QLogic 10000 Series Adapter Provides Application Benefits Through I_O Caching - Manual (Page 2)

SN0430935-00 Rev. D 04/14 2
White Paper QLogic 10000 Series Adapter Provides
Application Benefits Through I/O Caching
PROBLEM AND RESOLUTION
Increased server performance, higher virtual machine (VM) density,
advances in network bandwidth, and more demanding business
application workloads create a critical I/O performance imbalance
between servers, networks, and storage subsystems. Storage I/O is the
primary performance bottleneck for most virtualized and data-intensive
applications.
As application workloads and VM densities increase, so does the
pressure on these potential hotspots and the time required to access
critical application information. Slower storage response times result in
lower application performance, lost productivity, more frequent service
disruptions, reduced customer satisfaction, and, ultimately, a loss of
competitive advantage.
A new class of server-side storage acceleration is the latest innovation
in the market addressing this performance disparity. The idea is simple:
fast, reliable, solid-state Flash memory connected to the server brings
faster data access to the server’s CPU. Flash memory is highly available in
the market and promises to perform much faster than any rotational disk
under typical small, highly random, I/O enterprise workloads.
The QLogic 10000 Series Fibre Channel Adapter from QLogic shows
increased transaction throughput from 7 to 22 times with OLTP database
workloads, while reducing application latency query (as measured
with TPC-E benchmarks). The 10000 Series Adapter is a PCIe-based
I/O device that provides integrated storage network (Fibre Channel)
connectivity, I/O caching technology, integrated flash memory, and the
embedded processing required to make management and caching tasks
transparent to the host server. The QLogic solution delivers the application
performance acceleration benefits of server-based cache without the
limitations of solutions that require separate, server-based storage
management software and operating system (OS) filter drivers.
The key benefits identified in this paper are the improvements by
the QLogic 10000 Series Adapter on the overall OLTP performance in
transactions per second (TPS). These improvements provide real-world
objective performance data to enterprise users. The primary objectives
are to optimize performance, maximize transactions, and increase the
value of the existing infrastructure, thereby reducing the costs of server,
management, and overall infrastructure.
DATABASE WORKLOAD QUERIES WITH TPC-E
TPC-E, an OLTP workload, provides a mixture of read-only and update-
intensive transactions that simulate the activities found in complex OLTP
application environments. TPC-E queries were used without publishing
any official TPC-E benchmark results. The queries exercises a wide range
of system components associated with such environments, which are
characterized by these attributes:
• Simultaneous execution of multiple transaction types that span a
breadth of complexity
• Moderate system and application execution time
• A balanced mixture of disk I/O and processor usage
SINGLE-SERVER QLOGIC 10000 SERIES ADAPTER DELIVERS
7.75 × PERFORMANCE
The single-server test setup shown in Figure 1 was configured with an
Intel
® Xeon ® Server connected to an HP ® Enterprise Virtual Array (EVA)
storage array, through the QLogic 10000 Series Adapter and the QLogic
5800V/5802V Fibre Channel Switch. Benchmark Factory
® for Databases
TPC-E runs were executed with cache levels varying in size from 0 to
100 percent of the actual SQL database size.
Figure 1. Single-Server Database Query Setup
Figure 2 shows how performance improvements are directly related to
caching the application data closer to the server processor. The range
indicated by “1A” shows a significant application performance available
with small configurations: 3 × TPS at just 20 percent cache size. The range
indicated by “1B” shows a large application acceleration available with
modest cache configurations: 7.75 × TPS at 30 percent cache size.