1 2 Previous Next 18 Replies Latest reply on May 4, 2015 3:26 AM by prashant.thakur Go to original post
      • 15. Re: Anyone in real world using large distributed cluster *with querying*?
        sannegrinovero

        By option [2] above when I mention "ad-hoc" indexing machines I mean that some people are using it but did disable the automatic election of the indexer and use a dedicated machine to act as master indexer. This machine would be configured differently than the others, according to the master/slave architecture described here: http://docs.jboss.org/hibernate/search/5.0/reference/en-US/html_single/#_overview

         

        So that's not the same as using the InfinispanIndexManager, which is a convenience to not have to deal with dedicated machines but allow Infinispan to automatically pick an indexer machine and do failover automatically (with solution [2] you have to roll your own failover, people normally do that with a cluster singleton MDB by using our JMS based backend).

         

        The main reason for that is historical: the InfinispanIndexManager option is a fairly new option; since the question was about success stories, I'll rephrase my original answer this way: I'm not aware of people using InfinispanIndexManager successfully in production, with the exception of some who use it but have async indexing enabled (who are happy). Most other happy users have such a dedicated indexer.

         

        Still, if you made indexing async (option 3 as you say), then having a dedicated indexing machine should be unnecessary: either 2 or 3 will prevent you to exhaust the worker threadpool at load.

        Great to hear you updated to 7.1 ;-)

         

        Using version 7.1 and also having async indexing I'm confident it will work fine.

        • 16. Re: Anyone in real world using large distributed cluster *with querying*?
          prashant.thakur

          Hi sannegrinovero

          Would it be better to use Hibernate Search with Infinispan as store.

          Four main questions we had are

          1) Can we still plugin Infinispan's remote query module was remote query operations ?

          2) To store data in Lucene Indexes in this cases do we need to manage Query Interceptor like implementation provided by Infinispan ?

          3) Will this result in data storage saving if only Index based search is required and no Lookup is required from key value pair ?

          4) How much do we expect query to be slower than Infinispan key based lookup if its a pointed query returning single record. but can be done from different columns. For 1k entry size and 100M entries ?

          • 17. Re: Anyone in real world using large distributed cluster *with querying*?
            sannegrinovero

            Hi Prashant,

            the short answer would be that of course you can use Hibernate Search with an Infinispan stored index, but that's a very different use case than Infinispan Query so your reference to it and you mentioning key/value pairs will need some clarifications: could you start a new thread please as we're getting very off topic here.

            • 18. Re: Anyone in real world using large distributed cluster *with querying*?
              prashant.thakur

              Starting a new thread

              1 2 Previous Next