Stacking Capacity Ratio

[Mrxlazuardin]

Hi,

Stacking port capacity to access port capacity seem to be lower on recent products. It is even lower than MLAG ISC which only limited to the number of LAG active ports of highest available ports. But of course MLAG doesn't have single control plane. Any thought of having better stacking capacity ratio? Or maybe Extreme PM can explain the reason of lower stacking capacity ratio.

Best regards,

[Mrxlazuardin]

Hi Grosjean,

I don't really get your meaning. Can you explain more? I'm thinking why don't ExtremeOS support more stacking ports for recent products or at least same to MLAG ISC active links limitation for getting better stacking capacity ratio. I think it will prevent bottleneck of interswitch traffic.

Best regards,

[Mrxlazuardin]

Hi Grosjean, Yes, I'm thinking of stacking X770 or X870 for spine switching. I like X670 way which give me 1:3 stacking ratio with SummitStack320. That's why I'm curious why there is something likely with recent switch. Best regards.

[Stephane_Grosj1]

Hi,

I meant recent features allow you to keep local dual-homed traffic. Say you have a stack of 2 with several LAGs spread across the 2 units. In the past, traffic coming to the stack from a LAG and needing to flow through another LAG could use the stacking links (hash decision). Now you can keep it local and not use stacking links, reducing the effective traffic ratio.

So, recent platforms, eligible for that feature, can have a better ratio. Also, a platform like the X450G2 have a higher stacking speed than previous generation. So, all in all, for that platform, I disagree with your statement.

ISC in MLAG is just a regular LAG, that doesn't usually need to be that big, btw. I wouldn't compare a LAG with stacking technology. I understand your point to be able to have more stacking links, but I'm not sure if that's possible on the hardware level: this is not ethernet flowing on those links. And if we are speaking of a platform like x870, do you really want to do a stack?

[Stephane_Grosj1]

Hi, Can you please develop that statement? Since a few releases (I'd say 21.1 worth checking) you have the ability to do local switching in a stack, for traffic eligible (thinking dual-homing here), so that would help recent platforms to have effective better stacking links usage. I'd agree, however, that MLAG is more efficient, as you don't require in steady state a big ISC link. Assuming dual-homing, of course.