IP Traffic Optimization?

sireenmalik asks: "As part of Network Management Systems development team i am developing an IP traffic optimizer. There are both online optimization(non-deterministic) and offline optimization algorithms (deterministic). The question is if we do Online optimization on the flows in real-time, is Offline Optimization afterwards worth it, at all? Has anyone ever made quantitative measurements? Any studies undertaken? What is the improvement in network utilization? Is it significant? Any comments?"

Comments

New Slashdot Slogan

[Anonymous Coward]

Slashdot: We Do Your Job For You

Out of curiosity

[SuiteSisterMary]

If you're doing both anyway, why don't you, umm, do the measurements yourself? Because, as I'm sure you know, any conclusions drawn by anybody else will only very generally apply to any specific set up you might have.

Re:Out of curiosity

[sireenmalik]

ACtually online is not a big deal but the online stuff is on which many good brains are still struggling all over the world.its NP hard.There are many many solutions but they suffer from their problems. most of them they only work for a network upto a specific number of nodes 20-30 on most cases. Plus the the algorithms are time consuming. Very time consuming. Considering that internet traffic is very daynamic the solutions almost loose their validity!!!

coffer = (tr.verb) cynically offer

[Anonymous Coward]

Slashdot, it is an honor for you that I am able to come here to persuade you, freely and without obligation, to give me pointers to valuable research insights into network flow modelling which will help me and my corporation do something profitable and proprietary in return for which you might or might not see increases in my karma for getting my original request posted in AskSlashdot. I was of course planning to share with you all my own conclusions and insights after finishing my literature research and AskSlashdot consultation. I was also going to explain this offer in my post to AskSlashdot, thus encouraging as many of you to post your ideas as possible, but instead after the briefest pause for ethical considerations I sneakily decided to focus on maximising the value added to my product by your community. Thank you a tiny bit on a pro rata basis per contributing individual but really not very much at all in plain English.

Re:coffer = (tr.verb) cynically offer

[Anonymous Coward]

or possibly no one, including the original submitter, understood just what exactly was being asked. "Offline" equates to web servers doing gzip on pages they send; "online" is VJ header compression. Seek the knowledge of those that have been doing this stuff for 10 years now; ask not the slashdot crowd still looking for the "double your internet speed" registry trick for their new Mandrake installs.

Very Odd Question

[gilgongo]

Does anyone understand what this guy is talking about?

Why would anyone seriously put any effort into IP "optimisation" at this level? To get 5% faster download speeds? To save 2% off their monthly bandwidth bills?

What are we all going to start using 9600 modems or something? I'm all in favour of geek research and stuff, but this is pushing it. If you want to get stuck into IP programing and stuff, at least concentrate on some future thing like IP6!

G

Re:Very Odd Question

[nickmdf]

Why would anyone seriously put any effort into IP "optimisation" at this level?

Well I suppose that this person could have a very specific kind of traffic
that might benefit from an optimisation algorithm. And if that study is for something
like wireless, a small savings could matter a lot if it is consistent.

Re:Very Odd Question

[gilgongo]

Oh gawd yeah - wireless. I fully agree that needs all the help it can get.

I take back my previous post.

Re:Very Odd Question

[sireenmalik]

Think in terms of traffic engineering which is a very weekly understood concept. Instead of over-engineering your network, or throwing bandwidth, you create space by deterministically placing your flows over the network. Your objective function could be to minimise Maximum Utilization etc. The more SLAs you can pack the more money you earn!
IP6 Qos , yes interesting thought? But have you considered the fact that bandwidth will always remain an issue(my opinion) and thus optimization holds validity.
Moreover, optimization roughly translates to QOS.

Commercial or just for research?

[teambpsi]

Is this a project for which your company is hoping to generate revenue from ? Or is this "we have too much free time and money in our budget and should spend it here" ;)

Either way I'd suggest you you look at the 'competition' -- surely you would glean a lot of very useful information from the whitepapers of people like Packeteer Corp

Or if you don't have the bazillion dollars to buy their products, try a really nifty software only solution from the folks at Emerging Technologies -- their bandwidth manager is really pretty cool -- we use it to control our co-location client traffic.

Ultimately, any kind of traffic shaping you do is really about Quality of Service (QoS) issues.