12-19-2018 12:57 AM - edited 03-08-2019 04:51 PM
Hi All,
I'm looking at using either a WS-C3850-12XS or a C9500-16X switch for an internal layer 3 node which will be used as a Route redistribution device between BGP and OSPF. I've found in the past that Cisco often limit the number of BGP routes, possibly to force you to use a router! Does anyone know the maximum number of BGP route entries these switches will hold? Below is an extract for the 9300 switch although it's generic and doesn't specify BGP routes.
Cat9300 route entry details
Total number of IPv4 routes (Address Resolution Protocol [ARP] plus learned routes) |
Up to 64,000 indirect* Up to 80,000 host* |
Thanks in advance for taking the time to read this and hopefully provide a reply :-)
12-19-2018 02:51 AM
Hello,
below are the specs I could find. Directly connected/host routes are hosts connected via L2 to the switch (like a host on a directly connected subnet) whereas indirectly connected routes are routes that are routed via another IP address. I don't think it makes a difference whether these are BGP or OSPF or EIGRP or RIP routes...
According to the two attached data sheets, the 9500-16X supports:
Total number of IPv4 routes (Address Resolution Protocol [ARP] plus learned routes)
Up to 64,000 indirect*
Up to 80,000 host*
And the 3850
3850
24,000 (ARP plus learned routes)
Cisco Catalyst 3850 Series Switches Data Sheet
Cisco Catalyst 9500 Series Switches Data Sheet
12-19-2018 03:13 AM
Thanks for the reply Georg. I've seen the detail in the data sheets. We purchased a Nexus 5500 switch with a layer 3 card a few years ago to use at the network edge and found that although the switch had a very large route capacity it limited the number of BGP routes to 8k. I thought I'd ask around this time before choosing a switch for this project rather than a router. High throughput switches are a lot cheaper :-)
12-19-2018 03:54 AM
Hello,
I guess the switches are not really meant to be deployed on the (BGP) edge. Even the high end Nexus switches will not hold the full BGP table (which currently consists of about 750K prefixes)...
12-19-2018 05:05 AM
Hi Georg, The plan was the ISP would only send a default route. They failed initially which caused the switch to crash several times until we sorted it out. I understand your argument about using a router but when you can save thousands of dollars by using a switch for what is a simple function (BGP with one route and a few VLan interfaces), I for one would rather save the money.
thanks again for your input.
12-19-2018 05:09 AM
Hello,
if you only need the default route, the switches you mention are certainly more than sufficient. And indeed a lot cheaper...
12-19-2018 05:16 AM
Hi Georg, The previous post was an example of an old deployment. This new design will probably have anything up to 3000 BGP routes although that's a very rough estimate as I don't have access to the customers network. I need up to 10G throughput and 10G & 1G interfaces which is why I'm looking at a switch because an ASR with all the licensing will be expensive compared to a switch.
12-19-2018 06:18 AM
12-19-2018 06:39 AM
Thanks for the input Joseph, that's very interesting. Hopefully someone has come across the limits either through experience or by documentation.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide