We want to use a pair of juniper qfx3500 in virtual chassis mode and we have around 1000x SVI with 7 bgp peer and no full table and around 20gbps bps and 8m pps, can qfx3500 handle this amount of traffic with no impact on cpu ? my mean is 1000x SVI does not impact on cpu performance? Thanks
1 Answer
All the parameters you give are within the scaling capabilities of the QFX3500.
However, in several of your previous posts asking about equipment capabilities to meet your scaling needs, you later follow up asking if you can add ACLs or PBR. Those will be a problem for the QFX3500. It has an unusually small TCAM allocation for ACL -- only a few hundred entries IIRC -- and probably won't suit similar use-cases that you've asked about previously.
I wouldn't recommend virtual chassis / QFabric mode of operation on QFX3500. It's just my opinion, but the benefits never outweighed the increased bugs / malfunctions on this particular series.
Also, I'm not sure what the licensing requirements are for QFX3500 with BGP. You might need to buy something like Advanced Feature License which may make the switch a bigger investment than it's really worth, as compared to something slightly newer like a QFX5100/QFX5200 series.
-
may i have your own opnion ? if you have only 3 option , 1) n3k-c3064pq-10gx 2) juniper qfx3500 3) arista 7050s , which one do you take? if you will be able to choose only one of them. Commented Nov 3, 2020 at 20:50
-
I don't have the ACL/PBR usage in my networks (I am a consultant for mostly datacenter / small ISP networks) that you have. For my use-case, the QFX would be my preference, not because it's a great switch but because those others are just worse in a lot of ways, and Junos' config system has enough advantages for me that it outweighs the downsides. But for you, the Nexus might be a better choice; it has more flexibility for ACL/PBR/etc but in trade you have to deal with NX-OS. All these are kinda bad products IMO and you're just picking the least bad. Commented Nov 3, 2020 at 21:14