Apache: Big Data 2016

HBase has been powering variety of applications in the past 8 years.
In this presentation, I will talk about the following recent developments:

Enhancement to compaction: Properly selecting / tuning compaction strategy is at the heart of providing consistent performance. FIFO compaction policy collects expired store files. Since no real compaction is done, we do not use CPU and IO (disk and network). This results in improved throughput and latency for both write and read.

Region normalization feature: This serves times series data well (in combination with FIFO compaction policy) where non-default TTL is specified. As aging data is archived, adjacent empty regions are continuously merged. This keeps table in well managed shape.

Bulk Loaded HFile Replication: HBase replication is enhanced to support replication of bulk loaded data. This completes disaster tolerance scenario.