About throughput

[Michaelyang]

Hello,

When I went to check the throughput of F5, I found a lot of information as shown in the figure below :

I tried to understand through the official F5 link below, but there are still a few questions I would like to ask
https://support.f5.com/csp/article/K50309321

1. If I just look at the throughput of the client coming in to F5, the service value and In value of Throughput is not there, right? Should I be looking at the file with the In values for the TMM client?

2. What is the meaning of multiplying 8 within the formula? Why is it multiplied by 8 ?

3. I see are per second capture value, but the above has current time, interval how to bring into the formula to calculate the?

Any help is appreciate.

[mihaic]

1.  That article is pretty good at explaining. 
The traffic flow from a client is seen as In to a virtual server and In to a server. Likewise, the response from the server is seen as Out from the server and Out from the virtual server.

Client IN -> Server IN

Client Out <- Server Out

2. Because the counters are in bytes example sysStatClientBytesOut , to convert it to bits , it is multiplied by 8. Because a byte has 8 bits

3.  The counters have predefined intervals , like this for example:

sysStatClientBytesOut1m

sysStatClientBytesOut5s

sysStatClientBytesOut5m

[Michaelyang]

Hi mihaic,

Thanks for your reply.

About the counters have predefined intervals,
What is <interval> value in "tmsh show sys performance throughput detail" ?

[Michaelyang]

Hi mihaic,

I found Throughput In is  [<Clients bits in> + <Server bits Out>]

However, I found that according to the data on the table, there is a big difference in adding up to each other ...
( 15.3k + 7.9k = 208.1 k ... ? )

[mihaic]

I think those values are like a snapshot. They are the values at the moment of the capture, let's say.

[Michaelyang]

Hi mihaic,

Thanks for your reply.

So Current's interval is 1 ... ?

[mihaic]

probably it is 1 s. Not sure.

Client IN -> Server IN

Client Out <- Server Out

15.3k + 7.9k = 28.2 K    not 208 K

[Michaelyang]

Hi mihaic,

Thanks for your reply.
But why does the picture show 208.1K...

[mihaic]

well it is written in that article.

The total throughput in and out of the BIG-IP system collected from all interfaces, including traffic processed by all Traffic Management Microkernel (TMM) and Packet Velocity ASIC (PVA)

• In: The ingress traffic to the system through its interfaces
• Out: The egress traffic from the system through its interfaces

In : means total ingress traffic on all interfaces except management.

[Michaelyang]

Hi mihaic,

yeah...But my F5 is VE version...
So there should be no need to take into account the PVA piece...

It looks like TMM Client-side throughput already has the sum of all traffic...
So it should be 28.2 K, but it is showing 208.1 K... 

[mihaic]

You are comparing the value of 208 K which is all ingress traffic on all interfaces

with a Value of 28K that something IN and something OUT.

What are you trying to do? I don't think adding things will give you something that you'll expect.
That's because of  different things that affect the throughput. See below.

Have a look at the number of packets IN and packets OUT. You will see that you have more than twice the number of packets coming in than going out. 

I don't think it works that way. you have throughput in and throughput out.

Also be aware of different options that affect the throughput, like oneconnect, HTTP cache, compression, irules

https://support.f5.com/csp/article/K9077