Commit graph

431 commits

Author SHA1 Message Date
root 215544820f Serialize map output locations more efficiently, and only once, in MapOutputTracker 2012-09-05 23:54:04 +00:00
root dc68febdce User Spark's closure serializer for the ShuffleMapTask cache 2012-09-05 23:06:59 +00:00
root ed937a821f Merge branch 'dev' of github.com:radlab/spark into dev 2012-09-05 22:26:49 +00:00
root 1d6b36d3c3 Further tuning for network performance 2012-09-05 22:26:37 +00:00
root 3fa0d7f0c9 Serialize BlockRDD more efficiently 2012-09-05 08:28:15 +00:00
root 4a5d0d249e Merge branch 'dev' of github.com:radlab/spark into dev 2012-09-05 08:23:09 +00:00
root efc7668d16 Allow serializing HttpBroadcast through Kryo 2012-09-05 08:22:57 +00:00
root 75487b2f5a Broadcast the JobConf in HadoopRDD to reduce task sizes 2012-09-05 08:14:50 +00:00
root b7ad291ac5 Tuning Akka for more connections 2012-09-05 07:08:07 +00:00
root fc186dc18a Merge branch 'dev' of github.com:radlab/spark into dev 2012-09-05 05:53:18 +00:00
root 4ea032a142 Some changes to make important log output visible even if we set the logging to WARNING 2012-09-05 05:53:07 +00:00
Tathagata Das 7c09ad0e04 Changed DStream member access permissions from private to protected. Updated StateDStream to checkpoint RDDs and forget lineage. 2012-09-04 19:11:49 -07:00
Tathagata Das b8e9e8ea78 Merge branch 'dev' of github.com:radlab/spark into dev 2012-09-02 02:35:32 -07:00
root ceabf71257 tweaks 2012-09-01 21:52:42 +00:00
root 6025889be0 More raw network receiver programs 2012-09-01 20:51:07 +00:00
Matei Zaharia f84d2bbe55 Bug fixes to RateLimitedOutputStream 2012-09-01 00:31:15 -07:00
Matei Zaharia 44758aa8e2 First work towards a RawInputDStream and a sender program for it. 2012-09-01 00:17:59 -07:00
root c42e7ac282 More block manager fixes 2012-09-01 04:31:11 +00:00
root 113277549c Really fixed the replication-3 issue. The problem was a few buffers not being rewound. 2012-08-31 05:39:35 +00:00
Matei Zaharia 101ae493e2 Replicate serialized blocks properly, without sharing a ByteBuffer. 2012-08-30 22:24:14 -07:00
Matei Zaharia 1b3e3352eb Deserialize multi-get results in the caller's thread. This fixes an
issue with shared buffers with the KryoSerializer.
2012-08-30 17:59:25 -07:00
root c4366eb764 Fixes to ShuffleFetcher 2012-08-31 00:34:24 +00:00
Tathagata Das 4db3a96766 Made minor changes to reduce compilation errors in Eclipse. Twirl stuff still does not compile in Eclipse. 2012-08-29 13:04:01 -07:00
Matei Zaharia 17af2df0cd Log levels 2012-08-27 23:07:32 -07:00
Matei Zaharia b4a2214218 More fault tolerance fixes to catch lost tasks 2012-08-27 22:49:29 -07:00
Matei Zaharia b914cd0dfa Serialize generation correctly in ShuffleMapTask 2012-08-26 20:07:59 -07:00
Matei Zaharia 69c2ab0408 logging 2012-08-26 20:00:58 -07:00
Matei Zaharia 117e3f8c86 Fix a bug that was causing FetchFailedException not to be thrown 2012-08-26 19:52:56 -07:00
Matei Zaharia 3c9c44a8d3 More helpful log messages 2012-08-26 19:37:43 -07:00
Matei Zaharia 26dfd20c9a Detect disconnected slaves in StandaloneScheduler 2012-08-26 18:56:56 -07:00
Matei Zaharia 29e83f39e9 Fix replication with MEMORY_ONLY_DESER_2 2012-08-26 18:16:25 -07:00
Matei Zaharia 06ef7c3d1b Less debug info 2012-08-26 16:29:20 -07:00
Matei Zaharia 741899b21e Fix sendMessageReliablySync 2012-08-26 16:26:06 -07:00
Matei Zaharia 5a8015d2db Merge remote-tracking branch 'public/dev' into dev 2012-08-24 16:11:44 -07:00
Matei Zaharia 2c16ae36d7 Set log level in tests to WARN 2012-08-23 20:38:14 -07:00
Matei Zaharia deedb9e7b7 Fix further issues with tests and broadcast.
The broadcast fix is to store values as MEMORY_ONLY_DESER instead of
MEMORY_ONLY, which will save substantial time on serialization.
2012-08-23 20:31:49 -07:00
Matei Zaharia 59b831b9d1 Fixed test failures due to broadcast not stopping correctly 2012-08-23 19:59:55 -07:00
Matei Zaharia 7310a6f499 Merge pull request #147 from mosharaf/dev
Broadcast refactoring/cleaning up
2012-08-23 19:38:28 -07:00
Matei Zaharia 25a6a39e6d Added other SparkContext constructors to JavaSparkContext 2012-08-19 18:59:16 -07:00
Shivaram Venkataraman 0f4fbb057b Change BlockManagerSuite test cases to use a deterministic size estimator and
update the results to match the new estimates
2012-08-13 13:32:23 -07:00
Shivaram Venkataraman 22ba3a3f77 Add test-cases for 32-bit and no-compressed oops scenarios. 2012-08-13 13:32:10 -07:00
Shivaram Venkataraman 1f68c4b03b Update test cases to match the new size estimates. Uses 64-bit and compressed
oops setting to get deterministic results
2012-08-13 13:31:54 -07:00
Shivaram Venkataraman 1ea269110c Move object size and pointer size initialization into a function to enable unit-testing 2012-08-13 13:31:45 -07:00
Shivaram Venkataraman 44661df9cc If spark.test.useCompressedOops is set, use that to infer compressed oops
setting. This is useful to get a deterministic test case
2012-08-13 13:31:39 -07:00
Shivaram Venkataraman 0dd8fe73ba Use HotSpotDiagnosticMXBean to get if CompressedOops are in use or not 2012-08-13 13:31:29 -07:00
Shivaram Venkataraman 80104ce1da Add link to Java wiki which specifies what changes with compressed oops 2012-08-13 13:31:21 -07:00
Shivaram Venkataraman 00ab5490b3 Changes to make size estimator more accurate. Fixes object size, pointer size
according to architecture and also aligns objects and arrays when computing
instance sizes. Verified using Eclipse Memory Analysis Tool (MAT)
2012-08-13 13:31:11 -07:00
Matei Zaharia 6ae3c375a9 Renamed apply() to call() in Java API and allowed it to throw Exceptions 2012-08-12 23:10:19 +02:00
Matei Zaharia 0141879c40 Use Promises instead of having a Future wait on a thread in
ConnectionManager.
2012-08-12 22:16:32 +02:00
Matei Zaharia 845a870242 Return remotely fetched blocks in a pipelined fashion from BlockManager 2012-08-12 20:01:38 +02:00