Posted on September 22, 2015 7:27 pm
 |  Asked by Chintan Shah
 |  8159 views
RESOLVED
0
0
Print Friendly, PDF & Email

Hi,

Can some one help me to understand the difference between these two ? Say if I configure forwarding-table partition with option, I should get 128K L3 routes scale but then routing-table options show comparatively very low value. In addition this, Is there any limitation on LPM scale for v6 routes with prefix length /65-/127. I have seen this limitation in other vendors and can impact the scale for v4 and v6 (with prefix <65).

Switch1(config)#platform trident forwarding-table partition ?
0 288k l2 entries, 16k host entries, 16k lpm entries
1 224k l2 entries, 80k host entries, 16k lpm entries
2 160k l2 entries, 144k host entries, 16k lpm entries
3 96k l2 entries, 208k host entries, 16k lpm entries
4 32k l2 entries, 16k host entries, 128k lpm entries

Switch1(config)#platform trident routing-table partition ?
1 16K IPv4, 6K IPv6 (prefix len up to /64), 1K IPv6 (any prefix length)
2 16K IPv4, 4K IPv6 (prefix len up to /64), 2K IPv6 (any prefix length)
3 16K IPv4, 2K IPv6 (prefix len up to /64), 3K IPv6 (any prefix length)
4 16K IPv4, 0K IPv6 (prefix len up to /64), 4K IPv6 (any prefix length)

Thanks,
Chintan

Marked as spam
0
Posted by Jose Reyes
Answered on September 22, 2015 10:25 pm

Hi Chintan.

Thanks for getting in touch. Arista switches that use Trident ASICs (7300X, 7250X and 7050X series) support a feature called Algorithmic Longest Prefix Match (ALPM). ALPM is a way to extend the shared memory table (UFT) to store Longest Prefix Match (LPM) routes instead of Host routes. LPM routes don’t have the full /32 prefix length. For more details about UFT, you can take a look here.

But going back to ALPM. The route lookup is first done on Host table, using a /32 mask for IPv4 or /128 for IPv6. This lookup is based on the key (VRF, Destination IP address). If there is match for the destination IP address for the VRF in the Host table, the forwarding information obtained from the Host table will be used to route the packet.

In case the lookup in the host table fails, the switch will do a longest-prefix-match lookup on the LPM (TCAM) table. This table has as last resort a default route. The result will point a bucket in the UFT (ALPM) table. A bucket is just a smaller subset of potential routes into which the hardware will do a lookup. If there is a match resulting from this lookup in the UFT (ALPM) table, the resulting information will be used for forwarding. If there is a miss, then the results form the LPM (TCAM) table will be used for forwarding.

So you have three tables: host, LPM (TCAM) and UFT (ALPM).

To enable ALPM, issue the command ”platform trident forwarding-table partition 4”. IMPORTANT NOTE: this is a service impact command that causes the forwarding agent to restart.

Going back to your original question, the output of the command ”show platform trident routing-table partition” is listing details about the LPM (TCAM) table, as seen below:

Leaf-2#show platform trident routing-table partition
Max number of routes in the routing tcam table:
IPv4 routes: 16K
IPv6 routes with prefix length up to /64: 4K
IPv6 routes with any prefix length: 2K

If you enable ALPM, to verify the size of the UFT (ALPM) table,  you can use the command ”show platform trident l3 software alpm-table”.

More details about ALPM can be found in the 4.14.5F Transfer of Information (TOI) document, located in http://www.arista.com –> Software Download –> Tab ”TOI”.

Hope this clarify your question.

Regards,

José Luis Reyes
Systems Engineer LATAM

0
Posted by Chintan Shah
Answered on September 24, 2015 7:03 pm

Hi José,

Thank you for detailed response and explaining difference between all 3 tables – Host, LPM, ALPM.
I see that by enabling ALPM , switch can support maximum 128K LPM routes but when I look at ALPM table it shows 89/393216, does it mean that it can support up to 393216 routes in ALPM table ?

show platform trident l3 software alpm-table
ALPM table mode: 4, table usage: 89/393216
IPv4 routes: 89/393216

Also, can you please also clarify if there is any limitation specially when we use IPv6 in regards to prefix length ( e.x. IPV6 prefix with prefix length between /65-/127) that can affect the maximum routes ALPM can store ?

Thanks,
Chintan

Post your Answer

You must be logged in to post an answer.