F5 VM Sizing

Thanks for the reply, Torrance

the above para mentions that there are the limits for the vCPU and the throughput that is the one of the aspect for sizing, but my question was related to the datasheet snapshot

"for example if I have 1 application which is receiving 3000 L7 request/sec , and L4 2000 request /sec then i can purchased a license of 25 MBPS, that will suffice for my application."

Thanks

Dear F5 Expert

Anybody can reply to my query, i hope it is a generic question that i am asking.

The difficultly you have is it depends.

L7 traffic means you are doing some level of inspection, be that header inspection to TLS encrypt and maybe decrypt. L4 just means you are bring traffic in and passing it on with a much lower inspection level.

You also have to consider throughput or bandwidth, so this could be 5 connections but each one is pulling 10Mb each or 10000 connections running at 10k each. All of these use the resources of the big-ip and the hardware or virtual hardware its sitting on.

The more you do with it the harder you will push it, and there will come a time when you hit a limit.

Some of these might be like latency, so how long the big-ip takes to process the traffic some times this can be boosted by boosting the ve's resources up to a point. So more RAM or CPU or maybe making sure its on a fast ssd for example.

Other things like connections and bandwidth come over time and over usage. So sizing is a bit difficult.

Have you got running services already you can compare to get a benchmark?

If not i'd suggest looking at the ranges like 25Mb, 200Mb and 1Gb normally there is a pretty good costing model around these and go in low and fully explain a growth / upgrade path for if and when you get there.

Hope that sort of helps.

Dear thanks for your reply , u took ur precious's time to reply my post I really appreciate the same.

I have understood ur post 40 % on the sizing what need to be addressed when sizing, but lets leave everything aside can according to the f5 VM Datasheet snapshot can u explain me what f5 is trying to explain in that snapshot.

Regards

Dear Experts,

I m not asking something exclusive it's all a VM datasheet explanation for the snapshot.

Hi,

So what f5 are trying to give you is a rough idea of how much through the appliance can take,

Not on the left the size of server that's used and the notes about the NIC used all important when trying to compare like for like.

What I was trying to say is this is rule of thumb here, and you need to go in with some assumption which you need to validate.

So bandwidth and connections per second are key.

Bandwidth gives you a limit due to licensing.

Connections per second is key but is benched against the performance of the server or virtual configuration you are using.

If connection go up, the resource of the box will go up, as it start to saturate the server your performance will drop. This is then easily to fix in the first instance, up the virtual cores or memory to give more resources. This should then lower the latency again of your traffic and get you operating again.

So like server management and performance management in general monitoring and visibility is key to understanding what you are dealing with.

The the realistic approach of you can design to you are sick of doing it and getting the approvals but the moment the system is turned on your calculations go out of the window as you now have live data to benchmark this on.

So really the question is, have you got bandwidth and connectivity info from your app teams which you can overlay onto the f5 data?

Just be mindful that F5 VE's treat everything hitting the interface as part of the license limit, does not matter if the destination is your actual F5.  If you are not putting them on their own VLAN it will constantly hit licensing limits for any limited bandwidth licensing.

Edit: This has somewhat beed remediated since my last look at this issue. See the important notes on How the BIG-IP VE system enforces the licensed throughput rate (f5.com)