submission (workshop, tutorial, panel, presentation)
Presentation
- Author of the submission
- E-mail address or username (if username, please confirm email address in Special:Preferences)
- Country of origin
United States
- Affiliation, if any (organization, company etc.)
Wikimedia Foundation
- Personal homepage or blog
http://techblog.wikimedia.org/author/robla/
- Abstract (please use no less than 300 words to describe your proposal)
Once upon a time, Wikimedia Foundation was only just barely able to keep up with the growth of Wikipedia and related sites. We're still very lean by industry standards, but we're getting to the point where we can be more proactive about reliability and performance, and to make more frequent releases. The Wikimedia Foundation Engineering organization has several groups (Features, Mobile, Operations, Fundraising, and General); this talk will focus on the engineering activities of the General Engineering group at Wikimedia Foundation.
Over the coming year, the General Engineering group plans to address several aspects of site reliability, performance, and deployment velocity.
One effort to improve reliability and to increase deployment velocity, we plan to make a critical change to the way we do software deployments our "Heterogeneous deploy" project. This will make it possible to gradually roll out new versions of the software without having to deploy to all sites at once. We can monitor problems and reverse course before inflicting those problems on everyone, and we'll get the ability to more easily deploy test wikis that actually reflect real-world conditions.
With every new feature we add (be it something a software developer adds to the core software, or a new capability a Wikipedian adds to a widely-used template), we potentially increase the load on our servers. While Moore's Law suggests that normal hardware replacement schedules will allow us some headroom to increase our CPU usage, it won't buy us enough headroom on a site where anyone can edit. We have many different options we're exploring at this time, such as better caching to reduce the need for expensive parse operations, and the deployment of a HipHop-compiled MediaWiki.
Rob will talk about these topics and answer questions about the General Engineering organization and Wikimedia Foundation engineering as a whole.
- Track (People and Community/Knowledge and Collaboration/Infrastructure)
Collaboration/Infrastructure
- Will you attend Wikimania if your submission is not accepted?
Most likely.
- Slides or further information (optional)
Interested attendees
If you are interested in attending this session, please sign with your username below. This will help reviewers to decide which sessions are of high interest. Sign with four tildes. (~~~~).
- Vibhijain 10:47, 5 May 2011 (UTC)
- Add your username here.