6500/7600 Sup module performance specs
Bookmark this page:
 Yahoo!
 Google
 Windows Live
 del.icio.us
 digg
 Netscape
|
|
||||||||||||||||
|
Posted by linguafr on August 4, 2008, 1:36 pm
Please log in for more thread options including GRE and BGP (full routing table). From the router performance spec sheet the sup32 supports 15Mpps (64byte packets) which equates to 7.68Gbps. Yet the cisco rep we've spoken with claims the sup32 is underpowered for our needs and suggests the Sup720 which is 2x the price. What do you folks in the field think? Thx | ||||||||||||||||
|
Posted by Simon Leinen on August 4, 2008, 3:12 pm
Please log in for more thread options The Sup32 seems just fine for your needs. Note that the Sup32, and also the Sup720 (but not the Sup720-3BXL or the RSP720), have limited hardware forwarding tables, so they don't really work with full BGP routing (>260'000 entries) today.
Another thing that the Sup32 lacks is a switching fabric. Therefore you won't be able to use line cards with it that require the switching fabric, i.e. WS-X67xx. This basically means no possibility to upgrade to 10GE interfaces (short of upgrading to a Sup32-2X10GE or Sup720 or similar). But if you don't need any of that, you should be pretty happy with the Sup32 - it forwards IPv4, IPv6, and MPLS (including EoMPLS) in hardware. It also does GRE in hardware, provided that you use separate addresses (loopbacks) for terminating the tunnels. -- Simon. | ||||||||||||||||
|
Posted by linguafr on August 4, 2008, 3:40 pm
Please log in for more thread options > linguafr writes:
> > We need a sup module that will support up to a 1 gig of routed traffic
> > including GRE and BGP (full routing table). From the router > > performance spec sheet the sup32 supports 15Mpps (64byte packets) > > which equates to 7.68Gbps. Yet the cisco rep we've spoken with claims > > the sup32 is underpowered for our needs and suggests the Sup720 which > > is 2x the price. What do you folks in the field think? Thx >
> The Sup32 seems just fine for your needs. > > Note that the Sup32, and also the Sup720 (but not the Sup720-3BXL or > the RSP720), have limited hardware forwarding tables, so they don't > really work with full BGP routing (>260'000 entries) today. > > Another thing that the Sup32 lacks is a switching fabric. Therefore > you won't be able to use line cards with it that require the switching > fabric, i.e. WS-X67xx. This basically means no possibility to upgrade > to 10GE interfaces (short of upgrading to a Sup32-2X10GE or Sup720 or > similar). > > But if you don't need any of that, you should be pretty happy with the > Sup32 - it forwards IPv4, IPv6, and MPLS (including EoMPLS) in hardware. > It also does GRE in hardware, provided that you use separate addresses > (loopbacks) for terminating the tunnels. > -- > Simon. Thanks Simon What routes are left out once you exceed the 256K limit? Also, is this an ASIC based limit as you can install a gig of RAM in a sup32. | ||||||||||||||||
|
Posted by Simon Leinen on August 12, 2008, 12:58 pm
Please log in for more thread options
linguafr writes: >> Note that the Sup32, and also the Sup720 (but not the Sup720-3BXL or
>> the RSP720), have limited hardware forwarding tables, so they don't >> really work with full BGP routing (>260'000 entries) today. > What routes are left out once you exceed the 256K limit?
Thankfully (we only have about 40'000 routes), I don't know. Definitely an interesting question! > Also, is this an ASIC based limit as you can install a gig of RAM in
> a sup32. The FIB (hardware forwarding table) is part of the PFC complex; I'm not sure whether it is actually physically on the ASIC (EARL?). Anyway, RAM extension doesn't help; BGP will be able to handle many full feeds with 1GB of RAM, but - unless you route-filter pretty aggressively - the resulting FIB won't be able to fit in the hardware forwarding table on a PFC3[1]. It would fit on a PFC3XL[2], though, but that variant doesn't exist for the Sup32. -- Simon. [1] 192K IPv4 routes in the default configuration, can be increased close to 240K by sacrificing other types of routes. You can check using "show mls cef maximum-routes". sup32>sh mls cef max
FIB TCAM maximum routes :
======================= Current :- ------- IPv4 + MPLS - 192k (default) IPv6 + IP Multicast - 32k (default) [2] 512K IPv4 routes by default: rsp720>sh mls cef max
FIB TCAM maximum routes :
======================= Current :- ------- IPv4 + MPLS - 512k (default) IPv6 + IP Multicast - 256k (default) | ||||||||||||||||
|
Posted by Stephen on August 12, 2008, 3:58 pm
Please log in for more thread options
On Tue, 12 Aug 2008 18:58:57 +0200, Simon Leinen >linguafr writes:
>>> Note that the Sup32, and also the Sup720 (but not the Sup720-3BXL or
>>> the RSP720), have limited hardware forwarding tables, so they don't >>> really work with full BGP routing (>260'000 entries) today. >
>> What routes are left out once you exceed the 256K limit?
>
>Thankfully (we only have about 40'000 routes), I don't know. >Definitely an interesting question! On other boxes with TCAM forwarding it is much easier to get to the limit... on a Cat 3560 / 3750 once the table is full any additional routes overflow into software forwarding. >
>> Also, is this an ASIC based limit as you can install a gig of RAM in
>> a sup32. >
>The FIB (hardware forwarding table) is part of the PFC complex; I'm >not sure whether it is actually physically on the ASIC (EARL?). > >Anyway, RAM extension doesn't help; BGP will be able to handle many >full feeds with 1GB of RAM, but - unless you route-filter pretty >aggressively - the resulting FIB won't be able to fit in the hardware >forwarding table on a PFC3[1]. It would fit on a PFC3XL[2], though, >but that variant doesn't exist for the Sup32. And if you have DFCs the table gets replicated there - so another place for hardware overflows to happen? AFAIR when the box comes up it detects "lowest common denominator" for the various blades and pretends to be that. However who knows what happens when an engineer fits a non -3XL replacement card PFC / DFC and dynamically downgrades the chassis? -- Regards stephen_hope@xyzworld.com - replace xyz with ntl | ||||||||||||||||
| Similar Threads | Posted |
| 6500/7600 Sup module performance specs | August 4, 2008, 1:36 pm |
| CSS 11500 Specs | May 11, 2007, 8:22 pm |
| Looking for a radius server with specific specs | May 14, 2007, 1:46 pm |
| WAN performance | November 27, 2006, 2:10 pm |
| VPN performance | May 2, 2008, 4:00 am |
| intermitten performance on VPN | July 28, 2005, 1:38 pm |
| switch performance | April 13, 2006, 3:29 am |
| performance testing | August 9, 2006, 4:54 am |
| Poor FTP performance with 837 | October 1, 2006, 3:01 am |
| 851 and 871 Performance Inquiry | June 7, 2007, 6:09 pm |
| PIX VPN encryption performance? | September 24, 2008, 11:31 am |
| NetFlow Collector performance | September 27, 2005, 9:42 am |
| 2611 routing performance | October 4, 2005, 3:14 am |
| time-range performance | February 11, 2006, 10:03 pm |
| Help with Switches their performance and monitoring | April 26, 2006, 6:34 am |
> including GRE and BGP (full routing table). From the router
> performance spec sheet the sup32 supports 15Mpps (64byte packets)
> which equates to 7.68Gbps. Yet the cisco rep we've spoken with claims
> the sup32 is underpowered for our needs and suggests the Sup720 which
> is 2x the price. What do you folks in the field think? Thx