The round robin algorithm may not appear to work correctly in a Cisco
CSS 11800 series content services switch. An example of such a problem could
occur when you issue the show summary command. The
results may not show an even distribution across servers.
There are no specific requirements for this document.
The information in this document is based on the following hardware
The information in this document was created from the devices in a
specific lab environment. All of the devices used in this document started with
a cleared (default) configuration. If your network is live, make sure that you
understand the potential impact of any command.
For more information on document conventions, see the
Cisco Technical Tips
The CSS 11800 combines centralized processing and memory resources for
policy-based flow setup with distributed network processors that support
wire-speed flow forwarding.
The CSS 11800 Content Policy Engine consists of the following:
By distributing the incoming requests across the four SFPs, the CSS
11800 can handle high volumes of flows without sacrificing performance.
A CSS 11800 that is populated with two switch fabric modules (SFMs)
gives the system a total of four SFPs, two of which are on each SFM. Incoming
connections are distributed across the four SFPs for initial flow setup for
greater overall performance. The exact distribution across SFPs is established
by performing an X-OR on the protocol source and destination ports of the
incoming request and binding it to one of the SFPs based on the last two bits
of the result.
This distribution results in a statistical balance across the four
SFPs, each of which runs its own instance of the load-balancing algorithm. In
most cases, a perfect round robin load-balancing occurs because most
applications use sequential ports, which results in perfect distribution across
It may be possible to hit a different SFP, which could result in the
same server being hit in succession because a separate instance of the
load-balancing algorithm is being used per SFP. While this behavior can be
noticed if a small number of nonsequential streams are used, it should result
in better distribution as more streams are added. In real-world scenarios in
which there are many flows, all four processors are used. Distribution across
the servers should be fairly even.