From what I managed to grasp while reading their Java front-end, their HBase support is very limited: RowKey supports is limited to Strings only, and I think only one value (field) Occupies more space, and does not allow composite Primary Key WHERE supports only for Strings type fields No support for integer type fields, thus you are forced to occupy more space when saving your data No support for binary encoding of couple of fields, although they did mention they are using Avro - couldn't find it yet. This means each column is saved as a real HBase column -> occupies space and might hinder performance on very large scale data set No support for secondary indexes No support for user defined functions in SELECT, or as whole (Stored procedure like)
All and all, the code seems very solid: BE written in C++ Could be good for performance, but I'm not sure it's worth the overhead if you're running on HBase which likely to cause any bottlenecks or any query to be slow. BE - FE streaming communication, which supports canceling a query in the middle of execution, and I guess allows for streaming results to the client FE written in Java. Communications between Java FE and BE using Thrift Asaf Mesika On 24 באוק 2012, at 19:04, "Surendra , Manchikanti" <[email protected]> wrote: > Impala supports HDFS and HBase. > > Thanks, > -- Surendra Manchikanti > > > On Wed, Oct 24, 2012 at 10:23 PM, Arun Ramakrishnan < > [email protected]> wrote: > >> 1. Its Apache lincese< >> http://blog.cloudera.com/blog/2012/10/cloudera-impala-real-time-queries-in-apache-hadoop-for-real/ >>> >> 2. Trevni columnar format is something to look into. >> >> On Wed, Oct 24, 2012 at 9:18 AM, karthik tunga <[email protected] >>> wrote: >> >>> >>> >> http://blog.cloudera.com/blog/2012/10/cloudera-impala-real-time-queries-in-apache-hadoop-for-real/ >>> >>> It looks like impala has HDFS support as well. >>> >>> Cheers, >>> Karthik >>> >>> On 24 October 2012 12:16, [email protected] < >>> [email protected]> wrote: >>> >>>>> It's licensed to Cloudera only. >>>> >>>> I don't think that is entirely true. >>>> >>>> "*Cloudera Impala is an Apache-licensed open source project that was >>>> founded and is led by Cloudera.*"[1] >>>> >>>> [1] >>>> >>>> >>> >> http://www.cloudera.com/content/cloudera/en/products/cloudera-enterprise-core/cloudera-enterprise-RTQ.html >>>> >>>> On Wed, Oct 24, 2012 at 11:11 AM, Timothy Chen <[email protected]> >>> wrote: >>>> >>>>> I think this is right up to our valley. >>>>> >>>>> Http://github.com/Cloudera/impala >>>>> >>>>> It's licensed to Cloudera only. >>>>> >>>>> Supports llvm ir and looks like its planning to support all different >>>>> formats like we do. >>>>> >>>>> Tim >>>>> >>>>> Sent from my iPhone >>>> >>>> >>>> >>>> >>>> -- >>>> Swarnim >>>> >>> >>
