x
Optical/IP

Cisco's Telepresence

3:00 PM -- After sitting through a demo of Nortel Networks Ltd. 's telepresence service, I thought it would only be appropriate to see firsthand what its likely primary competition looked like. (See Nortel Trumps Cisco?.) So yesterday, I headed to Cisco Systems Inc. (Nasdaq: CSCO)'s New York office to check out its telepresence offering.

It probably won't come as a surprise that I found Nortel was not as far ahead of the game as it claims to be. I do agree with Nortel that its room delivers a more realistic experience than does Cisco's. Nortel's video screens were built nicely into the room and its camera placement behind the screens was also a nice touch.

But where Cisco scores big points is the near zero audio latency. During my meeting with Cisco's Erica Shroeder in San Jose, I found that Erica would react to whatever I said immediately after I was done speaking, whereas with Nortel's service there was a noticeable delay in response.

Based on that, while Nortel's service is flashier, Cisco's does seem to work a little bit better based on my own limited experiences. But a true verdict can only be determined by those who use the services regularly. Please feel free to share your experiences if you are one of those people. And check out the pictures below of Cisco's telepresence room in New York.



— Raymond McConville, Reporter, Light Reading

Page 1 / 2   >   >>
DCITDave 12/5/2012 | 2:57:42 PM
re: Cisco's Telepresence re: "But a true verdict can only be determined by those who use the services regularly."

So now we have to plunk down $300k each just to figure out for ourselves which system is more impressive?

Which one did YOU like better?

-ph
metroman 12/5/2012 | 2:57:40 PM
re: Cisco's Telepresence Unless the system delivers a similar experience to sitting in front of the person you are meeting with (presence) it is a waste of money. OK - you might think that the Cisco environment was not as nice as the Nortel one, butif I were in meetings with people who took 1 second to respond to what I was saying I would not have meetings with them any more.

I can live with a "nice" meeting room rather than a "special" one, I cannot live with unnatural meetings with delays and talking over one another etc. Anyone who has used an internet VoIP solution will know that you get long latencies in many cases making the call pretty tough. The problems in an HD telepresence environment will be jitter buffers for HD video. How do you minimise Jitter to be able to reduce buffer sizes and limit latency. Add to that the need to syncronise Voice. You have to have a network which can assure expedited delivery end-to-end. HP overcome this by building their own network. Cisco will overcome it by developing a new version of CPN and strongarming the carriers who want to carry this data across a Cisco network. That is the business case for Cisco, pull through of enterprise and carrier lock-in.

Not sure what Nortel are going to do as they have not built a delivery network and they don't have carrier leverage in data. I think that they show their lack of vision by allowing a demo to have so much delay.

One additional point to raise here is support. What if the CEO walks into the Telepresence suite and the system does not work (that's never happened before in Video Conferencing has it...?;-))? There needs to be a very slick way of handling this issue, tracking the fault and getting it up and running quickly. The success of Telepresence will in many ways be based on trust that the system will always work or be easily fixed.

Metroman
metroman 12/5/2012 | 2:57:40 PM
re: Cisco's Telepresence Sailboat

I agree, but I had made an assumption that as Nortel were showing off their system they would install it in an engineered environment that would show it at its best. If they didn't then they were foolish. If they did then I am not impressed with the anecdotal evidence offered.

Metroman

^Eagle^ 12/5/2012 | 2:57:40 PM
re: Cisco's Telepresence OK,

we have some subjective inputs on the quality of the experience on both platforms. Limited sample size (Ray).

While I value the observations so far, there is far too little data or concrete information for us to use to make any kind of informed decision.

What kind and how much bandwidth was provisioned between the Cisco nodes in their demo and between the Nortel nodes in their demo? OCx? DS3?... over ATM? or over lower speed Frame Relay? pure IP over SONET? IP over a wavelength?

OK.. how about what was the router topology behind the video node?

All most all the artifacts of low latency audio, or better video images, etc. can be seriously affected or degraded based on the actual network that the link is provisioned across.

I believe Cisco has very fat pipes between their video locations. And that their router network is configured to give highest priority to the video stream.

I think the ONLY way to do a real comparison is to make all other items equal: carrier connection speeds, carrier backbone links, link routes (packet delay and jitter) made the same, back end router config the same, and router priorities set the same (via MPLS or what ever other control method is needed). Test the end to end link for simple eye diagram and BER to make sure that the link performance is baselined.. and the do a head to head bake off.

Anything else is simple market hype with no data to back it up.

Maybe light reading should do a real bake off with a real lab running it (independent lab NOT connected to or beholden to either provider).

sailboat
prs6str 12/5/2012 | 2:57:39 PM
re: Cisco's Telepresence In addition to Metroman's comments, my assumption was that it would be challenging to find a modern IP network with 1 sec of delay between NY and Toronto, so clearly the issue has to be with the encode/decode/de-jitter buffer portion of the experience.

Yes the network should be held constant for a given test comparing 2 solutions, but there really is no excuse for that much delay for a specially engineered demonstration.
ron202 12/5/2012 | 2:57:37 PM
re: Cisco's Telepresence Sailboat,
Seems that you are not familiar with video technology. A HD video 1920X1080 = 2.1Mpixel X 8 bits per pixel = 16Mbits x 3 = 48Mb x 30frames =1.4 Gbps. Compressed (h264 - depends on movement on the scene but a video conference is not exactly a car racing) is 4 - 8 Mbps x 3 streams (Cisco has 3 screens) = 24Mbps .
Don't you believe that your requirements for 10G or 2.5G are overblown?

Now of course the delay depends a lot on video processing used. If Nortel has a delay of 1 second (or close) probably their codec is not so performant.
ron
^Eagle^ 12/5/2012 | 2:57:37 PM
re: Cisco's Telepresence I still contend it is all about the speed of the network connection.

And how the route links are configured in between.

Yes, it is entirely possible to build a network with that little delay. depends on bandwidth needed. We regularly do it for voice. it is called POTS.

For the kind of experience that Cisco is providing and Nortel is attempting to provide, I contend that buffer's will degrade the experience. buffers by definition hold the data as needed.. and induce delay. The only really good solution is a nice clean high speed link. Hook up a 10g connection dedicated end to end.. pretty straight forward to have a high quality experience. Could you do it with 2.5G link end to end? Probably... hmm... I am sure for the "virtual" projection used in Cisco's "on stage" version, where they use a scrim on a darkened stage to project the image lifesize, so it looks like a hologram (not a hologram.. just simple theater special effects projector technology brought to a high speed link.

Real question still remains, what level of guaranteed connection speed / throughput do you need to have this kind of admittedly great video connection of the telepresence solution, and how much more for the "on stage" virtual hologram set up? ?

It's all about the bandwidth regardless of what ultimate compression algorithms, jitter correction and buffer architecture is used.

I would appreciate some clarity from both vendors so we can make a better apples to apples comparison. What speeds did Cisco use and what speeds did Nortel use for the links? And what connectivity / network architecture?

sailboat
ron202 12/5/2012 | 2:57:36 PM
re: Cisco's Telepresence Sailboat,
Don't want to go in a message board dispute with you.
However a few technical (I hope ) points
1. the market for chips able to encode High Definiton (1920x1080) 30 or 60 frames is VERY limited. A few startups came later but their results are questionable. Probably both Nortel and Cisco had a very limited choice a few years ago when they started the development. I am wondering if they developed their asics for this application.
2. H.264 is quite established but H.264SVC is still in draft (at least last time when I checked)
3. I was told that Cisco is using something like OC3 in terms of bandwidth
4. agree with other reader comment - if Nortel demo had a slight delay problem (it was a demo and was supposed to look good) probably they have a technical problem or they should fire their marketing guy who organized the demo if had wrong setups.
4. regarding the price - well for Cisco 6 screens of high resolution industrial quality displays are probably around 20K , put another 10k for high resolution cameras (6 of them I believe) add the video codec assys , rooms furniture etc and you come in the range of 60-80K.Probably from the 300k they have some discount.
In any event regardless of the company the technology is not (yet) cheap.
It is a good looking video conference experience and probably are some customers who will purchase this type of equipment.
If Nortel had some minor problems during the demo with the delay I am sure that are fixable.
ron

prs6str 12/5/2012 | 2:57:36 PM
re: Cisco's Telepresence sailboat - I don't think anyone disagrees with what you are saying. If a system requires 1G vs. 20Mb/s then that is a consideration. Flexibility of that network design/service is also.

The point was that people sometimes get fired when they knowingly put their product into a public demo with anything other than optimal conditions. One would assume that Nortel provided a non-blocking low-latency path between NY and Toronto for this demo. It's not like they setup the room that day for the demo - it's a permanent setup. So, the issue was probably in a component they could not "fix." The next logical latency culprit is the encoder, IMO. Again, assumption is they wouldn't put it on an unsuitable network.

I'm not sure your comments on the availability of components and industry knowledge for 1080p video conferencing systems is valid. Why would these chips be in abundance if there are only a handful of vendors just now introducing these products? Encoding/decoding chips, yes...but I bet these are for systems that are one-way delivery where a second is irrelevant...not realtime communications that Telepresence requires.

I remember in the early days of VoIP, encode/decode/de-jitter buffer engineering was actually a distinguishing characteristic of VoIP products. HD teleconferencing with spatial audio is entirely more complex and would not surprise me if the maturity of chipsets makes hitting those end-to-end latency targets challenging without significant tweaking by the codec provider. I'm sure this will change quickly, but as I mentioned before, these products are only now emerging.
^Eagle^ 12/5/2012 | 2:57:36 PM
re: Cisco's Telepresence Ron,

to be more clear.

if Cisco needs to run theirs at 1gig (yes, this is often used as link method on various digital video feeds and approximately what your calculations showed), and Nortel needs 2gig or vice versa, that will tell a lot.

As the video encoding methods and video processing methods are well known and well understood and are largely based on industry standards (h.264 for instance), and there are limited suppliers of the chip sets to do this processing, I contend that rolling up the needed parts to make a video conferencing system can be done by many developers and integrators. I also contend the delay that the network can introduce as the link / video transits across the WAN can easily be a much more significant contributor to jitter, delay, quality of experience than the local processor and ram architecture. (assuming the designer was at least competent and didn't do a bad design.. this kind of stuff has been understood by video companies for almost 20 years). If a link runs across the continent or even across continents (like the link cisco shows on their web site showing the "on stage" experience between San Jose and India), I strongly contend that the link speed and CONFIGURATION will be the key ingredients to a great video experience. Configuration includes route engineering and architecture, switching and routing topology, priority paths through the local link (across the enterprise LAN) and over the firewall into the telco WAN, etc.

I cannot asses the relative performance of the two systems and their relative cost or cost to value proposition unless I know what is required to really deliver the experience.

How much bandwidth? and how must it be configured and set up? Should it be a dedicated link? or could I use a shared or switched infrastructure? could I use Frame Relay? could I use ATM? must I upgrade my entire corporate infrastructure and link design and speed to a new IP over optics network with all new Cisco MPLS routers to make this Telepresence work?

Doesn't matter what the answer is.. it could be over a T1 or even a fractional T1.. it could require a 2.5G dedicated link. I don't care. My point is that we cannot make a clear comparison without a full picture of the entire network architecture.

cisco and nortel both have a fairly large high bandwidth internal network linking all their sites together. If I am going to consider either solution and if both are making claims about performance and superiority, I need to know a lot more to make such a comparative analysis.

How about a bake off over the same infrastructure and routing backbone?

Other wise it is only marketing bs.

sailboat
Page 1 / 2   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE