This project is very interesting.
> Arrow is (amongst other things) a way to structure data in memory for high > speed processing and deals with records, fields and datatypes. Am I correct in assuming this is similar to Google Powerdrill [1] and IBM BLU [2]? If so some insight on how similar/different the project goals are would be great. Thanks! [1] Processing a Trillion Cells per Mouse Click - VLDB'12 [2] DB2 with BLU Acceleration: So Much More than Just a Column Store - VLDB'13 On Thu, Feb 18, 2016 at 3:52 PM, Sambit Tripathy (RBEI/EDS1) < [email protected]> wrote: > Sounds exciting. > > -Sambit. > > > -----Original Message----- > From: Jacques Nadeau [mailto:[email protected]] > Sent: Thursday, February 18, 2016 2:55 PM > To: [email protected] > Subject: Re: vs Tachyon > > A simple (albeit somewhat limited) analogy would be a comparison to Parquet > and HDFS. > > Parquet is a way to store data in a columnar representation on disk. HDFS > is a distributed way to read and write streams of data on physical media. > > Arrow is (amongst other things) a way to structure data in memory for high > speed processing and deals with records, fields and datatypes. Tachyon is > distributed way to read and write streams of data in-memory. > > There is a great opportunity to integrate Tachyon and Arrow as they are > very complimentary. Being able to share Arrow representations through a > shared in-memory storage system would be helpful to many applications. > > > > On Thu, Feb 18, 2016 at 11:45 AM, Sambit Tripathy (RBEI/EDS1) < > [email protected]> wrote: > > > > > Hi, > > > > I thank the community for this initiative. > > > > Just out of curiosity, is this similar to http://tachyon-project.org/ > > (Tachyon)? > > > > > > > > Best regards > > > > Sambit Tripathy > > > > > > > -- Snehasish Kumar School of Computing Science Simon Fraser University
