x
larytet 12/5/2012 | 3:34:22 AM
re: Allot Leaps on Nasdaq Debut Two most popular P2P applications eMule (eDonkey network) and Azureus (BitTorrent) implement strong and weak packet encryption. Filetopia had it from the beginning

There are couple of ways to limit P2P. I am not sure that DIP is efficient one.
ISP, for example, can charge for BW on per byte basis. Another approach is to count number of simultaneous TCP connections/streams to different IP addresses and blocking and traffic shaping "heavy users".

Deep packet inspection is going to be a hard way to figure out what is going on in the network. Allot fights itself. The more equipment the co sells more popular encrypted P2P will become.

Interesting link here

http://www.azureuswiki.com/ind...

"Avoid traffic shaping"

Since more and more users are affected by overly aggressive traffic shaping Azureus implements this traffic obfuscation feature to allow users to use their bandwidth properly. Since this only works when a shaped peer can connect to any peer in the swarm with the crypto header it is currently not possible to turn this feature off.
multithreaded 12/5/2012 | 3:34:20 AM
re: Allot Leaps on Nasdaq Debut Identifying P2P flows is only part of offering from DPA. Other benifits include (1) user profiling and targeted Ads. delivery; (2) QoS guarantte for triple-play services; (3) protection the ISP network from DOS attack.



canadian 12/5/2012 | 3:34:18 AM
re: Allot Leaps on Nasdaq Debut How do they provide the targeted ad delivery?
mr zippy 12/5/2012 | 3:34:18 AM
re: Allot Leaps on Nasdaq Debut Identifying P2P flows is only part of offering from DPA. Other benifits include (1) user profiling and targeted Ads. delivery; (2) QoS guarantte for triple-play services; (3) protection the ISP network from DOS attack.

And by the sounds of that (possibly sales or marketing) spiel, you'd also get a single point of failure, because all traffic has to go through one of these boxes to do all of those things successfully - or are these boxes replacements for routers / MPLS LSRs ?
larytet 12/5/2012 | 3:34:18 AM
re: Allot Leaps on Nasdaq Debut Theoretically ISP can "spy" what their customers do and allow 3rd parties to use the collected info for targeted ads. For example, if you run (non-encrypted) search in Google for mortgage your ISP (and government) knows (can learn) about this. And so on.

Another approach to the problem is not looking into the HTTP headers (can be costly), but instead classify the users by amount of bandwidth, number of simultaneous TCP connections, peak hours, etc. In some cases ISP could provide different packaged services to different categories of users.

I guess, that at least some of such scheme is going to be illegal in the US.
Sisyphus 12/5/2012 | 3:34:16 AM
re: Allot Leaps on Nasdaq Debut > .. because all traffic has to go through one
> of these boxes to do all of those things
> successfully ..

While I have mixed feelings about many of the applications that DPI is being talked up to target, thinking that the DPI appliance has to be inline is not accurate.

The DPI engine could just snoop traffic, and send information to a policy server that in turn instructs the inline network elements to enforce certain policies for the indirectly monitored session-flows. In some cases -notably Cisco's- one does not even necessarily need the dedicated policy server in some cases - the DPI can directly communicate with routers that are capable of enforcing local policies.

That said, what is it that we *really* need DPI for? I think it may prove to expensive a tool to truly be implemented at a level that always monitors 100% of incoming traffic in order to actually do something useful for a small fraction thereof. I have no doubt it's a great tool in some cases (compliance with lawful intercept, and the tried and proven application flow optimization), but I wonder whether a more specialized approach wouldn't be more cost effective if P2P traffic analysis is the major driver.

We shall see where and how widely and for what we truly implement DPI at a large scale. I kind of expect that what will end up happening is occasional redirects of samples, and also tap in points for big brother. I don't think it makes sense to analyze every packet that flows into the Internet up to L7, as some advocate. I am open-minded, though!
moneynottalent 12/5/2012 | 3:34:06 AM
re: Allot Leaps on Nasdaq Debut Does anyone have a view on technology advantages/disadvantages between the DPI vendors? Sandvine, Allot, PCube (CSCO)and Ellacoya play in the space I think. I hear not all of them are really scaleable, but I have no way to judge.
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE