Update of /cvsroot/cweb/bigdata/src/java/com/bigdata/journal
In directory sc8-pr-cvs4.sourceforge.net:/tmp/cvs-serv11927/src/java/com/bigdata/journal
Modified Files:
AbstractBufferStrategy.java Options.java AbstractJournal.java
Log Message:
Updated the benchmark for sustained IO, raw journal write rates, and both unisolated and isolated index write rates.
Index: AbstractBufferStrategy.java
===================================================================
RCS file: /cvsroot/cweb/bigdata/src/java/com/bigdata/journal/AbstractBufferStrategy.java,v
retrieving revision 1.15
retrieving revision 1.16
diff -C2 -d -r1.15 -r1.16
*** AbstractBufferStrategy.java 29 Mar 2007 17:01:33 -0000 1.15
--- AbstractBufferStrategy.java 4 Apr 2007 16:52:16 -0000 1.16
***************
*** 149,155 ****
if( maximumExtent != 0L && required > maximumExtent ) {
! // Would exceed the maximum extent (iff a hard limit).
! log.error("Would exceed maximumExtent="+maximumExtent);
return false;
--- 149,160 ----
if( maximumExtent != 0L && required > maximumExtent ) {
! /*
! * Would exceed the maximum extent (iff a hard limit).
! *
! * Note: this will show up for transactions that whose write set
! * overflows the in-memory buffer onto the disk.
! */
! log.warn("Would exceed maximumExtent="+maximumExtent);
return false;
Index: AbstractJournal.java
===================================================================
RCS file: /cvsroot/cweb/bigdata/src/java/com/bigdata/journal/AbstractJournal.java,v
retrieving revision 1.9
retrieving revision 1.10
diff -C2 -d -r1.9 -r1.10
*** AbstractJournal.java 29 Mar 2007 17:01:33 -0000 1.9
--- AbstractJournal.java 4 Apr 2007 16:52:16 -0000 1.10
***************
*** 132,136 ****
* solution resources would be migrated to other hosts to reduce the total
* sustained resource load).</li>
! * <li> Concurrent load for RDFS w/o rollback.</li>
* <li> Network api (data service).</li>
* <li> Group commit for higher transaction throughput.<br>
--- 132,151 ----
* solution resources would be migrated to other hosts to reduce the total
* sustained resource load).</li>
! * <li> Concurrent load for RDFS w/o rollback. There should also be an
! * unisolated read mode that does "read-behind", i.e., reading from the last
! * committed state on the store. The purpose of this is to permit fully
! * concurrent readers with the writer in an unisolated mode. If concurrent
! * readers actually read from the same btree instance as the writer then
! * exceptions would arise from concurrent modification. This problem is
! * trivially avoided maintaining a distinction between a concurrent read-only
! * btree emerging from the last committed state and a single non-concurrent
! * access btree for the writer. In this manner readers may read from the
! * unisolated state of the database with concurrent modification -- to another
! * instance of the btree. Once the writer commits, any new readers will read
! * from its committed state and the older btree objects will be flushed from
! * cache soon after their readers terminate -- thereby providing a consistent
! * view to a reader (or the reader could always switch to read from the then
! * current btree after a commit - just like the distinction between an isolated
! * reader and a read-committed reader).</li>
* <li> Network api (data service).</li>
* <li> Group commit for higher transaction throughput.<br>
Index: Options.java
===================================================================
RCS file: /cvsroot/cweb/bigdata/src/java/com/bigdata/journal/Options.java,v
retrieving revision 1.11
retrieving revision 1.12
diff -C2 -d -r1.11 -r1.12
*** Options.java 22 Mar 2007 21:11:25 -0000 1.11
--- Options.java 4 Apr 2007 16:52:16 -0000 1.12
***************
*** 133,139 ****
* mode that writes through synchronously to stable storage (default
* <code>No</code>). This option does NOT effect the stability of the
! * data on disk but may be used to tweak the file system buffering.
*
* @see #FORCE_ON_COMMIT, which controls the stability of the data on disk.
* @see ForceEnum
*/
--- 133,141 ----
* mode that writes through synchronously to stable storage (default
* <code>No</code>). This option does NOT effect the stability of the
! * data on disk but may be used to tweak the file system buffering (forcing
! * writes is generally MUCH slower and is turned off by default).
*
* @see #FORCE_ON_COMMIT, which controls the stability of the data on disk.
+ * @see #DEFAULT_FORCE_WRITES
* @see ForceEnum
*/
|
