{"id":2217,"date":"2013-11-12T23:38:10","date_gmt":"2013-11-13T04:38:10","guid":{"rendered":"http:\/\/hstore.cs.brown.edu\/?p=2217"},"modified":"2013-11-12T23:38:10","modified_gmt":"2013-11-13T04:38:10","slug":"new-release-november-2013","status":"publish","type":"post","link":"https:\/\/hstore.cs.brown.edu\/2013\/11\/new-release-november-2013\/","title":{"rendered":"New Release (November 2013)"},"content":{"rendered":"

The H-Store project is pleased to announce the release of the latest version of its high-performance, distributed transaction processing database management system. The major enhancements in the system are related to transaction correctness and for speculative execution. This is a belated release, as most of the new features and improvements were implemented earlier in the summer. <\/p>\n

We are also excited to announce the start of new research projects in H-Store with our new collaborators at Intel Labs as part of the ISTC for Big Data<\/a> and at the Qatar Computing Research Institute<\/a>. More information on these projects will be available later in the winter.<\/p>\n

Major highlights of this release:<\/b><\/p>\n

    \n
  1. Improved speculative transaction execution. H-Store now supports optimistic concurrency control<\/a> for speculative tasks using a new tuple-level read\/write tracking mechanism. The PartitionExecutor<\/a><\/span> now uses a new conflict checking API<\/a> that can determine whether it is safe to speculatively execute a transaction before it is invoked or during the PREPARE phase of 2PC. [a2e3e09]<\/a> [5b94e06]<\/a> [3ea53e8]<\/a> [5dbfa36]<\/a> [b2fa101]<\/a>\n
  2. New automatic query prefetching for distributed transactions. The system can identify when a transaction will need to execute a query at a remote partition before it queued up and then piggyback that query invocation in the initialization request. These queries will be asynchronously executed and have their results returned to the base partition to avoid needing to have the transaction block waiting for the network roundtrip. [859972b]<\/a> [3bc4fbe]<\/a> [51cf9d8]<\/a> [e737233]<\/a>\n
  3. Various performance improvements and optimizations for the anti-caching<\/a> feature, including support for updating the internal LRU chain for tuples accessed through secondary indexes. [e33d394]<\/a>\n
  4. Added support for Mac OSX 10.7. [5c64bd9]<\/a> [b6714a9]<\/a>\n
  5. New Benchmarks:<\/em> Added the SmallBank<\/a> benchmark from Mohammad Alomari<\/a> and Michael Cahill<\/a>. [475bcd1]<\/a> [1773453]<\/a> [edccaa0]<\/a>\n
  6. FIX: The “early prepare” optimization for 2PC has been broken for some time. It was finally fixed in exchange for Aaron<\/a> giving Andy<\/a> a non-sensual massage at SIGMOD. [6d655dc]<\/a> [151802b]<\/a>\n
  7. FIX: Fixed support in the for queries using constant values. This prevents them from having to be broadcast to every partition. Query predicates that use a equality comparison with a constant value are now routed to the correct number of partitions. Non-equality predicates are broadcast to every partition. [6c7c911]<\/a> [81b0f02]<\/a>\n
  8. FIX: INSERT\/UPDATE\/DELETE queries on replicated tables now return the logical number of tuples modified. [f230114]<\/a>\n<\/ol>\n

    This release can be downloaded via Github:<\/p>\n