The VON show was entertaining as always. Probably a little over a thousand people attended, but those who did were enthusiastic about it. Clearly not as big as 2 years ago, but attended by a lot of interesting folks. The mood was certainly a bit somber because of the market condition, with again the folks tied to enterprise and asian markets a bit brighter. Was it an accident that this year’s VON logo souvenir was a hooded black sweatshirt that could on a dark night double as a grim reaper costume? Some common thoughts:
The Exhibit floor was large enough to be interesting even though there were a lot of familiar companies not there. Microsoft, Cisco, and Intel had large prominent booths. Microsoft was displaying some partner applications for call center and other enterprise oriented telephony applications. They were also showing messenger, with video and handwritten messages on a tablet PC. The tablet PC looks like a small laptop with a screen that swivels 180 degrees and folds down so that the screen is exposed on top. You draw on it with a special pen that it tracks. Messenger will send drawings or do handwriting recognition. Recognition rate looked pretty good even with my sloppy writing (occasional letters wrong but not enough to make it unreadable). Having spent a lot of the conference alternating between writing and typing notes, I know for me typing is MUCH faster, so I asked about that. The presenter acknowledged that particularly younger folks who grow up typing may not find this an advantage, though it occurred to me that it may be a major seller in parts of the world with more complex written languages that are less easily typed (i.e. china, japan, and the middle east).
Intel showed a software media server. 200 ports of conferencing, speech recognition, and other media intensive services on an 800Mhz Pentium. It was described as “Dialogics functions without the boards.) Interesting, though I’m not sure it’s really dense enough yet.
Netrake showed me something interesting – an IP session controller, built on network processors, which basically maps packet contents at the upper levels to allow SIP and other protocols to pass through NATs and firewalls. In effect it provides firewall and NAT traversal without having to reconfigure the firewall or NAT to be SIP aware.
Several application server and IP centrex vendors were there with small booths (pactolis, net.com, Nuera, Sylantro, SS8) Lucent had a largish booth featuring Softswitch and their converged server. Siemen’s had a good sized booth displaying enterprise solutions as did Avaya. Nortel, Alcatel, and Sonus were among regular VON vendors not there.
As the conference opened, instead of the usual preview of the featured bands at the party the audio played a medley of older songs about survival and fighting the good fight. Jeff opened by saying that he was looking for a theme song for the event. The last 6 months have been about survival. We need to keep going and know that there is fertile ground out there. Not all the ground is fertile. Jeff repeated his call for services that deliver "Purple Minutes'" -- high value services people will pay for that are different. He reminded us that there is no doubt about the fact that the future of communications is in IP. We forget that sometimes and it takes time to get the rest of the world to believe in it.
He said that there were more
people here from
He talked about lots of quiet accomplishments –
What gets in the news about the industry isn't how green OUR industry is but how bad it is for everyone else (i.e. those still selling circuit telephony. He put in a plug for the Pulver100, his list of 100 companies leading the VON effort. The Pulver100 are all private companies, not publicly traded. All have good wins. (Pulver.com/pulver100. there are many familiar names on the list)
What does the industry need -- Customers! How do we get Customers? Pursue ways that VoIP can enhance people’s lives.
Trends: 802.11 will seriously challenge 3G -- it's a solution that empowers people. People will be putting VoIP over 802.11 (Comment -- The professor part of me of course having just written an exam covering the Shannon and Nyquist limits on communication wonders how all those bits are going to fit in a couple of fairly narrow public spectrum bands).
Jeff's Marketing lessons:
Internet Change agents:
This is where "Purple minutes" come from -- change agents. e.g. adding voice to anything.. The personal central office -- you define your own services (e.g. hosted call centers!)
Purple minutes come from connectivity, community, applications. Some thoughts for people:
Jeff's wish list:
· Royalty free variable bit rate adaptive codec
· Low cost low end IP edge devices.
· Distributed service execution -- platforms exist, deploy them.
· Take advantage of the great QoS on the internet backbone.
· IPv6 needs to happen. Congress won't mandate it but it has to come because it solves problems.
· The battle cry for Purple minutes has yet to be heard. (The investors and customers don't get it yet)
· Drop "IP" from IP Communications -- we are the future of communications.
We are "On the upside of the downturn". He described the history.
David doesn't want to give his wireless network away because of security issues. Because the industry hasn't addressed security, we break protocols with firewalls.
· Integrated experiences using IM and presence
· Standard infrastructure based in SIMPLE and SIP
· People to people communication, not device based.
On standards, remember proprietary email (Comment -- is he talking about exchange and outlook?)
We solved the NAT problem, but firewalls are still a pain. (Comment -- did we really solve the NAT problem? Is this why when I hook to Pulver's wireless network I'm constantly annoyed by windows alerting me to home gateway (NAT) devices on my network?)
We have multi-modal devices, we need to interwork them with applications (Comment -- HP had a poster of a road warrior with a PDA, laptop, cellphone, pager, set up in the hall. At least a couple of folks I was talking with who saw it said this is exactly what the communication industry shouldn't be doing).
David showed a Cisco/Microsoft video on their integrated service. Basically windows messenger working with Cisco IP phones on IM. Also I think there was some SMS in this. Not all that exciting. It's all stuff we can do today.
David's predictions for the future:
· 2003/2004 -- Focus is enterprise IM, moving from proprietary to SIP
· 2005/2006 -- Seamless interworking. Conferencing begins to replace travel
· 2008 -- Seamless session transfer (never explained, maybe roaming)
· 2010+ -- IPv6 deployment, nobody needs phone numbers, VON is run online
(Comment -- This seems really pessimistic to me but maybe I’m not realistic about lead times for deployment).
Are we in "the perfect storm" or do we just have the bends? His main point was that the real technology disruption we are in is the result of DWDM and cheap fiber. VoIP is just a side show and not particularly significant to the telecom crash. (Comment – I never thought about this but I guess this is probably right.) He portrayed DWDM as "the Midas touch" -- something that looks like a real blessing but becomes a curse. The point here is that DWDM and cheap fiber looked like a real valuable asset when bandwidth was reasonably expensive, but they made bandwidth virtually worthless, so lots of people who invested on the basis of recovering money by selling bandwidth got burned. I can see this.
The telephone network has been all about managing scarcity. Trunks are expensive, switches are expensive, manage them frugally. The new reality is completely different. In an era of abundance, the result is inevitable -- bankrupcy for anyone investing in managing scarcity (Comment -- I see where he's coming from but on the other hand it's only abundant if you control it. Many companies out there are managing scarcity because the abundant resource is under someone else’s control.)
He gave an example of the gold rush and the fact that the merchants made more money than the miners selling aluminum pans and camp gear. Why? Aluminum was very expensive then. In the 1850's, people made jewelry out of it. In the 1880s they figured out how to make it from ores and the price dropped sharply. All the people who were making jewelry and other expensive things went bust. It took another 40-50 years for the market to really recover in volume (2 world wars and finally the aluminum can provided enough volume to recover the same amount of money being made earlier). Really interesting and a bit scary -- the timing of the recovery is independent of the crash. He claimed big dislocations like this are very rare.
He showed some views of Cisco’s VoIP market. It's been flat to declining in the last year. Cisco's PSTN stuff is still selling well, 30 softswitches/quarter, but not selling at a growing rate. He talked about some customers.
He talked about the spending freeze.
All of this will go away in time, but it's painful now. One thing that might unlock things is a competitive threat -- if VoIP becomes a viable differentiated service that everyone has to have, they will spend on it.
How should companies survive?
The freeze isn't global.
When the freeze unlocks there are 3 levels of vendors.
There is a lot of "Stealth VoIP" around. RBOC's are testing it, just not being very public about it.
WiFI(802.11) is going to be a huge disruption. VoIP on WiFi is going to be next. WiFI fills a niche that is compatible with 2.5G data. Use Wifi when you can, and 2.5G to extend range at lower speeds. VoIP will drive WIFI into the enterprise -- wireless PBX is a big win for managing enterprises. Have to solve security problems (Comment -- and again watch out for the radio bandwidth)
Conclusions: Freeze and turmoil will continue for a while. It's not the end of telecom. It will result in a more level playing field when things thaw.
Fiber is great, but it doesn't go to my house, and it doesn't go into most buildings. Until it does, that abundance doesn't get to me. He started with the characteristics of smart networks. They were okay, but you can only be so impressed with call waiting (i.e. only so much you can do). Attempts to smarten networks have broken services (interesting example he gave was that AT&T's TrueVoice service, which boosted the bass frequencies to compensate for lousy microphones in phones, broke some modems).
So, networks are best a bit dumb, but how dumb? The internet is real dumb. Do people really want to have to roll their own voice services from UDP and TCP? Not likely. Even the internet though provides some services inside (domain names, mail, etc.) Some services are best provided by the network but not necessarily by the provider of the bandwidth (Comment -- yes, this is a distinction few people make but it's a critical one.)
Don't try to reproduce the PSTN. We know how to do that, it's not interesting. What should we do different? Address the user, not just the network. Users come in multiple flavors. Turnkey users who just want basic phone service, some who want to customize.
Security is the hardest unsolved problem.
He went through a ton of applications, mostly presence based, that demonstrate the power of IP telephony. Nothing really new here.
Wideband voice -- make IP voice better than the PSTN (Comment -- yes, I think this is a real possibility, but it's a need people don't know they have right now. Cell phones have trained everyone to accept really terrible voice).
Combine the corporate directory and presence (comment -- yes, lots of interesting services. At some point though people will get worry about being watched).
One model of access based on URI's is take your phone with you. Plug your office phone in at home and have all the same services. (Comment -- sure, but why do I need to haul yet another technology toy around with me? Why not build the phone dumb and the intelligence somewhere else, not in the connectivity provider, so I can use a cheap generic phone on both locations and not have to lug it around?)
Video isn't for consumers for the most part. Consumers don't really seem want it for much. It is for enterprise -- avoid travel. (Comment, but it has to be very good to replace travel. My experience with video has always been poor. The problem is that you don’t control what you are looking at. It’s that ability – to control what you look at, that makes face to face meetings much more valuable).
User interface is the key. He gave some interesting examples. Mitel makes a plug for your PDA that lets you use a PDA as a big screen interface to a dumb little phone. OnRelay makes a PDA app that displays your office phone. He made lots of points about how adding a better user interface opens up new services. (Comment -- sure, but I think part of the problem is we just aren't being creative enough about what we can do with a voice interface. If Bill Gates wants you to talk to your PC, why is the phone industry moving towards typing at your phone? Wildfire and similar services prove you can build intuitive interfaces with built in help without a screen and with only a 12 button phone, but it takes some thought and DSPs to do it. Expensive DSPs and connectivity to DSP service nodes priced this out of the range most people would buy, but that's no longer going to be a limit. Wildfire is still out there and someone from Wildfire attended the conference.)
VON needs to be just as good as PSTN (reliable, easy to use, high quality, etc.), plus a lot more.
What’s the Path to VON
(Comment, this is a very incremental view, not surprising)
Platform requirements - they are in the midst of working these out. When it happens, that unlocks $3-4Billion in capital from BellSouth This will happen when the technology reaches “carrier grade”. To do that it can’t just be cheap to buy but must be scalable and be cheap to operate.
Must have service creation and evolution - not just a one off, but uses service creation to extend to whatever they want.
Summary - VON deployment driven by features and operations. Must be carrier grade.
He gave a very realistic and pessimistic view of the situation in which an ILEC is stuck – declining minutes and declining subscribers.
He started with a view of the world in 1995 -- no real internet, legally protected monopoly carriers, little use of wireless, a stable business. What’s happened since:
Competitive market now has no natural monopoly:
Question (Carl Ford) -- what about your capital budget – Answer going down.
What's the priority? -- whatever makes money!
Vingt Cerf is one of the architects of the early internet, and has been involved in a lot of the evolution of the network over a period of 30 years. He received the William McGowan award for communications, presented by Jeff at the conference.
He had no slides, saying "Power corrupts, and Powerpoint corrupts absolutely".
He talked a bit about McGowan (founder and CEO of MCI), and how when he first met him at MCI. McGowan told him that MCI didn't need R&D, they would just dangle a big check in front of a vendor and get the vendor to do the R&D.
Cerf came back to them to help them build MCIs first email system. The way they got people to use it was that McGowan directed his people that they would get all their instructions via email. Getting things done takes leadership. That's why we were here.
Worldcom today announced their Worldcom connection service, a SIP based VoIP service. One thing he noted is that it's a fixed price all you can eat service. This is fundamentally the way the internet has always worked. VoIP is transforming the voice market like this. (Comment -- didn't AT&T announce a fixed price long distance model recently?)
He sees this as the leading edge of a transformation in the industry. (Comment -- fixed pricing sounds like a great idea if you are a heavy use or if the thing being priced is cheap enough I don’t think many people go for those fixed price – all you can fly airline tickets)
He talked about how MCI/Worldcom moves terabytes of call detail records now. If it's fixed price, that goes away. (Comment -- really? Can they avoid detail recording for use by law enforcement.)
The phone network solved the equal access problem several years ago. Today, if you are a dialup user you get equal access, but broadband doesn't -- you are stuck with the ISP who provides the DSP/Cable modem. (He also made a comment that what we call broadband now will not see so in 10 years, I agree).
He wants to see equal access for broadband, feels it's technically possible, but we need regulatory framework and a business model. (Comment -- the consumer impact here is of course yet another monthly bill. Let's be careful).
He gave credit to Henry Sinnriech for SIP. SIP isn't merely a protocol for voice calls, it's a general framework for end-to-end negotiations of services. SIP not only supports peer-to-peer, but also client-server. (Comment -- not quite sure about this one, MGCP is more adaptable for this one). His example was negotiating whether or not to send large files to an end device.
Enum is very important element in defining interaction with the public net. It has value even if you stay on the internet. One of the problems with enum is political – The US and 17 other countries share country code 1, and nobody has resolved who owns it and populates it. Another problem is that there have been advocates for letting anyone register enum nmbers. That won't work. People can hijack phone numbers if there are multiple registrars.
He closed by reminding people that we are pioneers and need to stay the course.
Question: What kind of turmoil happened in the history of the internet. Answer -- lots. The early internet didn't come from connecting LANs, but from people connecting 3 wide area networks (one satellite, Arpanet, one other). This was 1973. Bob Metcalfe was still working on commercializing ethernet. Getting people to switch over wasn't easy. Computer science organizations didn't want to spend money on networking that they could spend on more graduate students. Arpa forced them to connect. In about 1982-1983, the internet had started to grow, and it was still hard to get people to switch. Cerf and Postel had to shut down the arpanet for a day to force people to convert. Had to issue a second warning (two day outage). Finally at the end of 1982 they shut it down for good.
That was only the
beginning. Now he had to fight with X.25
networks and the
Question: What's up with IPv6. He is of course a strong supporter. Billions of devices will be on the network. Have to maintain the end-to-end notion about the internet. End-to-end digital authentication was the one thing they didn't build into the original design of IP that he wishes otherwise. They couldn't enforce it because IP had to run over wires, with no device in between computers. It would have been timely because public key cryptography came in only a few years later. Nat/Firewall solutions won't support end-to-end encryptions (they don't really work either). The big problem with IPv6 is lack of momentum. It’s good that Windows XP implements it and routers implement it, but it hasn’t had a lot of testing. In particular there hasn’t been a lot of testing with the higher level protocols. Pressure for IPv6 will come from devices that will only talk V6. Talking between V4 and V6 requires proxy solutions (no way to manage the 128 bit addresses in the 32 bit fields in V4). Oddly enough, while the internet hates NATs, they may be the stepping stones to making it work. (Comment -> a freind of mine fought a long battle with the IETF for a solution he calls IPV8, which looks a bit like NAT boxes around a core of IPV4. One thing IETF didn't like was the NAT function). Note that secure IP connections doesn't make you secure. A secure connection can deliver a virus just as easily as an insecure one. Security means securing the operating system as well.
She put out an incredible amount of information in a short period of time. I can’t capture it all here.
SBC has been testing VoIP since 1996. (She had lots of information on various trials of softswitches, IP PBX, and other technologies.
The industry has to walk before you run.
There will be multiple phases.
· IP Enable your circuit switches (She had lucent AGCS all over the example viewgraphs.).
· Second is redo in IP (softswitch implementation).
She went through iMerge in some detail. One thing that she praised AGCS (lucent) for was publishing the spec for the phones, which gives them broad interoperability. She said that many IP PBX solutions stick you with the phones from the IP PBX vendor (she said SIP is a little better on this but not there yet).
She went through the softswitch solution -- 3 layers. Reality is that people bundle these in all different kinds of ways.
TRI has been trialing VoIP for years. First trials were not toll quality. Deployed some small scale trials of VoIP. One thing that limited them was that it was basically long distance, and they didn't have regulatory relief for it.
They have softswitches from 7 different vendors in their labs.
She described a 1999 trial in 3 locations. Lots of problems with echo control (too much delay). Also the echo surpression fix didn't work very well (People didn’t like the “push to talk” quality of the voice).
They had trials in 2001 of IP Centrex (Lucent and Nortel) (Comment – the lucent trial is the one in the Lucent facility in Lisle). She had some big concerns with security.
She also mentioned the Lucent deployment -- 1500 people in iMerge (Comment -- I never had the service on my phone, but it got mixed reviews from those I know who had it early on. Biggest problems were with things like conferencing and echo.)
Next Gen softswitch deployments -- one problem is that offices need to have some number of phones that work without power (emergency). Another issue is what to do about e911 when the phones move. (Comment -- when lucent deployed, they forbid people from moving their own phone because of concern over losing track of location for 911s. This wasn't a good solution).
New services: Web administration, integration with outlook, call logs, click to call.
There are lots of issues in getting ready for VoIP. You have to understand your LAN, moving to Switched layer 2 will improve things, but lots of engineering issues. There are also lots of issues with your usage pattern to understand how to engineer it and how much it will save you.
Key issues: QoS,
interoperability, standards (4 out there, none will go
reliability, "Need a packet buttset to test
connections). E911 and
(how do you reach beyond firewalls).
Some key VoIP data issues:
They are starting to see a 4th layer on top of the softswitch, like IN, that does things like understand how to get through NATs and firewalls.
She showed some Sylantro slides for IP centrex (softswitch based) for administration. Administered a followme/findme service. Some do time of day (not Sylantro).
She showed a Lucent solution for enhanced business services, it works with iMerge (conferencing, contacts, click to dial)l (comment -- this is like something I worked on at Lucent a while back, or more probably the descendant of the thing we were trying to productize).
She presented new and old market projections for next generation networks. (Basically bad news -- way down.) One factor is the expected move to all IP never happened. Instead we now accept that you have multiple networks and interfaces just a fact of life.
NGN isn't replacing the PSTN, it's different. There are some success stories:
question: Is the switch market saturated? Answer -- definitely not. There will be worldwide growth of 500M lines
over next 6 years. That’s just as big as
the last 6. It's not in the
I sat through just a bit of his presentation, his big theme was that broadband deployment can and must happen through communities, not necessarily through carriers. Communities in many areas have developed and deployed their own broadband solutions without support from carriers. It's a civic good and a vehicle for development.
Netrake is a manufacturer of something they call an IP
session controller – designed to help SIP sessions cross network
boundaries. He described the challenge as
interconnecting islands of VoIP in a
He went through lots of information on different interconnection scenarios and pulled out some common needs like address translation, firewall transition, mediation of signaling, etc.
The proposed solution is of course their product -- an IP Session Controller. This is based on network processor technology, not general purpose processing since it manipulates all the packets, including the bearer channel flows. You need to be able to manipulate the RTP packets and monitor them, measuring QoS and translating addressing and other issues. This is not a call control platform -- it's complementary. This is an autonomous system -- not controlled by something else.
The Session controller intercepts all of the signaling traffic in order to determine how to prepare the network for the RTP traffic. I visited the company’s booth and learned more. What they do is have a client behind the firewall that enables the to open all the connections they need outgoing (thus allowing them to poke through the firewall). The IP session controller coordinates with the client and maps the RTP and other information in the packets to allow them through the firewall.
Convedia builds a hardware media server (like Snowshore). I remember meeting Peter at an earlier conference when the company was still in “stealth” mode. Their media server is under control of an application server or softswitch (they implement VoiceXML and work with any. They have two models, a 1U minimal server and a rack mounted high capacity unit.
They are actively selling the platform in 150 countries.
It's ugly out there – The core
network is overbuilt, even in
His view was the media server
market is doing okay. Enhanced services
are one of the bright spots in the industry.
RFPs, trials ,and
deployments are everywhere (IP Centrex, Cable MOS's, ASPs,
IXCs, wireless, NGN -- mainly in
Bright spots in the market:
BUT -- things are going to get worse before they get better. There will be more consolidation (asset sales). Growth in services and new services are required to fill up the pipes.
Getting business is about
$$$, business cases -- new revenue or reducing cost, not philosophy. There is still $50 Billion/year being spent
out there (In the
The state of the industry: really bad. Intel's sales have fallen more in this recession than in any other (as a % and in real $). The semiconductor company stock index is now 1/6 of what it was at its high. He went through lots of depressing data about job losses in high tech and telecom. The telecom loss rate is higher in 2002 than in 2001. With that as preview though he reminded us that technology really does make a difference in everyone's life, and they are the ones doing it.
He talked about a couple of Intel technologies. The first was Hyper-threading. Micro processors today are only 35% busy internally. Rather than deploy two processors for boosting processing power, figure out how to make one microprocessor run two instruction threads. Lower power, and lower cost for double the performance. Expect to see this on desktops and workstations next year.
He talked about wireless LAN as a bright spot in PCs. Intel is building a new mobility oriented microprocessor (Banias). This includes 802.11b built in. Also lots of features for increasing battery time, such as batching of micro instructions (uses less power per instruction), and a power optimized bus (the bus is only powered when it's active). It also introduces a stack manager that increments and decrements the stack with it's own function units. (Comment -- the old PDP-11 minicomputers had instructions that automatically incremented and decremented for the same reason -- that's why the C language has those odd ++ and -- operations. The idea was kind of dismissed when stack pointers moved out of memory and into registers, but now with internal speeds so high it is attractive again.)
Finally he talked about integrated analog/digital processing. The latest technology is 90 nanometer feature size. Integrating analog and digital design modules lets them build things like software radios, integrated codecs with software defined coding, and more efficient processing than traditional DSP implementations.
He's the author of a paper on the Stupid Network, which claims that networks should be dumb. (Comment – the paper was written when he was with AT&T, and he left shortly thereafter). The talk was basically familiar stuff. For people who haven't heard it, the argument is that intelligence in the network just gets in the way of innovation at the edges, and we need to move to an internet model. That doesn't mean you can't do client/server with servers hosting services for end users, just that you don't want to build intelligence into the transport network. (Comment -- I think actually all the problems with firewall traversal, QoS, and internetworking are pretty good support. The only problem is that security and QoS are real problems and it's not clear how you solve them without making the network just a little bit smart).
He talked about how the architecture of IP, in particular the lack of link-by-link error checking is what enables VoIP. (Comment -- this was one of the big innovations in an earlier piece of work I did called Fast Packet Network. I think the idea was also in Datakit, though both of these were virtual circuit oriented networks. It was very controversial at the time but we could prove it was better for both voice and data to leave error checks and retransmission at the edges).
He said that none of the recent killer applications were invented by Telcos (he had a familiar list of internet applications) (Comment -- gee, what about wireless?)
Stupid networks can fail cheaply -- services at the edge that don't work don't hurt anything, can be introduced cheaply, and then dismissed. He mentioned AT&T had a big failure with an intelligent service in the network (Comment, I'm not sure which one he is talking about since there have been many things that could be considered here).
He said that Gateway based VoIP with phones or PSTN at the edges is not a stupid network. It's limiting, telephony focused and basically a sustaining technology response to what is fundamentally disruptive (VoIP over stupid networks). He said that VoIP gateways are actually a temporary business, as VoIP endpoints multiply the need disappears.
He went through who should build and own the stupid network and what the business model should be, but it was more questions than answers. He gave an example of why it shouldn't be the Telco's. Steam power killed all the sailing companies in the 1870's. Steam power was really very little different except that it allowed ships to operate on schedules, and it required fuel at ports. Apparently the steam companies couldn't operate it. VoIP is much more different so from circuit switching, so why should we expect telephone companies to manage it? Here were some of his candidates for running a stupid network:
Nothing really on the economics except to point out that it's much cheaper to fiber the world today than it was before.
They are a hosted services IP PBX company. He gave an analysis of service provider spending leading to the crash. Historically it rose at 3%. Then from 1994 it rose at 13%. 13% might be sustainable but by 1998 it was rising 42% a year, clearly unsustainable. We overbuilt $50B and are digesting that bubble now.
Overbuilding has a benefit. T1's are cheap -- $75 point to point in the SF Bay area
All the traditional revenue sources for carriers, including services, are really commodities. That means declining prices and margins. What will work is bundling and personalization. By bundling services and personalizing to the customer you add value people pay for (the analogy was a custom PC from Dell). (Comment -- maybe for enterprise, I'm not so sure in general. I thought most PC’s purchased were pretty generic.).
Hosted service provider model is emerging as a new way to deliver services -- You host services for virtual network operators, who customize services on your platform and deliver to end customers. ) (Comment -- I did some work on this at Lucent in 2001. We had a hard time figuring out how to develop the channels to the virtual network operators. The business is typically worth $15-$20/user/month to the application service provider and $80/month (including usage) to the VNO.
He invited someone from Telia to complete his presentation. (Comment -- Lucent work with Telia on stuff like this too. They are quite innovative). Telia is the Swedish PTT. They are merging with Sonera (Finnish PTT). Why are they interested in hosted IP PBX?
Question -- what's the compatibility with CPE, is that a problem) Answer-- it hasn't been an issue for them. CPE based PBX's don't really coexist with their hosted deployments in the same customer.
(Comment -- Wow, practically an empty room. With all the vendors here I can't believe it. It would seem prudent for any vendor to show up when your customers offer feedback if only to show you are interested. A few more people showed up over time but still less than I’d have expected)
Ibasis is an international carrier. The have had lots of issues with vendor interoperability.
(Comment -- is it any simpler in the SIP world? I doubt it.)
They have done interoperability with a lot of vendors on H.323.
TheybBelieve that interoperability won't necessarily stop VoIP, but we will have more proxies and workarounds to interoperate in the mean time.
Interoperability is more than just getting Voice through it. Addressing, signaling, services, and AIN on the PSTN also need to interoperate on the softswitch.
VoIP Table stakes include popular PSTN features, including business functions. You have to provide interoperability of these features with PSTN implementation. Have to support things like calling name (this requires accessing the name database using SS7 or some other interface they give you).
Value added services (web provisioning and real-time provisioning, rich signaling and user interface) come next. These have to be usable.
Potential evolution: Web based control center, personalization, PAM, device adaptation.
Lots of new interoperability issues raised by the new stuff –
Carrier interoperability in the IP domain including services becomes an issue. IP to PSTN, IP to ATM, IP to premesis networks, etc.
Some service interoperability issues:
Consider how bad it can get when you have voice mail involved -- leave a message through one coded connection, code/decode in the voice mail system, then retrieve through another encoding (comment -- ouch! I thought two incompatible cell phone technologies interconnected with a low bit rate VoIP link was about as bad as it could get. This is worse).
Ultimate test is the user -- universal service, feature compatibility, high quality.
A real argument for open standards since carriers have to interwork with others.
Bob Weinski (Verisign (AKA Iluminet)).
They run a nation wide signaling infrastructure as well as providing billing and other services. 1000 carriers use their infrastructure. His talk was about IP Enabling their signaling infrastructure.
Steps in evolution to VoIP:
It's happening at all levels now. Not a major part of the infrastructure yet, but we aren't waiting until everything is done to move up to the next step.
He gave a long tutorial on call setup and interworking. Piles of protocols and structure. Bottom line is that they have:
Convergence is a myth, we are diverging. The pictures are much more complicated now than they were back on the good old days. (Comment – he’s right. The fundamental rule of evolution is that the old stuff takes forever to go away and clutters up viewgraphs and requirements documents for years after it becomes irrelevant to revenue and profits.)
Question -- which protocol gave the panel members the most interoperability troubles, SIP or H.323. (Someone immediately said "MGCP") iBasis said H.323. Chet McQuade said things were more complex with H.323 thoughH.323 is more mature.
Question -- how does Bell South see the importance of operations. Very important Don't make your operations for VoIP fit the PSTN mold though, start by doing competent support for new technology while they transform their legacy support systems, then hook up.
This one had a bunch of people in it. That probably fits the interests of the audience in finding ways to define the value we can provide and sell products.
Packet has won, but the key is value added services. This session is about how to do that.
Went through the basic motivation for softswitch, and what the applications are. Why is the softswitch consortium important? -- carriers can't do one stop shopping any more so they need interoperability, can't rely on the vendor.
Service providers are not completely frozen, they have plans for network modernization. Major problem is that Marketing people control the expenditures and the VON community hasn't educated the marketing teams in RBOC/ILEC companies on the value for deploying packet technologies. (Comment – I’ve seen examples of this one too)
They are an IP service provider they provide wholesale services to alternative access providers for enterprise customers. Market opportunity -- $9 Billion hosted services market. 60% of it up for grabs in next 3 years. Big growth in IP Centrex and IP Voice services (web voice services). Softswitch market is growing (140%/year).
Some examples. Small multi-site business. For $50K investment, you get $40K savings in telecom costs in the first year, and $250K in operations cost over 3 years. Pays for itself in 9 months or less. Very good.
Cost to the end user is $370/endpoint capital on average (phones, routers) plus about $73/month in usage and service charges. (Comment -- again, this isn't cheap telephony, but business telephony isn't cheap now).
This is managed IP networking, not public internet
He presented a very complete
and complex case based on work with a large ILEC in the
Residential (high end users) -- Voice over IP using IADs. Video, voice access to internet services (voice web?)
They use an IAD to address POTS and ISDN phones. Also support SIP phones. They analyzed it with 2 softswitches in 2 cities plus one to act as a SIP proxy. The solution interworks with IN.
He went into a lot of analysis on the economics. Basically the business case for going IP is positive if you are first to market with or without significant competition. It's neutral to negative if you are last to market. He went into lots of sensitivities, I'm not sure whether he really covered all of them.
One message -- Voice is the big driver. Without putting voice on your IP network you don't get a return.
-- how do customers take advantage of VoIP replace the PSTN or something else.
For him most are combining new and old objectives for it. VoIP gives them the chance to look at it as an opportunity to extend their solutions (remote workers, new solutions to CRM, etc.
Service providers are used to one stop shopping, lucent and nortel did it all. Nobody does that now. Vendors and Service providers are pointing fingers at eachother looking at who does integration/interoperability testing. The service providers don't want to do this because they don't yet have the competncies.
Question -- what's the driver to move from arbitrage to services. The impediment is complexity. Arbitrage was easy, minimal interworking. Doing services is hard. Really tough to do something hard with a customer who is just trying to save money.
Where is the money spent in networks -- endpoints! Getting VoIP deployed means spending at the endpoints. The PBX replacement cycle is 6-10 years. it's 3 years till you get to 50% penetration, 10 years until you replace them all. The good news is there's 10 times the revenue in edges and endpoints as their is in core carrier networks (Comment -- a very interesting view, and a correct one. edges and endpoints are huge opportunities).
He was very upbeat. Siemen's enterprise business is up 300%. This was his focus (Comment -- Siemen's had booth in the show and it was large, prominent, and enterprise focused).
More risk in staying TDM than going VoIP now. VoIP will be the way to go for enterprise and the risk of not going is missing services and getting stuck with old technology. The benefits of this go to the IT department and the customers who have small locations. FedEx was a cited example. (something like 200 regional centers using VoIP)
Enterprise market will turn in 2003, maybe later if uncertainty over Iraq continues. The critical problem is lack of benefits to the end user. A PBX on a packet network is not exciting. If that’s all we do businesses will replace circuit with IP when it breaks which isn't happening fast. He cited another customer example, Volvo, with a huge network. They are using a wireless phone as the primary phone and implement roaming calls using the company VPN (if I use my phone in another Volvo site I suppose this means the call goes over the company net).
PC/Internet was driven by applications, we need them to drive VoIP deployment.
What's the impact of the CLEC collapse? less than you would think from the press. 5 years ago the RBOCs wanted ATM. There were early adopters of VoIP (QWest, Global Xing, and others). Early adopters suffered some pain. The good news is nobody wants ATM any more. The early adopters of IP have functional networks. 3 Billion minutes/month go over Sonus gateways and softswitches. millions of ports, 100's of softswitches. BellSouth, QWest, Deutsch Telekom are all customers.
Moderator -- how can everyone here be so optimistic?
Basic answer -- everyone needs a phone. Equipment needs to be replaced. right now we are living on borrowed time. Spending must resume soon. All the panel participants plan to be there when it does.
This was really a coordinated show of the partners in Equant's hosted services network. Pretty well attended, probably about 25 people, which is good considering it was the first session of the day.
Equant is the largest global IP and data service provider for multinational companies. (huge business, 250 people staffing help desks). Talked about the needs for business hosted services in 1999-2000. Service differentiation for service providers (Comment -- he had one of the most annoying presentations I've ever seen/heard, a LOUD doorbell every time he changed slides. Don't do this!)
Cisco's offerings (in partnership with call control, billing, network management partners) include:
Multiple companies can share
the same network. He talked about the
network built for Equant that was built by HP, Equant and NetCentrex. They emulated the real world environment with
a mock up network in
Cisco supported Equant's marketing of this service with tools and presentations.
Netcentrex provides a centrex (hosted PBX) function in both IP and SS7 networks. 50% of their business is circuit. They began to build a solution based on H.323, and then wound up making it independent (SIP/H.323). They provided a centralized call control center for Equant that drives H.323, SIP, and mixed networks. There is a front end that addresses the individual domains (proxy/gatekeeper and even a controller for MGCP endpoints).
Equant manages their whole worldwide network with 2 softswitches (1
HP did the global integration for this. The VoIP VPN Application and interfaces where done by Atos on HP open call.
Cisco was selling 3,000 VoIP phones a day 3 months ago. (Comment -- that's a million a year, believable, though at something like $500 each that's a lot of money, more than many people gave as the market size). Problem is nobody provides a VoIP network for them so companies buy gateways and interwork in circuit. That's the opportunity for Equant -- provide a hosted network solution for all those endpoints.
Equant began by offering a VPN core connecting to PBX's through VoIP gateway. Then they deployed large IP PBX's connecting to it, then on to small systems, and finally net centrex (no local call control). (His slide mentioned "support of the IETF call processing language here. SIP CPL? I guess so).
She described the architecture. Cisco provided the infrastructure, NetCentrex provided call control, HP Opencall provided the VPN service platform and provisioning platform. This was a redundant fail over solution.
He came on again to address what the situation was now in 2002 after this deployment. They are offering this service to their worldwide enterprise customers.
· Much more attention now is being given to managed IP versus public internet.
· Security is a big concern.
· Adoption speed is slower than expected.
mostly migration. A few are willing to
do radical replacement (radical may be complete within 2 years.) Only a few
· The migration case is hard because customers want to keep current features and interwork during the migration (Comment -- yes, this is the hybrid network case I believe is going to be very important).
What are equant's needs in 2002 -- IP telephony integration (better integrated products
The vision hasn't changed from 1999. What changed:
· A shift towards private networks
· Still looking for the killer app.
· Irrational pessimism has replaced irrational exuberance.
· Not everything wants to be outsourced -- you want to keep people working.
LOTS of speakers in thissession Moderated by a consultant (Lee Quintanar) He gave a very complicated hybrid network picture on what's in carrier networks and a set of requirements for it. This was an overflow crowd, probably over 100 people
Dialpad was one of the first IP telephony companies to have to restructure, but survived.
They have done over 1.8 Billion VoIP minutes. 700K call setups a day with 4.5 Million minutes/day. They are consumer focused (PC to Phone and Phone to Phone, calling card focused. International (Korea/Japan) and US.
They have a server (Vega) that controls call setup and does authentication. (looks softswitch like). Dialpad doesn't own the network. Their data center is 250 square feet (a medium sized office!) They use relationships with ITSPs to deliver. They are basically a retailer for Internet Telephony.
Their server will do H.323 and SIP, and they do dynamic call routing (least cost, Qos, time driven, etc.) Not a lot of demand yet for SIP (no SIP endpoints?)
Service is basically toll bypass. Some plans to expand (voice mail, presence, etc.)
They started as a free model,
65K customers signed up a day! They
needed a call center. Lots of users are
international inbound to the
Good news: broadband lets them go G.711 and even if they use G.729 they get better and better QoS. Price is down to 1.5 cents/minute. Broadband glut is making their value proposition better and better (because QoS is now much closer to the PSTN, but that glut has also driven down the price for circuit)
Support via Live chat and call center (they have email, but it's not a good fit with customers paying for service).
They are an all packet IP Clec aiming at small business. Denver and Washington based. Still VC funded (since 1999).
He gave an example customer being charged $1575/month for 16 lines of circuit voice, local, and long distance, plus a 384Kb/s DSL for data. Talking nets charges 1K for the same stuff run over a T1 with VoIP, which gives faster (up to T1) internet access, and more features. (Comment -- good value, but again not cheap).
They have a reverse gateway to connect to the local switch to do 911, a Broadsoft feature server, A sonus/TTI softswitch, Cisco IADs. They usually use a Cisco architecture on premesis with Cisco IP phones if they use IP phones instead of IADs
· nobody has it all, they must integrate best of breed.
· Interoperability is the big challenge. 3-5 months to get message waiting light to work. Vendors were helpful, but it takes time.
· They need to know the vendor roadmap -- don't surprise them with upgrades. They can give vendors good feedback as the conduit to the real end customer.
· Vendor sales force turnover is a big problem -- they have to retrain the vendor SE's on what they do.
· The cost of IP endpoints (including IADs) is still a big problem in making the economics work.
Deploying on customer premises is a big deal. You have to know how to do it.
Their system changes the way businesses work. They typically replace some creative jury rigged system to save money on telephony. Some businesses want to change their behavior, some don't. Don't assume your customers are going to want to adopt change, and don't assume they are dumb!
Talkd a lot about BT Ignite -- They inherited much of the Concert Assets (variously partnered with MCI/Worldcom and AT&T). They have an international VPN product. Their big product is inbound services (tying together call centers for global customers). Their core is a standard MPLS network. Lots of different applications (VoIP and standard IP. The talk had lots of detail on the architecture. They support a lot of different endpoints including their own endpoints.
Main VoIP issues:
Why should they deploy VoIP?
Sure, IP transport is cheaper, but why should I trash a fully depreciated 40% utilized network for 3 cents a minute. Look at the economics:
They are launching an IP Based multi-media ACD (partnering with Cosmocom). Features include Click to talk, web interaction (chat, co-browse, video, email). This gives them the ability to marry a fully functional circuit based call center implementation with a IP technology (not exactly clear what they really did).
What the big benefit is for this is that when a customer needs to set up in a new country, they don't need to set equipment up there. The equipment can be centralized (just like equant).
This was basically the same message as the earlier session on Equant. They are a part of France Telecom. They have 3700 business customers, 2/3 of the top 100 companies, $3B revenue)
What isn't working:
This is AT&T's service that interconnects VoIP carriers as a wholesaler. Provides routing/rating that lets them interwork, track minutes, resell minutes, and do routing.
They are a true clearinghouse. Their members who terminate traffic offer it at a price, AT&T adds their margin, and sourcing carriers decide who to used based on the price.
Their network is based on an OSP (Open Settlements) server on each end and a settlement back end system. These systems produce the CDRs from which they produce the bills. Their scorecard:
Between vendor inter-domain connectivity is a problem. Gateway implementations don't scale, gatekeeper inteworking has compatibility issues, border proxies look like a good solution, not mature.
They don't yet see SIP in the wholesale market. Everything they do is H.323. SIP might happen, but not there yet.
Question -- what's the traffic mix? (by revenue most interesting)
Question: everyone is on H.323, interoperability is a big issue. There is now an H.323 forum working on a certification process. Need participation from service providers.
I was there for presentations from David Gurle and Henry Sinnriech. Nothing new here. Both preaching SIP as the answer, walled gardens as bad, and the current state of the industry lacking standards as impeding development. David made a particular plea that Presence and availability is more than just IM or click to call. It's your on line identity. It's fundamental and key to get right. There are serious issues with privacy and identity theft that people aren't getting right now. (Comment – All of this is somewhat ironic coming from a Microsoft person, since I've heard many complaints about privacy and security in Passport, but he doesn't necessarily represent Microsoft in all his views).
Party. Jeff's 40th birthday party/concert was as all the parties, good food and good music, even if I thought it got a bit loud. I continue to be amazed by the Herding Cats, a band he had for an earlier party that he brought back for this one. I'm not sure any of the songs they played repeated the last show, and all were well done. The second band (The calling) was good, but the sound system was just too loud for the hall in my view. Almost painfully so at times. Really too bad because I enjoyed the music, from a safe distance. I have no idea how long the party went on, but suspect we wont see some folks today.
This session was fairly sparsely attended, maybe 10-15 people, with several leaving during the session and few questions.
They are an application server company. The motivation for carriers to look for new services is obvious, everyone is looking at service revenue to offset comoditization of their business and retain customers.
He had some slides comparing past and current requirements and characteristics of telecom services. In the past, requirements were precise, multi-year payback was acceptable, product offerings were generic packages, and services needed lots of up front investment. Today, payback has to be in months, requirements are never concrete, services are tailored, and nobody wants to make up front investments.
This demands a service creation platform approach that can allow a new service to be added easily. (Comment – yes, but if nobody is making up front investments, how do you sell investment in a platform?) The platform has to be easy to implement and integrate (shared application platform). Large vendors are not doing nearly as much system integration now so it can’t require effort.
Doing IP and SS7 services is important. Carriers are buying services now, in advance of the decision to go VoIP, but don't want to strand investment.
Net.com is another application server provider. I first encountered them as supporters of a new forum on service creation that was really focused on integration of all the components needed for new services. I haven’t seen any recent news on this.
He went through a lot of the same background as the IP Unity speaker. His service creation architecture is tools creating scripts. Scripts run on an execution platform, which in turn drives some kind of network interface. He talked about being able to drive H.323, SIP, or SS7 off the same platform.
Their platform (Shout) -- has a built in media gateway, as well as SIP to H.323 interworking. (The platform seems to have all the usual APIs)
He gave an example service (
He gave an example of a legacy case, a US LEC asked a vendor for calling line ID enhancement on its class 5 switches. The project cost $4.5 million in development, took 14 months, and cost $25M to deploy. Not exactly what you want to have for service creation. (Comment – from the questions after the session I believe this was a Nortel switch modification, but it is a credible story for any switch based development)
Pactolus started building their solution in 1999. They have an application server based on SIP and MGCP. It can use an integrated or separate media server. Their Prepaid solution does 150 Million Minutes a month. They also have a big (96 port) conferencing solution.
Their competition is basically TDM platforms. They have an integrated media server (software), and can run several thousand call sessions of prepaid on one 1U Netra.
Requirements for SCE -- Standards based, pure IP software architecture, etensible SLEE, efficient and reliable. (Nothing seemed really new).
He talked about their SLEE as being XML based. Not clear what that means. He highlighted the ability to customize units in the SLEE: Examples being multiple languages, different signaling, etc.
Their service creation is based on a palate of building blocks, with the ability to call java scripts. Building blocks for call control, media control, and some with raw SIP in them.
They delivered prepaid to a carrier that delivers 80M minutes of prepaid a month. It was a 6 week job to integrate their solution with the carrier's customer databases.
Reliability – Their platform is designed for 5 9's. Scalable to 1 Billion minutes of prepaid a month.
I asked the only question -> Is there any service portability between your platforms. The answer was basically only with support and work between the vendors. I think the industry will score a milestone when we can achieve service portability.
We had 8 people, which was 6 more than the attendance in the session before us (Operations support), and all seemed engaged and stayed to the end, which I thought was positive.
I gave an overview of what hybrid networks are, where the arise. Basically hybrid networks include multiple protocols and multiple architectures for service delivery. They arise from carrier migration, mergers, and new service offerings. They are a fact of life for many years to come.
The issues include duplicating effort in developing services, duplicating equipment in delivering, and hairpinning or tromboning. The need for continuing investment in old technologies can also be an issue.
I overviewed several approaches:
Everyone seemed engaged with this.
Doug gave a Jain overview focusing on Jain SLEE. He was very supportive of me and Personeta, and gave a lot of good arguments for JAIN and for JAIN SLEE. I was glad to see him highlight service portability, particularly after the earlier session.
Lou is a former MCI/Worldcom employee and architect of their IN system. Net Direct is an integrator. He was there to give real world experience, of which he has plenty. He described two projects.
One was a voice portal with both IP and PSTN endpoints. Several vendors, several 3rd parties delivering content, and lots of integration. One interesting thing was that they had a major problem with speech recognition in SIP. This happened even with G.711 (64Kb/s) coding and good networking. During the questions the reason became clearer. The trouble is that the sampling rate of voice generated in the SIP endpoints isn't synchronized to the network (your SIP client in your PC uses its own clock. This isn't a problem with normal voice, but the recognizer was quite sensitive. Even after a lot of tuning it's much less good than PSTN. Even tone recognition is an issue. DTMF recognition rates were 92%, which means in each credit card number one digit is wrong. not good. They had lots of interoperability issues too. Perhaps the biggest experience is that the project was a success, but the business, hosted VoiceXML portals, hasn't been. The customer signed up far fewer enterprise customers to host portals than they had planned.
The second project was a huge call center: 1200 agents growing to 5400 agents. Here the customer had specified a what may be worst case test for the open systems approach: Multiple vendors for component, a total of a dozen. The network brought calls in using a softswitch and Sigtran based signaling to an SCP that directed calls to VoiceXML based IVR and eventually to agents using SIP terminals. The main issue here was integration headaches with incompatible protocols. Even after getting it all right, everything broke with each major release. The project presents an object lesson in not betting too heavily on open, standard interfaces to insure interoperability.
He had a lot of questions from someone who turned out to be with Wildfire about the speech. It's getting better, and SIP signaling for tones (RFC2833?) will help the DTMF problem but it will take a long time in deploying.
The SIP-IT event was going on next to VON, as was an ETSI TIPhon meeting, a JAIN meeting, and Pulver’s Presence, IM, and Location events. (Comment – in better times these all were separate events, but actually co-locating them was I think a real benefit in bringing together a lot of people) I visited the SIP-IT event to see some old friends. SIP-IT is actually the 11th SIP interworking event. (The early ones were called “bakeoffs” before Pillsbury objected to the use of the term). One interesting thing was just how different this looks. A room full of programmers hunched over laptops. The SIP-IT folks had separate meals and rarely mixed with the other groups. My contacts (Lucent and ex-lucent people primarily) told me that this event was about half the size of the largest ones. Very few companies had integrated the latest SIP RFC (3261). That struck those who had as odd since the draft has been out a while. There were several participants who had been at most of the events, though there were established companies missing and new entrants. Much of the effort is still going into basic interworking – getting a basic call to work. Not advanced scenarios like 3rd party call control, refer, record route, re-invite, and others. The events do, however, no doubt have a positive impact on the industry, allowing the developers to work together one on one in a non-competitive environment to get the implementations right.