VoIP - one size doesn't fit all
http://www.ietf.org/internet-drafts/draft-stein-great-00.txt
OK - I have read this and I can't let this one pass without some comments - I just can't do it.
The author argues that packetized voice over the internet is of suspect quality, especially when compared to voice over a circuit switched network. That is analogous to stating that car 'A' drives better than car 'B', when at the time of comparison, car 'A' was driven over a crater-laden mine field, and car 'B' on a newly paved highway. Then he makes the following claim,
"Why would anyone use a VoIP systems if the user experience is significantly inferior to that of the standard telephone system?"
This is akin to asking why anyone would use car 'A'. It's not the car, but the road. It's not voice over IP that is in question here, but voice over IP over the public Internet. He does go on to acknowledge that the design of the Internet is the crux of the issues.
"Of course, much of what was said above is specific to the present state of the public Internet, while well engineered, highly overprovisioned, networks suffer much less from these troubles."
Some more comparing apples to oranges.
"The above discussion focused on VoIP, but similar statements could be made concerning other forms of real-time traffic transported over the Internet, such as videoconferencing. On the other hand not all real- time traffic is as problematic. For example, streaming audio that can be delivered after a certain delay may be able to exploit retransmission mechanisms, and thus be immunized to many of the above hindrances. The essential ingredients are real-time constraints and delay insensitivity, characteristics present in interactive real-time applications. "
"On the other hand not all real-time traffic is problematic…" Comparing streaming UNIDIRECTIONAL audio to a live BIDIRECTIONAL conversation is moronic at best. I just can't stop there, there's more orthogonal argument making in the very next section.
He goes on to argue, "The design philosophy of the Public Switched Telephone Network (PSTN) presumes that routing is expensive but bandwidth plentiful, while that of Packet Switched Networks (PSNs), such as the Internet, presupposes bandwidth to be dear while routing affordable." I challenge anybody to find me a circuit-switched based engineer to state that bandwidth was/is plentiful in the PSTN. The key to good voice quality is the _dedicated_ nature of the transport medium, not the abundance of it. The PSTN dedicates bandwidth for the duration of the call. It's a point-to-point connection-oriented communications network. IP is inherently connection-less. This is the key difference. The remainder of section 2 goes on to describe the variables that affect latency, and how prominent those variables are in the Internet. Some of the points that he makes, variables in one-way network paths are addressed by ensuring symmetry, so that the bi-directional flows would have increased odds of similar network characteristics. This helps ensure that the sum of the one-way network characteristics are similar, if not the same, as the observed two-way results: 2+2=4, instead of 3+1=4.
Another point made that I am struggling with…"Even assuming a perfect PSN, i.e. one with no packet loss (PL) nor packet mis-ordering and only minimal packet delay variation (PDV), the perceived voice quality of VoIP calls is highly dependent on bandwidth reduction mechanisms. "
There were no quality issues that I recall as a result of our enabling VAD or supporting compression. It all in the quality of the implementation. I have sitnessed that enabling VAD, reduced bandwidth per call to approx 55k, which is less than a 64k DS0 for those that are arithmetic-challenged. I will acknowledge that the amount of savings from VAD is directly proportional to the volume of calls. VAD on a DS1 will not have the same impact on bandwidth per call as VAD on a DS3.
I
n section 4, the author points out issues with Delay and Delay variation. One of the suggested solutions is as follows: "…..a sensible approach would be to start with a specification of the network delay, and to derive allowable buffering and processing budgets. This would probably require smaller frame sizes and minimization of lookahead, and innovative designs would be needed to keep bit rates reasonable. "
This point was made without acknowledging the potential ramifications of that tactic. Reduce the frame size would greatly skew the ratio of overhead to payload towards overhead, thus reducing bandwidth efficiency. Not a desirable result.
To conclude the rant I am on, I believe that the author has simply missed the point of VoIP altogether. The author states, "... some enthusiasts have suggested that the attraction of VoIP is due to the additional functionality that is, or will be, available (e.g. instant messaging, video). However, in most cases it is probably either the economics (free calls) or the ready accessibility for people already seated at a PC (along with presence indications) that induces most people to tolerate the poor quality." but doesn't acknowledge a logical evolutionary path to mass voip adoption. I argue that VoIP is on 5 completely separate evolutionary paths, and therefore VoIP can no longer be thought about in general term. There will be more to come on that point in the future.....
OK - enough - time to get back to work.
-Adam "voiploser" Uzelac
http://www.ietf.org/internet-drafts/draft-stein-great-00.txt
OK - I have read this and I can't let this one pass without some comments - I just can't do it.
The author argues that packetized voice over the internet is of suspect quality, especially when compared to voice over a circuit switched network. That is analogous to stating that car 'A' drives better than car 'B', when at the time of comparison, car 'A' was driven over a crater-laden mine field, and car 'B' on a newly paved highway. Then he makes the following claim,
"Why would anyone use a VoIP systems if the user experience is significantly inferior to that of the standard telephone system?"
This is akin to asking why anyone would use car 'A'. It's not the car, but the road. It's not voice over IP that is in question here, but voice over IP over the public Internet. He does go on to acknowledge that the design of the Internet is the crux of the issues.
"Of course, much of what was said above is specific to the present state of the public Internet, while well engineered, highly overprovisioned, networks suffer much less from these troubles."
Some more comparing apples to oranges.
"The above discussion focused on VoIP, but similar statements could be made concerning other forms of real-time traffic transported over the Internet, such as videoconferencing. On the other hand not all real- time traffic is as problematic. For example, streaming audio that can be delivered after a certain delay may be able to exploit retransmission mechanisms, and thus be immunized to many of the above hindrances. The essential ingredients are real-time constraints and delay insensitivity, characteristics present in interactive real-time applications. "
"On the other hand not all real-time traffic is problematic…" Comparing streaming UNIDIRECTIONAL audio to a live BIDIRECTIONAL conversation is moronic at best. I just can't stop there, there's more orthogonal argument making in the very next section.
He goes on to argue, "The design philosophy of the Public Switched Telephone Network (PSTN) presumes that routing is expensive but bandwidth plentiful, while that of Packet Switched Networks (PSNs), such as the Internet, presupposes bandwidth to be dear while routing affordable." I challenge anybody to find me a circuit-switched based engineer to state that bandwidth was/is plentiful in the PSTN. The key to good voice quality is the _dedicated_ nature of the transport medium, not the abundance of it. The PSTN dedicates bandwidth for the duration of the call. It's a point-to-point connection-oriented communications network. IP is inherently connection-less. This is the key difference. The remainder of section 2 goes on to describe the variables that affect latency, and how prominent those variables are in the Internet. Some of the points that he makes, variables in one-way network paths are addressed by ensuring symmetry, so that the bi-directional flows would have increased odds of similar network characteristics. This helps ensure that the sum of the one-way network characteristics are similar, if not the same, as the observed two-way results: 2+2=4, instead of 3+1=4.
Another point made that I am struggling with…"Even assuming a perfect PSN, i.e. one with no packet loss (PL) nor packet mis-ordering and only minimal packet delay variation (PDV), the perceived voice quality of VoIP calls is highly dependent on bandwidth reduction mechanisms. "
There were no quality issues that I recall as a result of our enabling VAD or supporting compression. It all in the quality of the implementation. I have sitnessed that enabling VAD, reduced bandwidth per call to approx 55k, which is less than a 64k DS0 for those that are arithmetic-challenged. I will acknowledge that the amount of savings from VAD is directly proportional to the volume of calls. VAD on a DS1 will not have the same impact on bandwidth per call as VAD on a DS3.
I
n section 4, the author points out issues with Delay and Delay variation. One of the suggested solutions is as follows: "…..a sensible approach would be to start with a specification of the network delay, and to derive allowable buffering and processing budgets. This would probably require smaller frame sizes and minimization of lookahead, and innovative designs would be needed to keep bit rates reasonable. "
This point was made without acknowledging the potential ramifications of that tactic. Reduce the frame size would greatly skew the ratio of overhead to payload towards overhead, thus reducing bandwidth efficiency. Not a desirable result.
To conclude the rant I am on, I believe that the author has simply missed the point of VoIP altogether. The author states, "... some enthusiasts have suggested that the attraction of VoIP is due to the additional functionality that is, or will be, available (e.g. instant messaging, video). However, in most cases it is probably either the economics (free calls) or the ready accessibility for people already seated at a PC (along with presence indications) that induces most people to tolerate the poor quality." but doesn't acknowledge a logical evolutionary path to mass voip adoption. I argue that VoIP is on 5 completely separate evolutionary paths, and therefore VoIP can no longer be thought about in general term. There will be more to come on that point in the future.....
OK - enough - time to get back to work.
-Adam "voiploser" Uzelac
posted by voiploser at 11:54 AM
0 Comments:
Post a Comment
<< Home