A recent study has found the quality of VOIP calls declining steadily over the past two years, and points to network congestion and lack of service prioritization as the culprits. (See VOIP Quality Dropping.)
Brix Networks Inc. has come up with a clever way to study the quality of VOIP calls -- a portal called TestYourVOIP.com where consumers themselves place dummy calls. Brix says the site has generated nearly a million test calls since it launched in March 2004. (See Report Hisses at VOIP Quality.)
Brix says almost 20 percent of those calls rated “unacceptable” against a long-used quality testing algorithm called the “E Model.”
“It’s pointing to congestion and competition with other media,” Brix CTO Kaynam Hedayat says of the study’s implications. “There’s no differentiation in the network between VOIP and other kinds of traffic like video or gaming -- it’s just pure democracy.”
Test calls with a MOS of 3.6 or better, Brix says, are typically regarded as having satisfactory quality. Only 81 percent of the test calls achieved the 3.6 threshold.
Some in the VOIP business feel such studies show that VOIP providers are victims of their own success. “You’re dealing with consumer expectations as much as you are changes in call quality,” says SunRocket Inc. spokesman Brian Lustig. “Consumer expectations of land-line service are very high, and as more consumers begin to feel comfortable with VOIP, expectations rise.” (See VOIP Testing Goes Live.)
Lustig disagrees that overall VOIP quality is on the decline. “There’s been an explosion of VOIP use in the last couple years, and part of the reason for that is VOIP quality has improved,” Lustig says. Indeed, the global market for consumer VOIP has grown to nearly 20 million subscribers, researchers say.
Lustig says call quality improvement is one of the main things driving VOIP providers toward forming peering arrangements with one another. (See VoicePulse Peers With XConnect.) VOIP providers can avoid the “uncertainty” of the PSTN by forming direct IP handshakes with other VOIP providers, Lustig says.
Global IP Sound AB CEO Gary Hermansen points out that when a voice call leaves the IP domain for the PSTN, it must be transcoded into a lower-quality sound stream. He says quality is further challenged by the fact that VOIP services are now used by millions of people, and by the fact that video packets have been added to the service in many cases. (See Skype Uses Logitech, On2.)
TestYourVOIP.com displays a real-time “weather map” of the Internet's current ability to support real-time services, such as VOIP and IPTV. Registration is required to use the site.
Brix’s interest in the subject stems from its own product line -- real-time VOIP and IPTV testing and monitoring gear. Brix is one of a growing number of companies marketing the technology; others include:
Within the U.S. wireline operators I agree that within the short term DPI's most likely broad use is for P2P. However, look around the world and you already see traffic being blocked for other reasons. I also think that DPI is a technology, not a philisophy so it can be used in different ways. For example, if a consumer gives an operator permission to use DPI to implement a specific policy for the benefit of the consumer, that is different, even if not neutral, than using DPI without the consumer's permission or knowledge.
As for the effectiveness of security a few comments. I disagree with the idea that the use of security is a quasi-admission of doing something in avoidance of the operator's business model. Some people genuinely believe in end-to-end security without any consideration to the implications of network-based policy. Also it is worth noting that application signature recognition may be possible without DPI, so application developers may end up having to implement other strategies (changing the traffic profile for example) to deal with all kinds of network-based recognition techniques. The bottom line is any technology can be used in any number of ways. Don't focus on the technology, focus on the administrative policy of a network.
With respect to using DPI for methods other than P2P shaping/blocking, I spoke to a senior executive of a carrier earlier this year and he said, when asked if they would use DPI for blocking "we have not decided if it is worth our while yet". This both supports Seven's assertion that it may not be practical while also supporting the assertion that carriers may be willing to use this technology for blocking if they find it is worth their while. The frustration carriers feel over technology and operator non-neutral regulation should not be underestimated and certainly leads them to competitive feelings that *IF* acted on could lead to a variety of things that might be viewed as anti-competitive. Part of the solution then is relieving this tension by getting to neutral regulation (the path the FCC is currently on).
I agree with Fred that the relaxation of common carrier regulation has implications. But as I said before I think the old bucket of laws considered "common carrier" were overloaded with too many requirements that dated back to a monopoly period and need to be revamped so a *much* smaller set of common sense requirements can be applied to all that transport traffic for the public (not just dominant carriers), and perhaps even those that process traffic as well. Perhaps relaxing carrier obligations is just a first phase in rationalizing them to fit for a more competitive envirnoment. However, the Internet is the only major transport infrastructure with no carrier obligations. That may be an important regulatory break through, but I have as yet to hear that argument, and until I do it is an observation that can not be trivially dismissed.
Seven, DPI has been sold to mobile networks for the simple reason that they were the ones exempt from common carrier requirements. If you wanted to run fast data across VZW, you used their data service, not your choice of ISP. They were not subject to Computer II. Now that Computer II is history, wireline networks are free to become like the wireless ones. In other words, wireless was a model for a deregulated environment, wireline a model for a regulated one with enforced neutrality in the bottom layers (below IP). Cingular is owned by fATT, VZW by VZ, so you figure out what they had in mind when asking for Computer II to be lifted.
DPI networks can charge for content quite easily, since they do read into the actual application payloads. Bytemobile (the former Proquent part, specifically) is quite proud of its capabilities there. Yes, PGP can hide the content, but PGP only works when both sides use it, and public application-layer services (things for sale) can't encrypt any more than the browser handles internally. And even that's defeatable by a man-in-the-middle attack by a network operator. A DPI network can also charge a stiff premium for all encrypted bits; after all, if youre encrypting, you're obviously trying to get out of paying them what they think you owe.
Telco entitlement mentality, after all -- think of the "phantom call menace" (the current Missoula Plan treats it as public enemy number one) and all of the other arguments telcos have against access charge evasion.
Not disputing your point per se, but your example is for mobile networks - which are primarily walled gardens in any case. The discussion here is about wireline networks. I recognize you were using a technology example.
I think the reality in the short term is that DPI is used for capping P2P traffic. No content charging is done, and in fact it would be very hard to do. Anybody with any savvy would simply PGP the content. So at best, they will be able to work at the protocol levels. Protocol detection and charging is more like the FedEx example (assuming that QoS rules are actually obeyed by these networks - which of course today they are not).
FedEx most certainly does not inspect the inside of its packages. It's a common carrier, regulated as such. It can charge for weight and volume, both of which have a cost impact. I suppose it can charge for insurance, based on the sender's declaration of value. But it most certainly doesn't "deep inspect" its contents. It doesn't care if you're sending a pair of old sneakers or a gold ingot.
Bytemobile, on the other hand, markets DPI gear based on the relative "value" of different bits, with several orders of magnitude of difference between the highest-cost bits (SMS) and the lowest-cost bits (TV streams). Their core value proposition is to allow carriers to charge for anything that competes with SMS, including SMTP email, at a high enough price to protect SMS revenue, and at the same time to allow low prices -- in their model, below actual cost -- for video. This was shown by Rod Randall (whose VC firm has Bytemobile in portfolio) in the "Don't Get Skyped" lecture/video. Cisco competes with them.
What you're describing is not DPI but Measured Service. While consumers don't like it, it's a legitimate form of common carriage. Charging for number of bits, QoS, distance, and other things that can be gleaned from headers is not DPI, and is not a violation of core neutrality principles (though I agree that some extremists, myself not among them, still object to it).
The core issue is DPI, pricing based upon the private-carrier's perception of the "value" of the content as perceived from DPI (reading beyond the application layer headers), and selective blocking or gratuitous degradation (e.g., of VoIP providers). DPI vendors advertise those capabilities.
Packet inspection in the FedEx world is done by actual measurement (weight, dimension, distance) and by proxy. Delivery SLA is actual. Packets are thus "rated" and carriage fees are commensurate with these values.
It goes to create a measure of the load put on the resources of the delivery network. Similarly, diesel fuel taxes, transport licenses and other fees levied against the common transport carriers are meant to rate fees against the load they put on the highway system. An analogy that has been used on this thread many times as a way to argue for utility-based network services (such as those soon to be provided by DPN, Inc.).
(We'll leave aside the dirty little secret that many DPI technologies do not actually have any "D" in their DPI).
The same holds in an access network. The analogy - deliberately simplistic and meant to inject some humor (it would seem that you have none!) still holds quite well.
Rate away - it doesn't trouble me in the least. 38 others you say? I had no idea my post was so well-loved.
slideruler: "My goodness – this IS serious. Where will it all end? Next thing you know, FedEx will want to charge content shippers…err…I mean customers…differently for how much space each packet…ooops…I mean each package takes up in their cargo holds - or what the delivery time guarantee is. Such absurdity should be stamped out!"
This is not analogous to DPI. The correct analogy for FedEx would be opening up every package and deciding how much to charge based on the contents of the package
sr: "Nice touch on the post rating by the way….well done. ;-)"
38 people collectively gave your previous post a 1, I was one of them. I suspect it had to do with poor logic like your false analogy above.
“...actively, openly selling DPI-based billing systems...openly talk of new revenue models”.
My goodness – this IS serious. Where will it all end? Next thing you know, FedEx will want to charge content shippers…err…I mean customers…differently for how much space each packet…ooops…I mean each package takes up in their cargo holds - or what the delivery time guarantee is. Such absurdity should be stamped out!
Personally, I believe in the future of completely commoditized, dumb network transport. Preferably, with as much regulatory control as possible. I have a great pitch for a new breed of SP startup, here goes -
“With the planned Series A investment of US$100 Billion, Dumb PipeDream Networks seeks to take a dominant position in commodity bit-pipe network transport services - seeking to control the zero-to-negative margin segment of basic bit transport. It will aim to provide a world class next generation NN-compliant (RFC PWWG-NN-0110) transport network, becoming the network of choice for third party application-service providers utilizing DPN’s assets in order to create additional value for their shareholders. By eliminating the need for application-service providers to pay value-based rates, we can cede the pursuit of incremental service & content revenues to third parties, enabling us to focus on our core competency of subsidizing the business operations of others.”
What do you think - want to sign up as the lead on this one? I attempted to use as many consulting buzz-words as possible, but am sure I fell miserably short.
As I have stated in the past – I am not affiliated with equipment manufacturers, industry groups, Telcos (incumbent or other), associations, etc, etc, so please do not jump to any data-free conclusions.
Nice touch on the post rating by the way….well done. ;-)
Mr. Zippy, I'm happy to hear that FECN is finally catching on in the IP world. I put it into FR (took about three years) but then the IP folks didn't want it, and ignored it, with all sorts of silly excuses. We did some serious simulation on it at DEC before putting it into DECnet Phase V, OSI, and FR, and it made a lot of sense.
However, I can see why it's a mismatch for streaming -- the whole TCP/IP thing is really a mismatch for streaming. ECN gives more feedback to TCP than mere packet loss (the default VJ mechanism). TCP naturally adapts to the available capacity; its core apps (data transfer) don't have an inherent speed. But VoIP has an inherent speed. I could imagine that an end-to-end Skype-like app could adjust its parameters in real time, but most VoIP runs through media gateways, and is fixed rate.
So we hope that streaming doesn't overwhelm the network, and optimize for it, and hope everyone's happy. I dunno... it doesn't seem to me to be the most stable approach. Just because something works most of the time doesn't make it well designed. Heck, even Windows 3.0 worked more than 50% of the time. Well, barely...
Slideruler, my consulting work has little to do with DPI, NN, or any of that -- I'm a policy wonk at heart, and express my views even if it is not what my clientele thinks about. My "conspiracy" theories come directly from the equipment vendors in question, who are actively, openly selling DPI-based billing systems to carriers. It comes from carriers who openly talk about new revenue models, like taking a cut from ecommerce, and blocking sites who don't pay double. I don't need to make it up.
Let's face it -- your Bell friends have gotten people upset, and now they're trying to change the spin on their plans, at least until Congress gives them a pony.
Please enlighten me why this RFC4340 mechanism will work a lot better than previous attempt like the FECN and BECN mechanism of FR that works a lot more faster.
The IP version of BECN and FECN, namely
RFC3168, "The Addition of Explicit Congestion Notification (ECN) to IP"
It's becoming more widely deployed, although I'm not sure about Windows boxes (I'm guessing it is enabled on those though, there are some smart networking people at MS who would be making recommendations on the right things to do by default), it has been in the Linux kernel and on by default for something like at least the last 5 years. If SPs are suffering from congestion, one of the best things they could do is enable ECN on their routers, so that sending hosts would backoff their send rate before rather than when packet loss occurs. Of course, and unfortunately, it is up to UDP based applications to perform congestion control, and I think in most cases they don't, which is why DCCP is the better protocol for these types of application. One alternative possibility might be for the IP layer in a host to be ECN aware, and drop UDP packets that would cause congestion before they're sent (it's quite legitimate to drop a UDP datagram before it's hit the wire - UDP isn't reliable, so the unreliability could occur anywhere between the IP endpoints, including inside the host that is being asked to send the UDP datagram in the first place).
The blogs and comments are the opinions only of the writers and do not reflect the views of Light Reading. They are no substitute for your own research and should not be relied upon for trading or any other purpose.
Mobile Broadband Summit India Tuesday June 5, 2012 Thursday, June 7, 2012
Shangri La Hotel in New Delhi & ITC Grand Central in Mumbai, Delhi & Mumbai, India
To save this item to your list of favorite Light Reading content so you can find it later in your Profile page, click the "Save It" button next to the item.
