Forum Discussion
getsbi_34190
Nimbostratus
NimbostratusViprion Scalability
Hi,
I am trying to design a solution that require a single VIP running a throughput of 52Gb. I suspect we can do it with the Viprion box but have no Idea how it will work with the device clustering. Is it possible to create a cluster of three blades processing the traffic of a single VIP. Did anyone build such a design with HA (interfaces, device, etc) built into it?
Thank you,
G
7 Replies
L4L7_53191G: the newest Viprion systems (i.e., the new Puma II blades) will be able to get you up to 72Gb/sec, so your basic requirement sounds reasonable. To hammer out the specifics of your HA setup, it's probably best to contact your pre-sales engineer at f5. They'll be able to help answer your questions. It'll also be useful to share what L7 actions you may be doing, as well as requests/new connections per second if that data is available...
-Matt
getsbi_34190Thank you Matt,
I was wondering about the feature of the SuperVIP that I heard about but couldn't find any scalability reports or any in-depth information of the feature. Do you happen to have any information or experience with the feature on Viprion?
G- L4L7_53191To me at least, the 'super vip' is really just a marketing-speak way to describe how a single VIP can be federated/represented across multiple blades and CPUs across the chassis. The meat of this idea is that Viprion has what's call the 'disaggregator', which will manage traffic flows destined for a specific Virtual Server across the chassis blades and CPUs - pretty slick stuff.
Each CPU has an instance of TMOS assigned to it. Effectively with PII blades we've now got 32 or so TMOS microkernel instances to distribute traffic flows to. This is where the disaggregator ASIC comes into play - it knows how to pin flows to a specific module/TMOS instance and spread that traffic efficiently. Think of it as a load-distribution ASIC for TMOS, hence the 'super-vip'!
-Matt - getsbi_34190But don't I have a bottleneck in the backplane? Is there an information regarding the way it scales?
- L4L7_53191I don't know of a bottleneck on the backplane per se - it'll handle more than the fully loaded chassis. You may actually lob a ticket into support and see if they can get you that information, I have a feel for what it is but not a definitive number.
-Matt - getsbi_34190Okay I will issue a ticket and try getting the information. Out of curiosity what do think the number is?
G - L4L7_53191It's around 300Gps if I recall.
-Matt
