Intel 10GbE Adapter Performance Evaluation for FCoE and iSCSI

ManualsBrandsIntel ManualsOtherIntel 10 Gigabit XF SR Server Adapter

Table Of Contents

Intel

10GbE Adapter Performance

Evaluation for FCoE and iSCSI

Evaluation report prepared under contract with Intel

Corporation

Introduction

As the interest in converged networks grows, and as the vendors from the traditional Ethernet

adapter and Fibre Channel adapter marketplaces produce converged products, questions arise

about the capabilities and performance of each type of adapter. Intel

commissioned Demartek to

compare the performance of their Intel

Ethernet Server Adapter X520 that supports Ethernet and

Fibre Channel over Ethernet (FCoE) protocols with those of the two leading Fibre Channel

adapter vendors.

Evaluation Environment

A series of application performance workloads were run and repeated for each of the three

competing converged adapters connecting servers to storage in a Data Center Bridging

(DCB)/FCoE environment. The goal of this testing was to evaluate products in environments

similar to actual customer environments. As a result, these tests were performed with well-known

disk storage arrays with spinning disk drives in FCoE and iSCSI configurations similar to those

typically found in customer datacenters, rather than testing theoretical performance with

specialized hardware not typically found in customer environments.

Evaluation Summary

In examining the results of these comprehensive performance tests, we found that for the most

part, the performance of all three brands of adapters fell into a fairly narrow range. In some tests,

one adapter had the highest performance; in other tests, a different adapter had the highest

performance.

Because the adapter performance was reasonably close in most of these tests, IT professionals need

to consider the cost of these adapters, especially in environments where many adapters are

required. In addition, Ethernet features and functions in converged adapters should be considered.

For these tests, the Intel

adapter provides the best price for performance in these real-world

application tests. In some environments, the additional cost of more expensive adapters could be

applied towards faster CPUs or more CPU cores that would benefit all applications running on

that host.



Demartek

September 2010

Summary of content (18 pages)

PAGE 1
  Demartek September 2010 Intel® 10GbE Adapter Performance Evaluation for FCoE and iSCSI Evaluation report prepared under contract with Intel® Corporation Introduction As the interest in converged networks grows, and as the vendors from the traditional Ethernet adapter and Fibre Channel adapter marketplaces produce converged products, questions arise about the capabilities and performance of each type of adapter.
PAGE 2
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 2 of 18 1 – Converged Networks and Adapters “Converged” or “unified” networks have been garnering considerable attention recently, especially as the 10-Gb/sec Ethernet and Fibre Channel over Ethernet (FCoE) standards have been made official within the last calendar year 1. However, the measured fusion of the Ethernet and Fibre Channel worlds triggers many questions among datacenter directors, managers and administrators.
PAGE 3
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 3 of 18 2 – Intel® Installation and Management Interface Installation of the Intel® Ethernet Server Adapter X520 is performed in the same manner as other Intel® Ethernet adapters. The software and drivers use the standard Intel® installation process known as Intel® Network Connections, and is available via download or on a distribution CD.
PAGE 4
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 4 of 18 Intel® uses the same management interface for its Intel® Ethernet Server Adapter X520 as it does for its other Ethernet adapters. The Ethernet settings are in the same places administrators have come to expect, while the FCoE settings have their own page. Each are accessible through the familiar “Device Manager” management console under the “network adapters” and “storage controllers” sections, respectively.
PAGE 5
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 5 of 18 DCB/FCoE The DCB/FCoE properties can be accessed from the Ethernet properties by using the “Data Center” tab and clicking on the “FCoE Properties” button, as shown below. The FCoE properties show parameters such as the World-Wide-Names (WWNs), link speed, and other basic Fibre Channel parameters that are typical in this environment. In addition, some parameters can be changed, such as the device I/O queue depth.
PAGE 6
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 6 of 18 3 – Real-world Adapter Testing For this evaluation, we performed a spectrum of application performance tests using real-world configurations of servers and storage that customers actually use in production. We used two different, well-known enterprise storage targets that use a large number of spinning disks configured as they are typically configured in customer environments.
PAGE 7
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 7 of 18 4 – Evaluation Environment The tests were conducted at the Demartek lab in Arvada, Colorado and the Intel® lab in Hillsboro, Oregon. Basic functionality tests were performed in the Demartek lab. The performance tests were performed in the Intel® lab. Servers: • • Demartek lab: Supermicro X8DTH-6F, Dual Intel® Xeon processor E5540, 2.53 GHz, 5.86 QPI, 48GB RAM, Windows Server 2008 Enterprise.
PAGE 8
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 8 of 18 5 – Performance Results Two sets of tests, Microsoft Exchange JetStress 2007 and Microsoft SQLIO, were run using iSCSI and FCoE configurations. Both performance and CPU utilization are important, so we tracked both. The fundamental performance is critical to determine if the storage can support the application at a particular workload level. The CPU utilization shows how much load that workload is placing on the CPU.
PAGE 9
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 9 of 18 In most cases, the performance numbers were fairly close for all three adapters. In some cases, one adapter performed better and in other cases a different adapter performed better. The SQLIO output includes thread count (“T”) and queue depth (“Q”) for each data point. FCoE vs.
PAGE 10
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 10 of 18 JetStress – NetApp FCoE Single Path JetStress – NetApp iSCSI Single Path JetStress – NetApp FCoE Multi-Path © 2010 Demartek  www.demartek.com  Email: info@demartek.
PAGE 11
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 11 of 18 JetStress – NetApp iSCSI Multi-Path JetStress – EMC FCoE Multi-Path JetStress – EMC FCoE Single Path © 2010 Demartek  www.demartek.com  Email: info@demartek.
PAGE 12
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 12 of 18 SQLIO – NetApp FCoE Single Path © 2010 Demartek  www.demartek.com  Email: info@demartek.
PAGE 13
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 13 of 18 SQLIO – NetApp iSCSI Single Path © 2010 Demartek  www.demartek.com  Email: info@demartek.
PAGE 14
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 14 of 18 SQLIO – NetApp FCoE Multi-Path © 2010 Demartek  www.demartek.com  Email: info@demartek.
PAGE 15
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 15 of 18 SQLIO – NetApp iSCSI Multi-Path © 2010 Demartek  www.demartek.com  Email: info@demartek.
PAGE 16
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 16 of 18 SQLIO – EMC FCoE Multi-Path © 2010 Demartek  www.demartek.com  Email: info@demartek.
PAGE 17
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 17 of 18 SQLIO – EMC FCoE Single Path © 2010 Demartek  www.demartek.com  Email: info@demartek.
PAGE 18
  Demartek Intel® 10GbE Adapter Performance Evaluation September 2010 Page 18 of 18 Summary and Conclusion In running a fairly broad set of application tests that varied parameters such as queue depth and number of threads, we attempted to reproduce many of the types of environments the customers have in their production environments.