sFlow, CEE and FCoE

From: Peter Phaal <peter.phaal@inmon.com>
Date: 10/14/09
Message-ID: <E4130EE32F294DAB8D9DC95AE9BFE182@PHAALPC>

sFlow is currently able to report on FCoE traffic since fiber channel
addresses, message types and traffic volumes can all be obtained by decoding
and analyzing the sampled packet headers.

However, fiber channel has its own forwarding logic and the addition of an
extended flow structure to capture the forwarding decision associated with
the FCoE packets would be a useful extension.

Here are some initial ideas, based on a reading of RFC 4625
http://www.rfc-editor.org/rfc/rfc4625.txt
Note: This RFC is quite old, so there may be a more recent specification
that should be used as the basis for this specification.

/* Extended Fiber Channel Routing Entry */
/* opaque = flow_data; enterprise = 0; format = 1018 */
/* See RFC 4625 */

typedef unsigned int fc_address; /* 24 bit fiber channel address,
                                    most significant byte = 0 */

struct extended_fc {
 unsigned int src_mask_len; /* Source FC address mask,
                                see t11FcRouteSrcMask
                               (expressed in number of bits) */

 unsigned int dst_mask_len; /* Destination FC address mask,
                                see t11FcRouteDestMask
                                (expressed in number of bits) */

  fc_address next_hop; /* Next hop FC switch
                                see t11FcRouteDomainId */

  unsigned int metric; /* most significant byte,
                                most significant bit = t11FcRouteType

                                least significant 7 bits = t11FcRouteProto,
                                least significant 3 bytes = t11FcRouteMetric
                             */
}

Note: the information content is very similar to the extended_router
structure already in sFlow.

Does anyone have comments? Are there other types of information that could
be usefully exported in a converged enhanced Ethernet environment?
Additional counters perhaps?

Peter
Received on Wed Oct 14 16:57:31 2009

This archive was generated by hypermail 2.1.8 : 02/17/10 PST