Average Bandwidth Usage for Genesys

[pspenning]

Hello Everyone,
I know this is a VERY broad question because of so many factors, but I am just curious to know if anyone has done bandwidth studies on their Genesys Traffic.  If so what sort of calculations have you done, if any, to determine the amount of Bandwidth your particular Genesys platform is eating?

We are in phase 1 of our installation and I am just wondering what we're up against.
We are installing the following:

Location 1:
T-Server
Framework Server
Reporting Server
Workforce Management Server (Backup)
DB Server

Location 2:
GVP
SIP Server
Load Distrib. Web.
Nuance Speech Server
B'up Reporting Server
Framework Server (Backup)
Multi-channel Router
T-Server (Backup)
Workforce Management Server
DB Server (Backup)

Location 3:
SIP Server
GVP
Load Distrib. Web
Nuance Speech Server

We have large, redundant pipes between the locations but we have a ton of other traffic so I am trying to estimate what we will need to carve out for these servers to talk.

What other factors might I need to look at to get a best guess?

Thanks for the assistance!
Perry

[tony]

Perry,

You've described your environment but not the intrinsic value of your throughput.  The basis for data traffic throughput would be along the lines of how many calls to GVP, how many calls Self Server, how many calls reach Agents, what data are you passing with the calls, etc. You also mention MCR, so you would need to determine what data comes in on what channels...

There [i]are [/i] sizing documents available from Genesys.  Perhaps you could approach your product/account manager for a copy of the templates which you can complete to determine your expected throughput for each component/solution?

Tony

[pspenning]

Thanks Tony,
We had our weekly meeting with our install team yeasterday and I posed these questions as well.  They are getting me some information based on their installation processes and procedures.  I posted the question here because it's always good to get other input from the outside.  You have helpped greatly on pointing me in the direction of things to look at.  I appreciate it.

Thanks,
[glow=red,2,300]Perry[/glow]

[tony]

My pleasure! 

[victor]

Looks like a pretty standard layout.

Obviously, since you have SIP Server, it means you are using VoIP, and this will wreck havoc on your whole infrastructure unless you have QoS. Setting number of operators and number of calls and the rest aside, you will need to focus on three distinct categories:

1. CPU-intensive resources
2. connection intensive resources
3. traffic intensive resources

[b]CPU-intensive[/b]
I am always very weary of statserver, since they always tend ot eat the most of my CPU.
GVP is another product that is not too optimized and usually runs at pretty high CPU rates.

[b]Connection Intensive Resources[/b]
These applications experience high rates of connection request and have many clients connected to them at the same time.  SIP Server and MAYBE StatServer.

[b]And last but not least, traffic intensive!
[/b]
[u]SIP Server: [/u]
SIP Server usually has a lot of clients connected to it and probably gets the most signaling traffic out of all Genesys applications. Think of it as your gateway - it is dumb, it is there, and usually is the cause of all your timeouts or delays.

Here is what we use usually to calculate bandwidth for our t-server:

softphone clients - 1K/s per client
statserver - 10K/s + 1K/s PER softphone client
URS - 10K/s + (size of attach data in K)*number of t-server requests (attach, route)*2 PER strategy *　calls per second (per URS)
gateways: 1K*CPS (per gateway)

total is the bandwidth we would reserve for our t-server. It might be a total overkill for many installations, but we seem to be doing well with this.

[u]URS bandwidth is much easier:[/u]

10K/s + (size of attach data in K)*number of t-server requests (attach, route)*2 PER strategy *　calls per second (per URS) * number of URS in HA.


[u]Stat Server:
[/u]pretty hard to pipoint the exact bandwidth, because of number of Stats in templates and number of templates open, but here is what we use:

statserver - 10K/s + 1K/s PER softphone client + 100K/s PER CCP * number of templates open in one CCP * CCP REFRESHES PER SECOND (if you update your stats every 3 seconds then it is 1/3 = 0.333)

Usually statserver bandwidth will need to be re-adjusted after you know more about your reporting templates and how people use their CCP.


In most instances, we found this a bit of an overkill, but we would rather be on a safe side, because the wya you design your network is to make sure that VoIP traffic takes precedence, and more than once we ended up having our attach-data requests dropped simply because there was not enough room for all of RTP packets we were getting during the peak hours.  And that is DEFINITELY something you would not want to do. :>

I hope it helped,
Vic