HP 2424M - Manual (Page 27)

HP ProCurve Switch 4000M / 8000M / 2424M / 1600M Reviewer’s Guide
For 100Mbps Ethernet the formula is:
LIFO = FIFO - (packet-length (bytes) * .08 µsec)
At what point latency is measured is also a variable. Some tests show latency based on a single packet
sent into the switch. Other testing has latency measured for a packet in the middle of a stream of
packets. Generally, a switch will have more favorable latency times for the mid-stream packet than for
a single packet, so how the testing is done should be spelled out.
Latency is usually measured with the NetCom SmartBits tester, a special-purpose test unit that can
measure latency down to the nearest bit time for the technology under test.
In normal testing, latency should be measured for a variety of packet sizes and all ports should be
tested since latency, being sensitive to the architecture of the switch, can vary from port to port.
Latency for the HP ProCurve Switches 4000M/8000M/2424M/1600M are very good at around 3.5-12 µsec
for a store-and-forward switch. As is seen in the figures in Appendix A, this latency varies little port-to-
port for unicast or broadcast packets. This is an ideal situation for multimedia environments where
both low latency, and low variability in latency, is important. See section 2.2.2, Backplane Design, for a
discussion on how this low latency is achieved.
3.4 Broadcast Latency
For broadcast latencies, the difference in times of the emerging broadcast packets from all ports
relative to the source port is an important consideration. The lower the time variance between ports,
the better the switch is considered to perform in the broadcast latency test. The HP ProCurve
Switches 4000M/8000M/2424M/1600M have very low variance in times between ports making these
switches excellent in multimedia and other time sensitive environments.
In addition to the actual broadcast latency data from the SmartBits tester in Appendix A, the computed
mean and average deviation is also shown. The mean shows the average latency across all ports and is
indicative of overall latency of broadcast packets through the switch. Comparing the mean to the
typical latency with unicast packets shows that the HP ProCurve Switch 4000M/8000M/2424M/1600M’s
ability to handle broadcasts is excellent.
Average deviation is the measure of deviation across all ports from the computed mean, showing the
actual variability of values between ports from the mean. The smaller this deviation, the better. The
variability seen in the HP ProCurve Switch 4000M/8000M/2424M/1600M is very low, a direct result of
the backplane architecture.
©1998, 1999, 2000 Hewlett-Packard Co Revision 3.2b – 1/15/2000 Page 27 of 36