WOPR19 was held in Palo Alto, California on September 27-29, 2012, and was hosted by Facebook. Rob Sabourin was the Content Owner.
Attendees
AJ Alhait, Scott Barber, Goranka Bjedov, Dan Downing, Jane Fraser, Mieke Gevers, Doug Hoffman, Andy Hohenner, Paul Holland, Pam Holt, Yury Makedonov, John Meza, Blaine Morgan, Eric Proegler, Raymond Rivest, Robert Sabourin, Greg Seaton, John Struble,
Mais Tawfik Al-Khoury Ashkar, Mark Tomlinson
Theme: Performance Failures
WOPR19 is taking both a broad and a deep look at performance-related failures. The conference will address questions that surface after a system fails:
- What did we break?
- How did we know that anything was broken?
- How did we localize and isolate the problem?
- What technical weakness did we expose?
- What technically goes wrong in a system (or project) leading up to performance failures?
- What had to be fixed, tweaked, tuned, upgraded or replaced to correct the defects exposed by performance, load and stress testing?
- What technical weakness did we expose? Examples include problems with meeting throughput requirements, challenges with always-on/24×365.25 availability, lack of scalability, and problems with end-to-end recoverability.
- How did you separate failures (symptoms) from faults (causes)? Many failures are more subtle than breakages; consider subtleties like user perceptions, expectations and psychology.
- Did we validate our models of expected behavior? Were they realistic, reproducible and statistically valid? Consider treatment of outliers, and predictive analytics.
- How do teams work together to isolate bugs related to performance, response time, throughput or latency?
- Do certain failures recur? Do patterns emerge? What can teams do to avoid them in the first place?
We are looking for experience reports from performance related failures. Especially interesting are experiences related to tricky to isolate, hard to repeat and otherwise challenging bugs found during testing or live operation. During open season discussion, common and contrasting failings, isolation approaches, and fixes will be collected and discussed.
Work Product
Private workshop collaboration space here.