Extreme Application applaince scaling above 1 Million flows per minute for a Traffic Domain

[Antonio_Opromol]

Hi, I've seen the EDE application analytics course on Extreme's website, and I've understand the best practice explained. But I've got a questin regardind the rare cases where someone needs to manage in a single traffic domain more that 1 Million flows per minute (that seems to be the maximum numbers for a single extreme application analytics appliance (I think the same limits are also valids for the virtual analytics engine). I know that a single sistem supports until 6 analytics appliances and up to 3M/min flows, but in this case I can reach this numbers with 6 different traffic domains (6 different applainces), or with 3 applainces (each of these up to 1 M/min flows). But how to do if I want to see in OneView analytics all the 3M flows as a unique traffic domain?

[Antonio_Opromol]

Thanks Mike for the details. There is no a real project, so for me is enough your answer.

[Mike_Thomas]

Hi Antonio, the limit is set by the EMC/Netsight Server, so 3M is the limit their, and 50,000 clients.
If there is demand for something beyond this, it's up to sales/development to entertain the potential for more flows/feature request/future technology etc. We could make a feature request if their was a sales potential/current implementation issue, but a case would need to be started for that to occur, and feature request launched off of that.

[Antonio_Opromol]

Hi Mike, the question is if we need to collect from several sensors more than 3 million flows, and we want to have all these flows on the same traffic domain, how can we do these? Is not a real necessity, but only a theoretical question.

[Mike_Thomas]

I think if you are asking if 3 appliances at 1M each is okay. It is, a few things to consider.
it that the distribution from say a PV-FC-180 sensor may run at a higher CPU rate, as well as the appliances themselves as you near the maximum recommendations for the product. One should consider too the possibility of the native 1G port being oversubscribed, and using a inbound 10G port on the hardware appliance OR VM's.
So the worst case scenario is that some of the data is dropped somewhere, and than we have some missing flows to match up against, so there is a possibility of the data being imperfect. We typically only see this in support however when a VM is present and is really underpowered as a whole - with too many other systems present already on the VM's hardware, so the overall VM itself is running quite high - less a result of the PV appliances, then the quantity of VM on the server.