Richard Stastny identifies (and pointed out by Andy) a paper from Columbia University that tries to empirically observe the workings of Skype. Many of the findings confirm earlier claims. Some findings are new and have implications to us who are interested in developing an open system. The following is a collection of excerpts from the paper followed my editorial comments. The authors indicate that many details are hidden. I feel that it is not critical that we have to replicate each component of Skype; we have to focus our attention in addressing the problem on hand: how do we facilitate consumers to easily setup a mechanism so that they communicate with others, with out the aid of service providers.
“Any node with a public IP address having sufficient CPU, memory, and network bandwidth is a candidate to become a super node.”
The name to fame for Skype is that it eases the NAT/Firewall traversal problem. This does it with the help of Supernodes. How many of them will be available? As I have said before, either Supernode is the Achilles heel of the architecture or there is no need for them in the first place, meaning there will be no need for Skype.
“Skype uses wideband codecs which allows it to maintain reasonable call quality at an available bandwidth of 32 kb/s… We measured that the Skype codecs allow frequencies between 50-8,000 Hz to pass through.”
Now we know why Skype is not encouraging an ATA solution – traditional phones have only 4 kHz range. So those VoIP systems that are based on ATAs better rethink their requirements.
”The buddy list is local to one machine and is not stored on a central server.”
Probably the API will allow us to store the list in such a way that it can be loaded onto the next computer at login time. Let us not concern about finding such a storage space. What is gmail for anyway.
”SC cannot prevent itself from becoming a super node.”
This must be a new development. Discussion forum suggested the opposite. Given the processing and bandwidth load, I suspect that the wider public will find an artificial way to disqualify themselves, like inserting a NAT/Firewall.
“It thus appears from the reverse lookup that the login server is hosted by an ISP based in Denmark.”
The CEO coyly stated in an interview that the location is hidden in a secret spot. Assuming that this is indeed true, then the Denmark server must be a misleading proxy with the real login server hidden behind this machine.
“The caller SC established a TCP connection with the callee SC. Signaling information was exchanged over TCP.”
Harbinger of doom for VoIP service providers: If the end-points can locate themselves, there is no need for Skype or any other service provider. I would think real believers of intelligent at the end will be working to address this issue, rather than spending energy in dumping on PSTN or extolling the virtues of Skype.
“The caller sent signaling information over TCP to an online Skype node which forwarded it to callee over TCP. This online node also routed voice packets from caller to callee over UDP and vice versa… The negative side is that there will be a lot of traffic flowing across this node.”
Many other traversal techniques use optimization techniques to ease the demands on “supernodes” It is disappointing that Skype does not use such techniques.
“Thus, the total uplink and downlink bandwidth used for voice traffic is 5 kilobytes/s.”
I think this is just the bandwidth consumed by the codec. One has to add the transport overhead to compute the consumed bandwidth.
"We think that three factors are responsible for its increasing popularity. First, it provides better voice quality than MSN and Yahoo IM clients; second, it can work almost seamlessly behind NATs and firewalls; and third, it is extremely easy to install and use."
It is worth pointing out that it is not difficult to replicate these three characteristics using open technologies.
Posted by aswath at December 7, 2004 11:54 PM
Interesting comments, especially about the capability to replicate the simplicity of Skype, but one point is to speak about it the other is to deliver. I have installed one SPA-2000, 2 x SPA3000 and about 10 softphones, and I can say that it is significantly more complex than using Skype.
Skype also leave SIP implementations for dead when it come to connect to the servers behind changing network architectures which is what happens when one commutes (and this is one of the key points put forward for VOIP).
I believe SIP has some good things, and I use it extensively, but the applications need a major effort to improve their ease of use and functionality. We should applaud what Skype is doing, duplicating the good part and try to interconnect.
For what it's worth.
John
Posted by: John at December 13, 2004 08:46 AM