Latest Posts



Translate

Total Pageviews

Thursday, 10 October 2013

Do I choose the PVSCSI or LSI Logic virtual adapter on ESX 4.0 for non-IO intensive workloads? (1017652)

Details

VMware evaluated the performance of PVSCSI and LSI Logic to provide a guideline to customers on choosing the right adapter for different workloads. The experiment results show that PVSCSI greatly improves the CPU efficiency and provides better throughput for heavy I/O workloads. For certain workloads, however, the ESX 4.0 implementation of PVSCSI may have a higher latency than LSI Logic if the workload drives low I/O rates or issues few outstanding I/Os. This is due to the way the PVSCSI driver handles interrupt coalescing.

One technique for storage driver efficiency improvements is interrupt coalescing. Coalescing can be thought of as buffering: multiple events are queued for simultaneous processing. For coalescing to improve efficiency, interrupts must stream in fast enough to create large batch requests. Otherwise, the timeout window will pass with no additional interrupts arriving. This means the single interrupt is handled as normal but after an unnecessary delay.

The behavior of two key storage counters affects the way the PVSCSI and LSI Logic adapters handle interrupt coalescing:
  • Outstanding I/Os (OIOs): Represents the virtual machine’s demand for I/O.
  • I/Os per second (IOPS): Represents the storage system’s supply of I/O.
The LSI Logic driver increases coalescing as OIOs and IOPS increase. No coalescing is used with few OIOs or low throughput. This produces efficient I/O at large throughput and low-latency I/O when throughput is small.

In ESX 4.0, the PVSCSI driver coalesces based on OIOs only, and not throughput. This means that when the virtual machine is requesting a lot of I/O but the storage is not delivering, the PVSCSI driver is coalescing interrupts. But without the storage supplying a steady stream of I/Os, there are no interrupts to coalesce. The result is a slightly increased latency with little or no efficiency gain for PVSCSI in low throughput environments.

The CPU utilization difference between LSI and PVSCSI at hundreds of IOPS is insignificant. But at larger numbers of IOPS, PVSCSI can save a lot of CPU cycles.
 
 Source:-