How to determine bandwidth req'd over WAN...

[maxnpj]

I've searched this forum for answers and I'm sure this question has been asked a thousand times, but I'm being asked to provide hard numbers for bandwidth usage based on very little information.

We are looking at a new piece of software which does not exist in our enterprise now. We have tested this software using a WAN simulator but only between two machines sitting in the same room. We used Wireshark to capture the data going between the two machines and then were able to determine the amount of time for particular transactions at particular "bandwidths' and latencies. The WAN simulator allows us to introduce latency and rate shaping to simulate, for example, a 128k connection with 200ms of latency.

The WAN infrastructure is also not in place for these facilities either. They currently have DSL VPN connections (which I guess is a WAN infrastructure of sorts) and I'm being asked what impact will 22 sites have on the MPLS network when these facilities are brought online. They will be using the tested software as well as additional software such as an e-mail client and a web browser.

So my question is, how can you determine bandwidth requirements for apps that you aren't sure about? For example, we know how much time it takes one machine to execute one transaction over our simulated WAN, and we can simply extrapolate that information for the same transaction from a number of different workstations, but that assumes each worker is doing the exact same thing...which is unlikely. Also, how can you determine bandwidth requirements for an e-mail client? or for a web browser? I can't possibly know how much e-mail a user may get, or how much they may use the Internet.

I'm not saying I have *no* idea, but I 'm being asked for a hard number that is a guarantee that the users experience will be fast. I don't see how you can do that unless you just go waaaay over top.

Anyhow, thanks for reading and thanks for responding.

--Brian

[spremkumar]

Hi

Defining bandwidth for an application which doesnt have proper sizing in place is pretty tough,if i am were you i would have an acceptance with the team on their exeperience (slow/fast/normal) while accessing the application, based on that you can size an approximate value(kbps) per session and then move onto upgrading the bandwidth based on no of users who all are going to access the application.

For an example as you have mentioned the 128K speed access with 200ms latency better to check out the user experience, yes its always good to have more bandwidth but in some places the prices are not so cheap to afford the kinda bandwidth required. if you dont have any financial constraints i would suggest you to provide your team or management the pros/cons of having high bandwidth which will result in high productivity and good user experience.

Coming back to your MPLS infra query create a spreadsheet with the basic info like no of users,the applications which will be used and the b/w required for each applications. Based on this calculations you can get a ball park figure which you can make use of.

Also for your mail client and your web browser it doesnt use up much of bandwidth and its pretty defined untill unless they dont download any attachemnts or anything which will increase the bandwidth rquirement. Also there are various other solution with respect to mailing to have a local server to download everything and then your local clients can access the same locally instead of all the clients getting hooked up to the central mail server using up the whole bandwidth.

regs

[maxnpj]

spremkumar;

Thanks for the response. I will take those suggestions into consideration. It's difficult at best to know how much BW a particular app uses. I think manufacturer suggestions are based on response time, not necessarily "to the packet" bandwidth measurements. The app might run great at 512k and run crappy at 128k so they go in the middle for a recommendation.

Anyhow, thanks for the reponse.