On Sat, Apr 13, 2013 at 6:46 AM, Avery Ching <ach...@apache.org> wrote: > That's great Sebastian. I would also recommend taking a look at the > PageRankBenchmark for a performance comparison. It has been a lot of speed > improvements that should be a bunch faster than PageRankVertex. Even that > though, is not totally optimized. Hopefully we'll be adding a "how to > optimize performance" guide in the near future. Should we delay the release > or simply just ship a 1.1, say in the next month with this fix and > supporting YARN's 2.0.4? I'd like to get on a more normal release cycle > rather than once a year =).
Excellent point. The release process that I have seen work very well in the past is that you get on a regular date-driven release schedule (lets say quarterly) but you still allow for patch release with very limited scope from time to time. Path releases are NOT date driven -- they go out when all the patches required are committed to trunk and back ported to the appropriate branch. So, would the following work for us: we go ahead with 1.0 release as planned (unless showstoppers get identified in the next 72 hours). But we all agree to 1.0.1 with scope strictly limited to YARN stabilization on top of Hadoop 2.0.4. If we get all that we need for 1.0.1 in a week -- we'll release in a week if it takes us 3 weeks -- we'll release in 3 weeks and so on. Thanks, Roman.
