Starburst, the enterprise Presto company, announced the public availability of the Starburst Distribution of Presto. The company will provide enterprise-grade support to the rapidly growing Presto user base, while remaining focused on accelerating the development of the Presto engine through continued contributions to the open source Presto project.
There is a new cool feature in Presto that will speed up some inequality joins.
Since couple of months there is a new highly efficient connector for Presto. It works by storing all data in memory on Presto Worker nodes, which allow for extremely fast access times with high throughput while keeping CPU overhead at bare minimum.
I have wanted to experiment with Java for a long time to find out whether or not it can take advantage of Single Instruction, Multiple Data (SIMD) instructions to speed up CPU-intensive computations. I found very little information while I was researching this, so I decided to share my own findings.
One can say that Presto is a MPP (Massively Parallel Processing) kind of application. Well, I have never seen a data warehouse which did not follow this approach. Teradata, Netezza, Vertica and even Hive and many many more, all of these belong to this class of software. It is not only typical for data warehouses, but also for any distributed application which is processing vast amount of data, doing non-trivial and very costly computation on it.