HP 2424M - Manual (Page 24)

HP ProCurve Switch 4000M / 8000M / 2424M / 1600M Reviewer’s Guide
3. Performance Testing
Switches are normally performance tested under a variety of conditions for three main parameters:
throughput/packet loss rate, congestion control and latency. Definitions for throughput, packet loss
rate and latency are specified in RFC 1242. See http://www.rfc-editor.org/ . Unfortunately, there is no
clear definition for congestion control. There are numerous tests that test different aspects of
congestion control, none of which is considered the definitive test for congestion.
Keep in mind that most of these tests are performed in an artificial environment intended to measure
the outer limits of switches and rarely, if ever, reflect circumstances a switch would consistently find in
an actual network environment. This is particularly true for congestion type testing. The applicability of
these performance numbers to real world networks has to be interpreted carefully, as most switches in
these tests perform to a level significantly exceeding the needs of real world networks.
Additional comments on each of the parameters follows.
3.1 Throughput/Packet Loss Rate
Hardware is now available to the switch designer to allow designs that can handle media-speed streams
on all ports simultaneously. This does, in most cases, come at an increased price for the switch to
customers. As a result, designs for switches are a tradeoff of total throughput versus cost. This tradeoff
is most apparent for high port density switches, where the cost of providing media-speed throughput
on all ports simultaneously can be high, while the actual need for this level of throughput is doubtful.
3.1.1 HP ProCurve Switch 4000M / 8000M
This tradeoff can be seen in Appendix A in the throughput numbers for the HP ProCurve Switch 4000M
or 8000M, both high port density switches. With 40 port pairs active (i.e. all 80 ports connected and
passing half duplex traffic), the Switch 4000M or 8000M reported a 94.1% traffic throughput with 1518
byte packets and 78.5% with 64 byte packets. Forty full-duplex port pairs active at 100Mbps is the
equivalent of sending 4 Gbps through the switch. With the Switch 4000M and 8000M backplanes rated
at 3.8 Gbps (based on backplane clock speed) there are some packet drops to be expected at this traffic
level and this is seen with 1024 byte packets and larger. Backplane throughput, calculated from these
figures, is shown to be approximately 3.73 Gbps for actual throughput. The increase in drops (as shown
by the lower percentage) below the 1024 byte packet length is due to the switch engine overhead of
having to make many more forwarding decisions (as shown in the total packets column). This is
particularly evident in the 64 byte packet test.
This test shows the maximum packet per second rate for the HP ProCurve Switch 4000M and 8000M is
4.67 million 64 byte packets. This performance is excellent, particularly given the HP ProCurve
Switch 4000M and 8000M’s low cost.
3.1.2 HP ProCurve Switch 2424M
The HP ProCurve Switch 2424M has a backplane rated at 3.2 Gbps. When maximum full duplex traffic
is sent through the 24 fixed 10/100 ports and 2 Gigabit ports on an installed Stacking Module, the data
rate is theoretically 4.4 Gbps, so packet drops at this traffic level are expected. As seen in Appendix A,
the HP ProCurve Switch 2424M with 512 byte packets and above show a throughput rate of about
72.5%. This is equivalent to a backplane speed of 3.19 Gbps, which lines up nicely with the actual
backplane speed. With packets smaller than 512 bytes the throughput rate is affected the ability of the
switch to handle the higher number of packets seen with smaller packet sizes. The maximum packet
handling rate is with 64 byte packets where the HP ProCurve Switch 2424M handled 3.92 million
packets per second. This packet handling ability is more than sufficient for the typical desktop switch
environment and better than the other major vendors’ stackable solutions. Compare these numbers to
those posted in a test conducted by the Tolly Group on several stackable switches. The report,
No. 8286, is available at http://www.tolly.com/ under Custom Tests in the Test Results area of the site.
©1998, 1999, 2000 Hewlett-Packard Co Revision 3.2b – 1/15/2000 Page 24 of 36