[nsp] SUMMARY: Native IOS for 6509?

[Paul Kohler]

NetFlow AS support comes in 12.1(13)E
http://www.cisco.com/univercd/cc/td/doc/product/lan/cat6000/12_1e/swconfig/nde.htm#wp1054662

NetFlow interface support also comes in 12.1(13)E via flow masks
http://cco/en/US/products/hw/switches/ps708/products_configuration_guide_chapter09186a008007e6f0.html#wp1057334

The statement "they apparently ran into some Netflow bugs in August 2003, 
but they have been fixed" is correct. We've done a lot of work to fix any 
NF bugs and increase NF enhancements on the C6500/7600 and all platforms.

Paul

At 01:54 PM 5/20/2004, John Kristoff wrote:
>A few days ago I posed the following:
>
> > I'm soliciting opinions and experiences for running Native code on a
> > 6509 with a sup720.
> >
> > The environment of interest would be running at least a dozen 6509s with
> > well populated ports containing 10/100/1000 Ethernet and an eye towards
> > 10 gig.  The network would also have multicast (PIM-SM) widely deployed.
> >
> > Ideally respondents should have been running Native for some time or
> > would have switched back to Hybrid recently due to problems.
> >
> > I'd be especially interested in replies from large .edu's and those with
> > emails ending in @cisco.com.  :-)
> >
> > If people would send replies privately I can summarize back to the list.
>
>I received 6 replies.  Each paragraph below summarizes respondent info.
>Thanks very much to each of those 6 folks.  The info was very helpful.
>
>A colleague from a Canadian research-oriented organization reports
>running Native for at least 6 months.  They use 10/100/1000 ports with
>PIM-SM and are supplying multicast to a MSFC2/SUP2 also running native.
>High load apps are passing through the box.  No reported problems or
>complaints even with mixed code between the MSFC2/SUP and sup720.
>
>A colleague from a large national/international ISP reports running the
>sup720 with native for both L2 and L3 ethernet-based solutions.  They
>are running 12.2(17d)SXB currently, but have run 12.2(14SX), 12.2(17SX)
>and 12.2(17SXA) in the past.  They maintain full iBGP routes, plus
>unicast and multicast NLRI on both IPv4 and IPv6 eBGP.  PIM-SM is also
>running.  They report things are working fine.
>
>A colleague from a large Midwestern University reports running 3 7609s
>with native with PIM-SM on each.  They have a lot of 100 Mb/s ethernet
>with gig interconnections.  they are looking at 10 gig in a year or two.
>They are not used for off-campus peering, do not use them as their RP
>and do not have IPv6 on them.  IPv6 is being planned for the summer.
>They warn of potential Netflow problems, apparently some records or
>subset of record info does not exist.  Apparently AS and interface ID
>were not there at some point, but it is not confirmed.  They report that
>for what they do with these boxes they work well.
>
>A colleague from another Large Midwestern University reports that they
>are installing 720s this summer and they are planning to run native for
>better routing features.  They currently run hybrid on sup2s.  They
>believe for edge aggregation that CatOS may be more appropriate and that
>Cisco is positioning CatOS for the edge and distribution market, whereas
>IOS will be for core switching and routing.  They are looking for
>features like MPLS and IPv6 in the core, which native has and CatOS
>probably won't get.  They also indicated that the IOS image for hybrid
>generally lags behind the native image.  They believe native and hybrid
>are relatively equal, but that CatOS has better syslog support and does
>not impact the box like the 'debug' command on IOS has a tendency to do.
>They also point out that having the common IOS interface can make
>management and the management tools simpler.
>
>A colleague from a state university in New York reports that they have
>about 7000 nodes runing off a 6513 and a 6509.  Both have sup720s running
>native. They have multicast deployed everywhere.  They report no
>complaints and are very happy with them.
>
>A colleague from a large Texas university reports that they have been
>running sup720s with native since August 2003.  They have about 12 boxes
>with more on the way.  These boxes are on a dual fiber optic star
>backbone.  They also have about 40 backbone sites that are mostly sup2
>or sup1a boxes running hybrid. There are many 10/100/1000 ports throughout.
>The "buggiest thing" they've seen on the Ethernet ports is the 'input queue
>drop' counters that constantly increment.  They report that this appears to
>be a cosmetic problem only and likely due to trunking encapsulation frames
>being incorrectly counted by old 10/100 6248 cards.  They have multiple 10
>gig connections between some of the backbone and core as well as 10 gig
>available on many boxes.  They have PIM-SM widely deployed in the backbone
>and in the campus.  They get multicast from I2 via a 622 Mb/s link.  They
>don't report any problems with 'feature skew', but they wish they had the
>'show top' CatOS command.  They feel the config of QoS was easier on Hybrid,
>but perhaps because they were familiar with it there. It took them a lot of
>work to convert their CatOS QoS config to the native IOS config.  They
>apparently ran into some Netflow bugs in August 2003, but they have been
>fixed.  They report that the sup720s have been stable and performing well
>under native.
>
>John
>_______________________________________________
>cisco-nsp mailing list  cisco-nsp at puck.nether.net
>https://puck.nether.net/mailman/listinfo/cisco-nsp
>archive at http://puck.nether.net/pipermail/cisco-nsp/