OK, first lemme just ask this one question:
Are ALL traffic generated from browsing the net go through port 80 (HTTP)? Oh, and HTTPS for secure sites (what port number is that again)? No download, FTP, nothing else. Just plain simple browsing.
Now, to lay out the background situation:
I have about 40 clients in my office, and I want to limit the bandwidth used when the clients browse the net (no downloads allowed when browsing). Let’s say that I have 1.5MB total bandwidth, and I want to allocate 512k to web browsing.
Which one is the better practice according to you:
Setting the bandwidth to be shared by all clients (512k for all 40 clients)
Setting a limit for each client (say, 16k/client)
My office have about 10 programs that needs to connect to their maker’s servers to download updates & data files on a regular basis. The problem is, some of these programs use port 80 to download these patches, and I do not want them to use the 512k that is allocated for browsing. For this purpose, I want the programs to use the remaining 1MB bandwidth. In fact, I don’t want to limit them in any way, they can use whatever bandwidth is available (of the 1MB) to get these patches.
Now on to the question:
How do I set this up? I’ve tried reading the manual, but I still can’t get a satisfying result…
Would suggest simple queue with burst. Set max limit for 16k and burst limit 1 meg. Time around 10 sec with burst threshold 512k. This will cause your web pages to open up snappy but as soon as you open a web page for more than approx 5 sec (download) you will be limited to 16k. Only down side is you wont be able to brows when downloading. Of cause, you have to play with values that is best for you. Also will have to set up queue for each user.
@ashish:
OK. That’s what I got too from reading similar previous posts.
Well, back to the manual then.
Just one question though:
Like I said, there will be 2 kinds of traffic going through port 80 (HTTP): the regular web browsing and the patches. Can I use queue tree to separate these two? Or is this where the packet marking comes into play?
@Egate:
That’s what I originally had in mind. But using that method, wouldn’t the patches that uses port 80 be slowed down as well once it went past the 10-sec threshold?
My understanding of queues was for quality control, not limiting, why not make a rule that gives priority to the server/computers IP address that gets these updates, so when there is NO update traffic going on, the rest of the computers will have full bandwidth (shared) and when the servers get the updates they have priority over other traffic.
But the real question is, how many updates can their be in a day, if the software constantly gets updates during the day, I would suggest getting rid of it… sounds like a microsoft product!
