It might get us part way to HBASE-4047? One of the first stumbling blocks of an external coprocessor host is a high bandwidth channel between processes.
On Thu, Feb 15, 2018 at 10:39 PM, Stack <[email protected]> wrote: > On Thu, Feb 15, 2018 at 11:39 AM, Sean Busbey <[email protected]> wrote: > > > Heya folks, > > > > I've been reading about underlying technologies we might use to > > improve coprocessors, with an eye towards a long term goal of better > > isolating them. > > > > Today I started reading about Apache Arrow, which is a data format and > > support library that has a focus on in-memory perf. (in particular > > they claim their libs are all zero-copy.) > > > > Their main website[1] calls out Apache HBase as a use case and our > > community as backers. I can't find anything in our code base though, > > anyone know of any previous or ongoing efforts? > > > > > > > Some noise on instantiation but nought thereafter that I know of. > > What you thinking? We'd pass an external process blocks made by hbase for > it to play with and then we'd persist whatever it gave us back? > > Thanks Sean, > S (Apache Arrow PMC -- Smile). > > > > > [1]: http://arrow.apache.org/ > > > -- Best regards, Andrew Words like orphans lost among the crosstalk, meaning torn from truth's decrepit hands - A23, Crosstalk
