asterixdb-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mike Carey <>
Subject Re: Few questions about AsterixDB
Date Wed, 09 Dec 2015 21:13:37 GMT
1.  We parallelize pretty well for most things - we use parallel join 
methods, parallel aggregation, etc.
      Our one major/known issue is that for parallel sorting (e.g., 
parallel order by) we do the final merge on one node,
      which doesn't scale well for obvious reasons.  We are working on 
that and it won't be that way for much longer. :-)
2.  There is a companion piece of software called Pregelix which 
supports parallel graph computations using the
      Pregel programming model (borrowed from Google).  AsterixDB can 
interoperate with Pregelix, using queries to
      create graphs to pass to it, and taking the results back from it, 
through temporary datasets.  We have no built-in
      algorithms, but if one had Pregel algorithms for these, that'd be 
a way to get to them on Big Data.
3.  Same answer as for 2.
4.  Only through writing an application program (if I understand the 
question right).

On 12/9/15 11:18 AM, Malarout, Namrata (398M-Affiliate) wrote:
> Hi,
> I had for a few question specific to AsterixDB for a Data Container 
> Study I am working on. I can’t really address these as I have not 
> explored them yet.
> The questions are:
>  1. What, if any, are the parallelization bottlenecks?
>  2. Does AsterixDB support Anomaly Detection, Clustering or Graph
>     Traversal?
>  3. Can we find largest connecting components in graph?
>  4. Is automation possible for repetitive tasks?
> If anyone has worked with use cases which use any of the above 
> concepts or have knowledge about these and could help answer them, it 
> would be great.
> Thanks in advance,
> Namrata

View raw message