re: DPI Market Set for Mobile RampWhat would DPI be used for? Traffic shaping and prioritizing by service class (=cost)? Is there a major difference between wireless users and wireline users as far as DPI usefulness is concerned?
re: DPI Market Set for Mobile RampDPI for mobile is yet again a failed strategy for controlling class of service.-á Internet and video coming to mobile devices is chewing up bandwdith.-á Instead of designing a network to deliver services from any IP service platform to any IP client devices across a diverse set of technologies some are lobbying and selling for a flawed approach --- DPI.-á There is no way DPI can ever scale cost effectively or be used in a manner which enhances the consumer experience.-á DPI and IMS are the worst ideas since TMN and CMIP.
re: DPI Market Set for Mobile RampSo the basic question for service providers is either to keep investing in more bandwidth or into DPI to better manage the limited bandwidth they've got. Certainly investing in more bandwidth keeps the network simple, on the other hand DPI adds a lot of complexity (and cost) to the network. As an enduser I would always prefer a service provider who just provides me the bandwidth I need without looking into my packets. I believe the market will decide what approach will be taken.....
That sounds like a purist's comment. It would be expensive, and probably impractical, to build the kind of mobile network you require.
Mobile radio bandwidth is a shared resource, of 2 Mbit/s to 3 Mbit/s per sector on the downlink (today's best case).
To offer services at affordable price points, operators need to control traffic. DPI is one tool that might help. TCP optimization and content adaptation are other tools already widely used to "optimize" services and network performance.
re: DPI Market Set for Mobile RampI repeat myself: DPI is useless against encryption.
Any serious investment into DPI will drive applications simply to encrypt the traffic. The computation time is minuscule compared to the transmission time.
DPI simply delays dealing with the underlying problem: Statistical models of traffic that do not account for the tails correctly are going to run into trouble as high-bandwidth applications emerge.
re: DPI Market Set for Mobile RampWe may be talking past each other in regards to terms. DPI is used to police usage by specific users for services carriers want to restrict or eliminate. Class of Service (QoS) is what carriers offer to consumers who subscribe to services. The new mobile services need to upgrade the ability to deliver the right QoS for the video and other moible services they want to offer. This has nothing to do with DPI. As mobile subscribers roam they will indeed run into the issue that specific areas they are in will not be able to offer the QoS for the services to which they have subscribed and are trying to access. DPI solves none of these issues. The largest issue is in ensuring end to end timing and synchronization of higher quality video and audio services over mobile networks which means controlling latency and jitter from delivery platform, over the optical network, out to the base stations and then over the air. Again, DPI does nothing for this problem. Everything described above requires upgrades to all aspects of the mobile delivery network(s).
Is your scenario to the consumer: please subscribe to our new mobile services and if you use them to much we are going to restrict your access and usage. That is all DPI can do. If a subscriber has signed up for a level of access and quality then that should be delivered as agreed. In mobile networks where every device and device owner is identified by services subscribed upon network access I fail to see how something as complex as DPI or IMS are useful.
Probably said more than this topic is worth so I will leave it at this.
re: DPI Market Set for Mobile RampMaybe DPI is too specific a termGǪ but without question mobile operators will limit the capacity allocated to certain types of users.
Oddly, most of the deployments (I understand) of DPI in mobile networks today are between the GGSN/PDSN and the Internet and are intended to provide fairness to that part of the link. In that case you'd be right, surely it's easier to for operators to get more Internet bandwidth than deploy DPI.
It's in the RAN where bandwidth it's harder to come by. Regular Internet apps are already high-bandwidth from a mobile perspective. A 256 kbit/s video stream might seem trivial on a wired network, but it's hard to support more than four simultaneous users at that speed over the air in a 3G sector/carrier in a typical network configuration.
re: DPI Market Set for Mobile RampUnfortunately I am not a purist. I believe the cost of implementing a functional DPI feature that will work across networks and service provides is much higher than just building a functional network. Stevry has it right about encryption. I just do not see how DPI could ever be implemented effectively. DPI and IMS are both overly complex and driven by real purists who love to spend their lives writing complex standards and attending meetings.
re: DPI Market Set for Mobile RampStevery, I don't think you know what DPI is, encryption is not the problem. Please look up how DPI is done and also why people are complaining about Bittorrent is being throttle. Bittorrent is encrypted traffic. Also learn how VoIP is being throttle and remember VoIP is encrypted.
Is there a major difference between wireless users and wireline users as far as DPI usefulness is concerned?