Performance and Recommended Use of AD222A, AD393A, and AD221A PCIe Combination 4-Gbit/s Fibre Channel and Gigabit Ethernet Cards

How We Measured GbE and Fibre Channel

Efficiency

This article highlights the AD222A, AD393A and AD221A throughput. Throughput is the

data transfer rate, or data rate -- the amount of time it takes data to move from one place

to another. In this article, it’s shown for one-way signals as well as 2-way. Throughput

measures how well programs run with a certain workload and how quickly user requests

can be handled.

This article also provides the Service Demand for each throughput test. Service demand is

the amount of time (in microseconds) it takes one CPU to handle one kilobyte of data. It is

a normalized measurement because it eliminates disparities due to differences in

quantities, types, or frequencies of CPUs. Service Demand is an important capacity

planning & performance metric that is sometimes considered when comparing different

server models.

The performance results shown in this article were measured with the netperf4

benchmarking software. Tests were run with one AD222A card residing in slot 6 in an 8-

way HP Integrity rx6600 server. Details of the systems used and the software versions are

shown in Table 3. The Gigabit Ethernet transmit, receive, and bidirectional tests were run

using netperf4 with a socket size of 128K bytes, a message size of 32K bytes, and a

maximum transmission unit (MTU) size of 1500 bytes. The Fibre Channel read, write, and

bidirectional tests were run using netperf4 with a block size of 256K bytes.

TCP Request-Responses tests were run using a request size of 1-byte and a response size

of 1-byte.

Performance will vary when this product is used on different systems or software.

NOTES:

• The observed performance results are consistent across all of the same type of I/O

slots of the system.

• The core I/O card in the rx6600 carried minimal site LAN traffic during performance

tests.

• The line rate for GbE is 1.25 Gbit/s, removing the 8b/10b encoding overhead yields

1 Gbit/s un-encoded payload. The line rate for 4-Gb FC is 4.25 Gbit/s, removing

the 8b/10b encoding overhead yields 3.4-Gbit/s un-encoded payload. The total full-

duplex un-encoded payload is therefore:

6.8 Gbit/s (FC) +2 Gbit/s (GigE) = 8.8 Gbit/s.