What is the recommended hardware to run the Transit Control Plane functionality? We may have several sites in a MAN topology and would like to connect them via SD-Access Transit. Curious what is recommended for production hardware for Transit Control Plane.
Just to add to this - you can find the full list of supported hardware (and the recommended/supported software versions) for Control Plane nodes within the SDA Compatibility Matrix. Select Device Role 'Fabric Border and Control-Plane'
@Jonathan Cuthbert - I noticed that the above doesn't detail the Fabric Control plane scale for the new Catalyst 8K edge platforms (8000V, 8300 and 8500). Do you know what the numbers are for these devices and will these be published in the datasheet?
Just to add to this - you can find the full list of supported hardware (and the recommended/supported software versions) for Control Plane nodes within the SDA Compatibility Matrix. Select Device Role 'Fabric Border and Control-Plane'
@Jonathan Cuthbert - I noticed that the above doesn't detail the Fabric Control plane scale for the new Catalyst 8K edge platforms (8000V, 8300 and 8500). Do you know what the numbers are for these devices and will these be published in the datasheet?
Can the Catalyst Center / DNAC datasheet, or even better the SDA design guide be updated to include the current supported list of equipment with their maximum number of endpoints. Also it appears Cisco's design recommendations keep changing, no it appears the preference is to deploy co-located Border and Control Plane nodes, and not to split Control Plane out when scaling out. Currently have a TAC case due to what appears to be LISP CPU issues on 9500H well under the reported LISP entry limit
Hi Blake, the most appropriate BN/CP design depends on a range of factors including CP architecture (LISP/BGP vs Pub/Sub), budget, total connected endpoints and the roaming behavior of the endpoints e.g. 50K wifi robots always roaming between Edge Nodes vs slow moving wireless endpoints vs completely static wired endpoints. I am aware of your TAC case and have been assisting with it. The 9500H BN+CP should be able to very comfortably handle your scale. High LISP CPU in your scenario could be a bug, or possibly a network config/design issue (less likely). Let's wait for engineering RCA please and then we can determine best course of action. Regards, Jerome
Hi @jedolphi , do you know if Cisco has published numbers on the Endpoints supported by a Catalyst 8300 used as a CP node? The datasheet mentions that an ISR4000 and an ASR1000 can support 100000 endpoints with 8 Gig of RAM and 200000 endpoints with 16 Gig of RAM. Can we expect the same numbers with a Cat8k with 8/16 Gig of RAM? Can a Cat8k with 32 Gig of RAM support 400000 endpoints?
@frlevesque , we have not published. What is your use case for 8300 as Control Plane Node please? Can you use a low cost C9K switch instead e.g. non-POE 24 port C9300? Remember that SDA Transit CP only tracks summary routes, not host routes, so vast majority of customers usually only need a few hundred or low thousands of prefixes in Transit CP which C9300 can easily satisfy. I try to avoid routers in Site CP or Transit CP role because routers will not in the foreseeable future support LISP Extranet which may be required in your network now, or in future as new services are added. Thanks, Jerome
Learn, share, save
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
