I spent some time a week or two ago talking with Liran Zelka, co-founder of Database proxy provider ScaleBase. I first talked with Zelka at CloudConnect earlier in the year, and the reason for our catch up was to discuss the fact that ScaleBase is going into general availability this morning.
ScaleBase was founded on the premise that modern approaches to database management necessitate extensive database sharding. ScaleBase is a proxy that sits between the database shards and the application itself and acts as a kind of load balancer, routing database commends to the particular database shard they are intended for. In ScaleBase’s words, they’re helping create the concept of “transparent sharding”, a concept whereby data is assigned to multiple databases, yet looks to the application as if it were managed by just one. In this way a sharded database should look, and more importantly perform, like a standalone one. ScaleBase therefore call themselves a Database Load Balancer, although perhaps database router is a more apt term. ScaleBase have done some standard testing of their product and have seen both increasing linear throughput as the number of database shards increases;
and reduced overall response time as connections rise;
ScaleBase is up against primarily in house development of sharding code. Zelka pointed out that most companies who use sharding in their databases will also creating their own code to route database commends – ScaleBase believes that this is inefficient and unreliable and that their ability to abstract this part of database setup will be attractive to companies. In fact since starting a beta program earlier this year, ScaleBase reports over 500 in bound queries from companies, predominantly startups, wanting to trial the ScaleBase solution.
ScaleBase is downloadable software that can be installed in the cloud or on-premise. It also has the ability to be run in a clustered fashion to avoid the database load balancer becoming the single point of failure in the chain. Zelka reports that most beta customers are using ScaleBase in this way. ScaleBase isn’t charging by the number of load balancers used, rather they charge per database, currently this is set at $5000/year/server with obvious reductions for volume customers.
I put it to Zelka that they perhaps face a bit of a competitive threat from someone open sourcing a similar database load balancing product – especially given the widespread use of sharding and the pain of creating query routing code for every user of sharding. Zekla admitted that this is possible (in fact Twitter has apparently open sourced a product that, to an extent follows this concept). However Zelka felt that while a core product might be open sourced, it was unlikely to be sufficiently robust to support all the different types of queries etc.
ScaleBase currently only supports MySQL but has plans to roll out support for other databases. Oracle support is slated for mid 2012, while priorities for other databases have not yet been set – market demand will dictate this. Given the recent storm created by Michael Stonebreaker’s assertions that large MySQL implementations are a “fate worse than death”, it is interesting to hear about companies that ScaleBase that are easing the difficulties (if not reversing the existence of) massive sharded environments.