What do you mean by flat data? Csv files? You should use Orc or parquet format if you sort the data correctly. Hive+tez+Orc+optionally ignite HDFS cache of the most recent partitions in memory could be an interesting alternative for you.
Similarly spark can have a good performance even in older version. However I recommend spark mostly for iterative machine learning . > On 17 Jun 2016, at 14:48, Andrés Ivaldi <[email protected]> wrote: > > Jörn, you are right with that point, actually we don't have a data model we > just query flat data to perform a kind of ROLAP. > We are currently researchingk for the best option. The idea with Ignite is to > use in-memory cache to perform fast queries and as a layer for different kind > of data sources (not necessary RDBMS) > We where able to do it with Spark but is too slow for user experience(I've to > try 2.0 they said that the speed was improved), also I looked that Ignite can > be used as Spark chache with Ignite RDD maybe that could be another approach. > > Thanks > >> On Fri, Jun 17, 2016 at 2:29 AM, Jörn Franke <[email protected]> wrote: >> >> This depends on the type of queries! >> In any case: before you go in-Memory optimize your current data model and >> exploit your current technology. I have seen in the past often purely >> designed data model that do not leverage the underlying technology well. >> >>> On 16 Jun 2016, at 23:20, Andrés Ivaldi <[email protected]> wrote: >>> >>> Hello, I'm new with Apache Ignite, I'd like to know how can I use Apache >>> Ignite cache to bust up RDBMS queries. I saw it in apache conn that is >>> possible perform querys against RDBMS allowing to speed up it's execution >>> >>> Regards. >>> >>> -- >>> Ing. Ivaldi Andres > > > > -- > Ing. Ivaldi Andres
