Now that the yearly gathering of HBase aficionados is over by a couple of weeks, let’s take a look back at what was hot at the conference, and more generally speaking what is happening in the land of HBase. HBase is the open source BigTable data store and part of Lily’s core foundations. It brings interactive data management to the Hadoop ecosystem.
The HBaseCon show was well attended – 750 attendees over 600 last year’s inaugural edition – a sign of success and an increasingly more mature community. We at NGDATA were showcasing our HBase Triggers and Indexing library, as part of Lily – alongside Lily itself of course.
Overall, HBaseCon had a great vibe, and things got pretty hectic at our booth – we ran out of product datasheets long before time, so get in touch with us if you want one.
Session-wise, there were 3 clear themes present at the conference:
- Making HBase easier to develop against
- Indexing HBase
- Operationalizing HBase
To make HBase easier to develop against, there’s a variety of projects arising, such as Continuuity with its PaaS SDK & application server allowing to build and deploy HBase-backed applications in the cloud, and Wibidata with Kiji – a higher-level HBase schema layer and data application API.
Beyond HBase’s simple data management and CRUD API, a lot of organizations are exploring the addition of Search as a discovery mechanism for HBase. We were amongst a few others (Intel and Salesforce) to showcase research and innovations in the area of indexing HBase, with our solution now powering Cloudera Search’s HBase support. The race is now on to provide a commonly adopted solution for secondary indexing in HBase.
Finally, with HBase adoption growing, many sessions were on the topics of operationalization, performance tuning, optimal schema designs, and improving failure recovery time. In that regard, Steve Loughran’s early efforts at running HBase on Yarn seem like an interesting development to follow, to me.
