Fabric OS FCIP Administrators Guide v6.4.0 (53-1001766-01, November 2010)

76 Fabric OS FCIP Administrator’s Guide
53-1001766-01
WAN performance analysis tools
4
Ipperf performance statistics
The following table lists the end-to-end IP path performance statistics that you can display using
the portCmd ipperf command and option.
Starting an ipperf session
Typically, you start the WAN tool before setting up a new FCIP tunnel between two sites. You can
configure and use the
--ipperf option immediately after installing the IP configuration on the FCIP
port (for example, IP address, route entries). Once the basic IP addressing and IP connectivity is
established between two sites, you can configure
--ipperf with parameters similar to what will be
used when the FCIP tunnel is configured.
The traffic stream generated by the WAN tool ipperf session can be used for the following
functions:
• Validate a service provider Service Level Agreement (SLA) throughput, loss, and delay
characteristics.
• Validate end-to-end PMTU, especially if you are trying to eliminate TCP segmentation of large
Fibre Channel (FC) frames.
• Study the effects and impact FCIP tunnel traffic may have on any other applications sharing
network resources.
To start an
--ipperf session, you can use any port as long as the port (in combination with local
interface) is not in use. You must run the
--ipperf client on both the host (source mode, -S option)
and receiver (sink mode, -R option). See “Ipperf options” on page 78 for more information about
specifying source and sink mode.
1. Configure the receiver test endpoint using the CP CLI.
The syntax for invoking the receiver test endpoint using
--ipperf for slot8, port ge0 on an
FR4-18i is as follows:
portcmd --ipperf 8/ge0 -s 192.168.255.10 -d 192.168.255.100 -R
TABLE 13 WAN tool performance characteristics
Characteristic Description
Bandwidth Indicates the total packets and bytes sent. Bytes/second estimates are maintained as a
weighted average with a 30 second sampling frequency and also as an average rate over
the entire test run. The CLI output prints the bandwidth observed in the last display
interval as well as the Weighted Bandwidth (WBW). BW represents what the FCIP tunnel /
FC application sees for throughput rather than the Ethernet on-the-wire bytes.
Loss Indicates the loss estimate is based on the number of TCP retransmits (assumption is
that the number of spurious retransmits is minimal). Loss rate (percentage) is calculated
based on the rate of retransmissions within the last display interval.
Delay Indicates TCP smoothed RTT and variance estimate in milliseconds.
Path MTU (PMTU) Indicates the largest IP-layer datagram that can be transmitted over the end-to- end path
without fragmentation. This value is measured in bytes and includes the IP header and
payload.
There is a limited support for black hole PMTU discovery. If the Jumbo PMTU (anything
over 1500) does not work, ipperf will try 1260 bytes (minimum PMTU supported for FCIP
tunnels). If 1260 PMTU fails, ipperf will give up. There is no support for aging. PMTU
detection is not supported for active tunnels. During black hole PMTU discovery, the BW,
Loss, and PMTU values printed may not be accurate.