-Dhruv On 11/07/2013 02:43 AM, Timothy Chen wrote:
Yes as for internals they don't really say much, but they say a lot more in their blog post.Looks like they have their own ByteCode generation package, I don't believe they have a columnar format or vectorization. Tim On Wed, Nov 6, 2013 at 12:52 PM, Jason Altekruse <[email protected]>wrote:I thought the docs were very sparse, especially from a development perspective. I was wondering if they were using vectorization or doing record materialization at read time like Impala. Can't imagine they would get any performance out of it if they didn't do vectorization, as they already have to live within the constraints of java. Would also be interesting to see if they used code generation or the netty/unsafe APIs at all. Will be taking a closer look in the next few weeks to see if there are some good ideas we can incorporate into Drill. -Jason On Wed, Nov 6, 2013 at 1:24 PM, Timothy Chen <[email protected]> wrote:I've been waiting for this since their data conference earlier this year:)The concepts and architecture diagram also resembles like Drill, and also is developed in Java. I do like their docs that lists the functions available and high level diagrams: http://prestodb.io/docs/current/ Tim On Wed, Nov 6, 2013 at 11:08 AM, Jason Altekruse <[email protected]>wrote:Facebook has been working internally on a Java based distributed SQLenginefor the past year, today they open sourced it. Here is the official announcement:https://www.facebook.com/notes/facebook-engineering/presto-interacting-with-petabytes-of-data-at-facebook/10151786197628920-Jason
<<attachment: design.svg>>
