The Sun Storage 7000 series 2009.Q3 software release officially announced on September 22, provides a new iSCSI software stack via the Common Multiprotocol SCSI Target or COMSTAR framework, a new dataset Synchronous write bias property or logbias, and many other features. Focusing on the aforementioned features, a Sun blog by Roch Bourbonnais explores each of these two, detailing the ins and outs and why they are significant to performance.

Bourbonnais begins his blog entry iSCSI unleashed by explaining the new target code is an in-kernel implementation that replaces what was previously known as the iSCSI target deamon, a user-level implementation of iSCSI. With this new update, huge performance gains can be expected in certain circumstances, he writes.

A major flaw fixed in this update relates to iSCI requests. Bourbonnais points out that in the previous release, iSCI requests required multiple trips through a single queue, which was being read and written by two specific threads. The two threads fought for the same locks. This was compounded by the fact that user level threads can be put to sleep when they fail to acquire a mutex. "So while the iSCSI target deamon gave reasonable service for large request," writes Bourbonnais, "it was much less scalable in terms of the number IOPS that can be served and the CPU efficiency in which it could do that." For example, in a circumstance with 10 client initiators and 10 threads per initiators (so 100 outstanding request) doing 8K cache-hit reads, Sun testing resulted in the old target daemon recording 31K IOPS while the updated COMSTAR reached 85K IOPS for more than double in IOPS handling.

The blog continues through the testing Sun conducted and key areas where performance changes came into play, namely: iSCSI immediatedata parameter iSCSIadm(1M), low latency request response, and write cache enable (WCE) setting.

In another blog entry, Bourbonnais highlights the synchronous write bias property, which he notes is significant for managing performance of Oracle databases. The logbias property takes the default value of Latency signifying that the storage should handle synchronous writes in urgency, the historical default handling. With this new 2009.Q3 software update, the logbias property can be set to Throughput signifying that the storage should avoid using log devices acceleration instead trying to optimize the workload's throughput and efficiency for datasets holding Oracle Datafiles.

"We definitely expect to see a good boost to Oracle performance from this feature for many types of workloads and configs; workloads that generate 10s of MB/sec of DB writer traffic and have no more than 1 logzilla per tray/JBOD," Bourbonnais writes.

This blog entry also covers the management of synchronous writes by the ZFS Intent Log ZIL and setting Synchronous write bias : Throughput. Bourbonnais notes that for Oracle datafiles, using the new setting of Synchronous write bias : Throughput has potential to deliver more stable performance in general and higher performance for redo log sensitive workloads. Details of logbias deployments for Oracle are explained.

More Information

iSCSI unleashed - Bourbonnais blog entry

Synchronous write bias property - Bourbonnais blog entry

iSCSI enhancements in 2009.Q3 - The blog of Bill Pijewski, a member of the Fishworks engineering team

OpenSolaris Project: COMSTAR: Common Multiprotocol SCSI Target [...read more...]

		fullsource
Print Email Current Profile Archive Blog Generate newsletter