622 Adapters

ManualsBrandsHP ManualsNetwork HardwareHP-UX 11i v2 I/O Cards

Performance Report on ATM adapters, Part No. A5483A, A5513A, and A5515A 09/30/99 1

HP ATM 622 and 155 Mbps Adapters for HP 9000 Enterprise

Servers and Workstations

Performance Report

Product Numbers A5483A, A5513A, and A5515A

Introduction

A series of tests were run to evaluate the performance of HP ATM 622 and 155 Mbps adapters

on various HP 9000 server and workstation platforms. The tests also measured the effect of

Classical IP and LAN Emulation interfaces, extended packet sizes, and TCP configuration

parameters on the adapter’s performance.

A second set of tests evaluated the scalability with respect to number of adapters and number of

CPUs on various HP 9000 server platforms.

This document presents the performance results and provides tuning guidelines to achieve

maximum possible performance from the ATM adapters.

Appendix A provides the Test Methodology and Environment used for the performance tests.

Test Highlights

• The 622 Mbps adapter delivers bidirectional TCP throughput in excess of 1 Gbits/sec on the

N4000 platform. This is 94% of the theoretical maximum TCP throughput for a 622 Mbps link.

• The unidirectional TCP throughput varies linearly with the number of 622 Mbps Adapters. The

TCP throughput of 4.2 Gbits/sec is achieved with eight 622 adapters on N4000 platform.

• The unidirectional TCP throughput for the 155 Mbps adapter is 134.5 Mbps, which is equal to

the theoretical maximum throughput for a 155 Mbps link.

• The TCP throughput varies linearly with the number of 155 Mbps adapters. Aggregate

bidirectional throughput of 3 Gbits/sec is achieved with twelve 155 Mbps adapters on a

V2500 and N4000 platforms.

All the results above use the Classical IP as the interface with MTU size of 9180 bytes.

Single Adapter Configuration Test Results

The theoretical maximum TCP throughput is approximately 135 Mbps for the 155 adapter and

542 Mbps for the 622 adapter. See Appendix A for calculations of the theoretical maximum TCP

throughput using protocol overhead at various layers.

Table 1 below shows the TCP throughput numbers for the N4000 platform. See Tables 4 and 5

for results on other platforms.

Table 1: Maximum TCP Throughput on N4000 (One CPU , Classical IP Interface, MTU 9180 bytes)

N4000

(155 Mbps Adapter)

N4000

(622 Mbps Adapter)

Outbound Throughput Mbps

134.45 538.14

CPU Utilization %

6.5 26.85

Inbound Throughput Mbps

132.03 535.42

CPU Utilization %

5.65 23.61

Bidirectional Throughput Mbps

260.12 1005

CPU Utilization %

11.17 57.1

Summary of content (10 pages)

PAGE 1
HP ATM 622 and 155 Mbps Adapters for HP 9000 Enterprise Servers and Workstations Performance Report Product Numbers A5483A, A5513A, and A5515A Introduction A series of tests were run to evaluate the performance of HP ATM 622 and 155 Mbps adapters on various HP 9000 server and workstation platforms. The tests also measured the effect of Classical IP and LAN Emulation interfaces, extended packet sizes, and TCP configuration parameters on the adapter’s performance.
PAGE 2
The packet size (that is, the Maximum Transmission Unit [MTU] size) and the interface type influence the performance of the adapters. Table 2 below shows the maximum TCP throughput and CPU utilization figures for three different interface types for 155 Mbps adapters. Table 3 below shows the maximum throughput and CPU utilization figures for the same interface types for 622 Mbps adapters.
PAGE 3
The second set of tests consists of exchange of a single request and response packet of 1 byte each. The packet is at the TCP level. The performance metric is the aggregate number of request/response packet pairs (transactions) per second. This performance metric is an indication of the number of user level packets the adapter can process in a second. Test results show that the CPU and the platform type influence the single byte request/response performance.
PAGE 4
Multiple Adapter Configuration Test Results For a Classical IP interface with MTU size of 9180 bytes, the TCP throughput in outbound direction varies almost linearly with the number of adapters, for up to eight 622 adapters and 12 155 Adapters for V2500 and N4000 platforms. See Figures 2 and 3 below.
PAGE 5
Each netperf process was bound to the same CPU that the adapter uses to send interrupts to. The Classical IP interface on each adapter was configured with a different IP subnet; that is, there were eight IP subnets configured in an eight-adapter test. On V-series platforms, one 622 adapter was configured per PCI Controller, and a maximum of two 155 adapters were installed in a PCI Controller. On N-series platforms, the 622 Adapters were installed in the Twin Turbo slots.
PAGE 6
Table 8: TCP Throughput Results on LAN Emulation Interface (N4000:8 CPUs, 4 155 Adapters) Test/Interface Outbound TCP Throughput Mbps CPU utilization % Inbound TCP Throughput Mbps CPU Utilization % Bidirectional TCP Throughput Mbps CPU Utilization % LANE MTU 9218 bytes 538 2.75 529 4.04 772 6.8 LANE MTU 1516 bytes 475 6.3 268 3.28 338 4.
PAGE 7
Throughput–intensive applications can see an increase of about 20% when the application is bound to a specific CPU. This is done as follows: Find out which CPU the adapter is interrupting by using the sar command. With enough traffic on the adapter, the sar output should show interrupts on only one CPU. The application is bound to this CPU by using the mpctl system call.
PAGE 8
Appendix A: Test Methodology and Test Environment Test Methodology HP’s Netperf version 2.1 was used as the test tool in all the measurements. Netperf measures the throughput at the TCP level. This measurement models many real life applications since these applications use the TCP protocol. The target systems were configured with a minimum of 2 Gbytes of memory. The ATM software version used was I.11.00.10 with Build Revision 7.39; this corresponds to the DART45 release of ATM software. HP-UX 11.
PAGE 9
Table 10 Server and Workstation platform characteristics Platform CPU ICACHE/DCACHE Memory SPECint95 OS N4000 PA8500 440 MHz 0.5 MB/1 MB 1 GB3 34 11.00 May 99 V2500 PA8500 440 Mhz 0.5MB/1 MB 1 GB NA 11.00 Dec 98 V2250 PA8200 240 Mhz 2 MB/2 MB 1 GB 16.4 11.00 Dec 98 Protocol Overheads and Theoretical Maximum TCP Throughput The Throughput as seen by a TCP application is less than the OC-3 physical layer throughput of 155 Mbps or OC-12 (622 Mbps).
PAGE 10
Message size Theoretical TCP Throughput (155 adapter) Theoretical TCP Throughput (622 adapter) 128 Bytes 90 Mbps 362 Mbps 256 Bytes 103 Mbps 413 Mbps 1024 Bytes 126 Mbps 503 Mbps 4096 Bytes 133 Mbps 535 Mbps 64 Kbytes 135 Mbps 542 Mbps It should be noted that the above calculation assumes Classical IP interface. LAN Emulation adds additional 8 bytes as header. Therefore, the theoretical throughput will decrease accordingly. Performance Report on ATM adapters, Part No.