Another membase milestone was reached today – beta 2 was released and is available for download! Several cool features have been added, including support for datasets whose size exceeds the size of aggregate cluster main memory (i.e. supporting disk > RAM); very sexy, and useful, real-time and historical stat displays; and support for deploying moxi, the membase proxy, on a client-side machine. Looking back over the last three weeks, community reaction to membase has exceeded our collective expectations. We knew we were addressing an unmet need, but it is always a good feeling to hear it confirmed. We’ve had hundreds of downloads of membase beta 1 over the last three weeks and the feedback has been overwhelmingly positive: – “Membase appears to be the reliable, sharding and persistent memcached-alike we’ve all
been waiting for…” – “Membase is fast! like memcached fast. very low latency under load and good throughput…” – “Oh this is so hot, so very, very, hot…” But while it is nice to hear the good stuff, I tend to prefer hearing about the things people don’t like or the things users having trouble with.
There are always areas for improvement…and we are in the improvement business. One of the cool things about the open source software model is seriously high-value feedback one receives – not just “this feature is missing” or “I think you may have a memory leak,” but “you have a potential deadlock in this section of code…and here is a fix for the problem!” The top two complaints about membase beta 1 are “there is not enough technical information” and “are some of the features not working?” Technical information – We are still not where we want to be. Our objective is to make it insanely easy to get, build, change, contribute to and use membase, from the membase.org open source project site. We’ve worked hard over the last couple weeks to provide more information on the membase.org wiki, but there is more to come. We’ve been busy writing software, but are in the middle of transitioning to ensuring that software is easy to understand and work with. One thing we have decided to do, though, is to make it much easier to find and get information from the couchbase.com website. We have a reasonably detailed technical white paper that was a bit hard to find for people, and required registration to download. We’re highlighting that content now, and removing the friction involved in getting it. Features – We released membase beta 1 with some features not yet complete. We felt it was much more important to get the core technology in the hands of users rather than hold it back until all the bells and whistles were installed. Beta 1 delivered a high-performance, elastic, persistent, reliable, replicating key-value store which is on-the-wire compatible with memcached (so it already works with tens of thousands of applications and virtually every). Beta 2 begins to deliver the bells and whistles. The major piece of added functionality is the ability to store datasets that don’t completely fit in memory. The addition of monitoring tools and a plethora of statistics to select from provide deep insight into the inner workings of membase – providing valuable insight, troubleshooting, tuning and capacity information. Please keep the feedback coming! Beta 3 is shaping up to be an even more feature-packed release and we really want the community to help us drive the targets. We welcome the feedback via any channel you like – twitter (#membase or #northscale), support AT northscale.com or any of the other channels highlighted on membase.org: http://membase.org/gethelp.html.