Bandwith allocation for a site

Hi all,

This issue is a bit difficult and deservs more points but I'll try it a shot with my remaining points and will increase them when I get some points refunded..

We're planning to implement a corporate web site that will take lots of traffic and we need to estimate how much bandwith allocation is needed for optimal/stable performance, is there a tool/way to calculate such things? roughly? I know it's difficult since the bandwith will be shared by all concurrent users packets and a banwith upgrade need can be detected by slow site downloads.. how can such thing be simulated? I checked stress test tool.. it's reporting back the average response time of the targetted page.. but this is marginised since the site is tested from severeal client machines and doesn't reflect fully the production environnement.. anyways.. I'm ready for comments and collegue experience in such topic.



LVL 20
Michel SakrAsked:
Who is Participating?
TheNigeConnect With a Mentor Commented:
I'm not sure you'll ever be able to test it very accuratly with any tools....of course to find a general idea of what you need will be good.  Do your best to "guess" what bandwidth to start off with then monitor it there after to see if you need more or less.

I used to work for a small ISP with just over 5000 subscribers on dial-up and DSL.  They had 2 T1s and a fractional DS3 for a total of 9Mb/s to the backbone.  This capacity kept the usage under 60% of total.

A good tool to use for network usage graphing and monitoring is MRTG....we use it on our cisco router here.

hope that helps some
You should get yourself a Microsoft Web Application Stress
Toolkit. You can simulate any web traffic with this tool and get a detailed report for simulations.

Take a look here:

Michel SakrAuthor Commented:
I used to work for an ISP that had 3000 users and used a 512kb connection.. not to mention the hosted sites on its web servers.. I know it's hard to track.. and it needs trial and error.. and that the user's side performance depends on several possible bottlenecks along the way.. but hence.. is there a sort of for optimal performance a user should have xxkb of bandwith.. but here also.. what is optimal?.. hmm.. hard one.. especially when the commercial departement don't understand such thing.. they see a physical bandwith and at the opposit side users.. they want to know how much users can hold the bandwith without loosing performance...  and I don't want to send them the Famous "Technically not possible!" before doing some researches..
Hire Technology Freelancers with Gigs

Work with freelancers specializing in everything from database administration to programming, who have proven themselves as experts in their field. Hire the best, collaborate easily, pay securely, and get projects done right.

One of our clients once checked how much bandwidth he actually had. - To do so, he stress-tested "his" application.
Michel SakrAuthor Commented:
well that's what I did.. but it doesn't reflect the production environnement for several reasons I mentionned..

I'll double the points
I found this tidbit....

How many simultaneous connections can be served out at 28.8 Kbits/second?

50 'full' simultaneous connections (28Kbps) per T1. No more. No more. It's math - those are the numbers.

Now, let's say your typical business decides to put up a web site. They're likely to have ISDN or frame-relay, fractional T-1 or maybe even a full T-1. So, at best they should not allow more than 50 simultaneous connections (maybe up to 75, since other things may limit individual users' transfer rates, and allow a few more people to get part of the bandwidth).

.....thats at 28.8 ..... if you are talking 56k or dsl connections then the simultaneous connections drop I guess you could try to figure out how many simultaneous connections you will have at one time....or how many you want to support without the users seeing a degrade in service.....but that is tricky too...sice internet traffic is spotty and not constant.....

hope that helps some more
Michel SakrAuthor Commented:
50.. on E1
T1 will be 2048000/28800 = ~71

But this is supposed if all the users are taking the bandwith at the same time.. IN ISPs we usually multiply this number by 20-30 to see how much ISP users can be allocated.. I guess this should be the same since we will see daily sessions which can be taken similarly to ISP user sessions.. hence this is all theoretical.. it's like I'm asking you: what is the precise perimeter of the sea?

I'm gonna see how to evaluate this question.. I'll keep it open a while and set a new link to it... since it's a bit interesting

Actually, E1 is 2048KBps. T1 is 1.5MBps
Michel SakrAuthor Commented:
well yes.. but this was not the point..
Michel SakrAuthor Commented:
I found some related materials in microsoft IIS5 capacity planning issue..
I'll try to see if there's a link somewhere on microsoft for the document..
Michel SakrAuthor Commented:
The document can be found on MS TechNet Technical information CD august 2001 issue..

It has indepth analysis for IIS deployment and pre determining Bandwith...

I awarded you a B TheNige for best cooperation, I hate to give Bs but this was not the answer.. rgrds
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.