QLogic 2600 Series 16Gb Gen 5 Fibre Channel for Database and Business Analytics - Manual (Page 3)

SN0430953-00 rev. a 07/13 3
White PaPer
Figure 1. Port-Level Traffic Isolation
a series of iometer runs were performed, first using one port and then both
ports. the QLogic adapter scaled linearly as the second port was added to
the workload.
0
200,000
400,000
600,000
800,000
1,000,000
1,200,000
1,400,000
Single Port Dual Port
IOPS
Single to Dual Port
Scalability
512 Byte 100% Reads
Figure 2. Scalability Across Ports for the QLE2672 Adapter
QLogic’s isolated aSiC design architecture provides the following benefits,
which are key requirements for enterprise deployments of Fibre Channel
SaN infrastructures:
• Independent Functionality. reset or error recovery on one port does
not impact the performance of the other port.
• High Reliability. a firmware crash on one port does not crash the
other port.
• Best-in-Class Security. isolation at the physical level: Physical
Function (PF) for one port cannot access state information (registers,
memory info, etc.) of another PF due to physical isolation.
• Predictable Performance. i/O spikes on one port do not affect the
performance of the other port.
Real-World Enterprise Workload Analysis
to test the performance of the QLogic 2500 8Gb and QLogic 2600 Series
16Gb Fibre Channel adapters in a real-world environment, QLogic ran a
data warehousing (tPC-h) workload, not for the purpose of publishing
official benchmark results, using Microsoft
® SQL Server ® in a Microsoft
Windows Server ® 2012 environment.
Workload Overview
the read-intensive database workload consisted of a suite of business
oriented ad-hoc queries and concurrent data modifications. the queries
and the data populating the database had broad, industry-wide relevance.
the workload illustrated decision support systems and business analytics
that examine large volumes of data, execute queries with a high degree of
complexity, and provide answers to critical business questions.
Workload Database Queries
the workload consisted of a series of twenty-two unique queries that
exercised different sections of the database. each query placed a different
load on the combined CPU, memory, and storage resources. Because
each repetition of the workload performed the same work, the time to
complete the process was shorter for faster configurations of CPU, memory,
and storage.
real-world Variable i/O rates vs. Synthetic Workloads
Synthetic workloads, such as those produced by the iometer, serve a useful
purpose and are designed primarily to exercise the storage system and
its interface, with the ability to place a steady workload on the storage
system. however, a real database application workload requires varying
amounts of CPU, memory, and storage throughout the normal course of
activity, depending on the particular transactions being processed. as a
result, the i/O activity will vary during the progression of the workload. the
tPC-h benchmark provides a more accurate and real-life determination
of how choosing the right Fibre Channel adapter can significantly improve
real-world database analytical performance.
Performance Results
For the tPC-h tests, the time to complete the workloads and the i/O
bandwidth rates were measured during the tests. the data was captured
using the standard Performance Monitor (PerfMon) tool provided in the
Windows Server operating system.
as there are large, existing investments in 8Gb Fibre Channel infrastructure
today and 16Gb Gen 5 Fibre Channel technology is being adopted, this
analysis includes various combinations of performance tests using 8Gb and
16Gb Gen 5 Fibre Channel technology.