This was the 10th anniversary of the Voice on Net (VON) conference. Over the past 10 years the conference went from a small gathering of hobbiests and academics to a major commercial technology show by 2000-2001, then like the industry hit hard times and shrinkage. In 2005 though the show and the industry have recovered, making this one I believe the biggest ever and judging by the size of the exhibit hall probably the most commercially significant. I spent most of my time in the exhibit hall meeting with potential customers and partners, and some time in keynote and perspective talks and a few sessions.
Jeff Pulver has been using the show to promote the message that VoIP will happen and take over in the current year for many years. This year I believe he is correct. Voice over IP is mainstream now, with major investments being made by carriers of all kinds, and subscriber numbers climbing into the millions. Some indications of this:
SIP is no longer Simple – One speaker talked of 2,000 pages of draft specifications. In a simulcast from the IETF meeting, extensive work by 5 or 6 working groups focused on issues related to SIP was reported. I recall not that long ago (1998) hearing Jonathan Rosenberg (then at Bell Labs) walk through how SIP was so much simpler than other telephony protocols. The original view was no doubt far too simple, but looking at what happened I wonder how much of the complexity now being introduced is not essential complexity, but is instead an accident of a “loose” standards process allowing multiple competing usages and extensions to evolve and of an insistence on endpoint focused solutions to all problems while network based or hierarchical solutions better fit some problems. In an interesting twist, there are defectors from SIP in favor of new “simple” solutions based on peer to peer networking. An entire session and a large booth was devoted to an open source feature server and a simple interface protocol (AIX) gaining support, and the large base on non-SIP solutions like Skype continue to suggest the future will not be all SIP.
There is still no agreement on what “telephony” will mean in the future. One of the more surprising themes to me was that virtually all of the carriers who spoke on their VoIP plans basically talked about VoIP as an incremental technology, allowing consumers to get cheap flat rate phone service and use their existing analog phones with minimal features. (One cable operator even said their customers didn’t want a lot of features in their bundle because then they felt they were paying for things that they didn’t use). In past VONs and in my personal opinion, IP is a disruptive technology, and one of the major reasons to do it, if not the major reason, is to be able to offer voice services that don’t look like POTS, because they escape the limitations of a 12 button user interface with no visual feedback, and a 3Khz audio connection. In fact many including me, would argue that users (consumer and enterprise) really could care less whether their voice is in circuits or packets, but will go for solutions that provide better ways to manage their communications at lower cost, and many of those enhancements can be done by applying IP and IT technologies to the signaling and control of telephony without putting the voice in packets. (Putting voice in packets is of value to users by reducing costs associated with a separate circuit network, but no user visible functionality depends on it.) Some carriers are introducing IP controlled enhanced services (for example Verizon IOBI), but the focus seems to be on cheap Voice service transparent with POTS.
The focus on “Incremental” introduction was reflected in other ways – Vendors (Sonus and others) talking about building IP telephone out from the middle of the network. At least two financial community presenters presented projections of equipment spending that looked like a normal technology replacement – circuit falls off and IP picks up while total spending remains almost constant. This looks much more like the well regulated days of the Bell System (digital transmission replacing analog slowly over time) than the “wild west” environment always posited for IP and internet services. If IP telephony really creates a significant advantage for users and carriers (and I believe it does) I would expect to see a stampede – a large peak in spending in the near future as everyone rushes to get the equipment in and not be left behind, then a fall back. Perhaps we are still feeling the sting of the internet and IP bandwidth boom/bust cycle of the late 1990s, but I have to wonder whether the speakers and analysts have it wrong: Is the future really going to be that most of us will still be using analog phones over decades old twisted pair household wiring connected not to a line pair but an adaptor that converts it to VoIP to go out over our broadband connection – but with the only visible change being the flat rate unlimited calling plan? Looking at how people use Mobile phones and how they have evolved from Trimline phone look-alikes to communication devices Captain Kirk would have considered advanced I really wonder. The future I see has mobile phones, and IP communication devices (with screens, video, maybe broadband voice but definitely with better ways of calling and answering than pushing numbered buttons and picking up the phone with little information about the call), and makes me wonder whether the phone and cable companies of today are going to be the ones who provide it.
Network Interconnection is becoming a hot issue, and with it Border Session Controllers, edge routers, and other devices to enforce interworking at the IP level between carriers. I remember arguing that IP based carrier interworking was a key issue back in the late 1990s, but at the time everyone saw VoIP as either something buried inside a carrier network (with circuits at the edge), or end to end over the internet. To get QoS and support service level interworking though carriers have to exchange information about the nature of the IP packets that they are exchanging and to avoid cheating they have to enforce agreements, and that’s where these boxes come in.
I didn’t see a lot of NEW services, but that doesn’t mean services weren’t important. Several vendors demonstrated “Converged desktop” services. Push-to-talk mobile services were in evidence and presence enabled services were there as usual. A lot of vendors in the exhibition were showing video. My impression was there was some shift from focus on enterprise solutions towards carrier based solutions and solutions for the general public, but that may just be a reflection of more carrier people attending the show than in the others I recall.
I attended very few conference sessions, spending most of my
time in discussions on and outside of the exhibit hall floor, but here are some
I did attend.
VoIP Central Office session
This was a long session at the start of the conference, and
always pretty popular for the first day.
I attended a few talks in this session as they pertained to Personeta’s product offerings and customers. It was divided into several sections, the
first of which really covered “Class 5” like offerings
Bay Packets Speaker -- he covered basic call control models. I only heard the end. His messages:
There's a role for JAIN (abstraction but power) Parlay (3rd party security) and SIP Servlets (enable the Servlet programmers).
SIP Servlets are good (enabling to Web programmers) and bad (Not abstracted). They address the bad by providing building blocks (Comment -- looked like pre packaged handlers for multiple transactions in a call flow that implement some primitive, still fairly primitive though).
Feature interaction is key and has to be addressed. (Lots of questions/comments relating to this and in particular to IMS as the model of how to do it, but much detail or convincing argument on how this would really work.)
Ring Central – This company got its start doing Internet call waiting and grew into a full service VoIP company (Comment -- I dimly remember them, though the leader in that market was a Canadian company bought out by Nortel. I don't think they used AIN/IN for this, but used plain old call forwarding). They serve soft phone clients (Alpha site testing now. )
He showed a backbone oriented architecture (Looked a lot like the SBC Unified Messaging architecture, with web, billing, message storage, notification, and database as resources, plus some applications and interfaces. He said adding SIP was very easy (just added it as a proxy onto the backbone) (Comment -- but this tells me SIP isn't really integrated into the applications.)
Messages: Modular design is critical and powerful (not much to this talk really).
Questions -- IP QoS was missing, how do you address it. (Speaker didn't seem to have a good idea).
Another speaker commented that IP is by definition best effort and suggested huge pipes or RSVP and MPLS as a solution, but not everyone implements RSVP or MPLS (Comment -- Nothing new here, I heard this 5 years ago What was surprising was it did seem to be new to some of the audience, I guess this field is maturing and more people are coming in and being exposed now).
The other speaker (From Cablerock, I didn't hear his talk) talked about how they were a service provider (sounds like they wholesale to cable companies) and have to address QoS internally, but the issues around their interfaces (access, core network) aren't under their control).
Question -- how do we manage simple things like telling routing that paths fail. Answer -- no standard for that. Most use SNMP to manage softswitches to enable/disable particular paths if they have to. Otherwise it's up to the routers. This is not unlike what happens in the circuit world, but the packet world has another degree of freedom here which may not help. With circuits, either a circuit works or it doesn’t, and if it doesn’t you can’t set up calls and the switches detect it and route around it. In the packet world if a path goes down and gets rerouted at the router level it can continue to work in degraded quality -- you still have connectivity and call setup will still work, but delay and loss are out of limits and it's harder to do. Speaker said nobody is working on an end to end solution, but someone in the audience said MSF (forum for soft switching standards) is working on interface from routers to softswitches for this.
Cablerock -- doesn't use RSVP, does load balancing across 4 interconnects with partners, and does MPLS not end to end but with their partners (they can't get it end to end).
Many questions and discussion on how to get end to end QoS. Bottom line -- unless we are willing to sign up all the carriers to do end-to-end QoS solutions we won't get it.
Question to IP carriers audience, how many are facilities based and how many are "in the ether" -- half and half.
Question -- how many are looking to be broad, anyone-to-anyone carriers like Vonage, how many are looking to serve only internal communication to their customers. (The response was maybe 1/3 broad and 2/3 internal)
Interesting response from Cablerock
-- they have lots of calls to
Gateway Section
This section focused on the role of the gateway in the IP
Central Office.
Sentino
Messages:
Sonus (New employee gave the talk, not Grubner, the founder who was scheduled). Talked about evolution of VoIP – started out as chips and cards plus general purpose computers, low capacity, best effort, low bit rate coding, cheap.
Then came Long distance – replace PSTN – high quality, lots of controlled IP bandwidth, bandwidth isn’t critical so you don’t need low bit rate, but you do need echo cancellation (for quality) and modem detection for disabling any voice oriented processing like echo canceling. Here you separate softswitches and gateways.
Then came enterprise focused VoIP. This adds PBX outside of the gateway (Comment - looks like a plain old circuit PBX, not an IP PBX) Still no lines on gateways, just trunks, but more services and more complex trunking protocols. (DTMF detection mid-call, tones and announcements, bridging, etc.) (Comment – this is beginning to sound more like a media server not a media gateway.)
Also added detection of modem and redirection to TDM only routing. (Comment – I’m sure they do this but on the surface this is quite weird – the data traffic is shunted off the data network).
Also have to handle signaling – Mapping embedded SS7 (E1) to Sigtran.
The broadband – IP in from anywhere in the
role. He still sees a role for
media gateway – interworking with IP carriers and interworking with broadband access points. (Comment
– yes, in an all IP world if you are a gateway maker you position the gateway
as a border session controller, not unreasonable)
Packet Cable – have to do packet replication and DTMF detection for CALEA.
Wireless gateway/MSC – have to do transcoding and transcoder free end to end interworking.
Audiocodes (Alan Percy) – (He’s a regular VON speaker)
Audiocodes -- $82M revenue, 440 employees, 11 years in business. (Comment – still not very big for a market leader in their niche).
VoIP to the residence/business growing rapidly, key for carriers is differentiating their offerings. The past was very managed and controlled environments, so QoS could be achieved through engineering. In the past 6 months, many more deployments depend on shared infrastructure that’s not engineered to voice. (Comment – just wait until IPTV starts sharing the cable and DSL bandwidth with voice).
You can’t always control the network (have to use shared VoIP/IP infrastructure). Last mile is the hardest mile (and the one you probably can’t control).
Some effects – have to do dynamic jitter buffer adaptation (Comment - yup – I have a patent on this from about 1985).
G.723 (very low bit rate) can be better than G.711(64Kb/s) – low bitrate usually means less packet loss, plus G.723 has an adaptation to replace missing packets and produce degraded speech rather than just missing stuff. (Comment – but you still can lose phonemes and lose intelligibility).
Have to deal with problems like packet loss and background noise. There is now an ETSI speech quality testing regime for VoIP – (comment – Yes, I did an exercise on this for one of my classes a couple of years ago. )
Real message here seems to be that VoIP quality over the last mile is hard, it takes a lot of work and experience in the codecs to handle a bad network environment, and this is where Audiocodes excels. (Well, interestingly enough at Past VONs I’ve heard some really slick technology in this area from Global IP Sound. Those guys are here still, but I don’t know how real their solution is, clearly not as widely deployed as Audiocodes).
Intel (Walt Brown) – he is another old VON speaker.
Showed a bunch of architecture charts from 3GPP and TiSpan. Apparently they put these things on T-Shirts upside down so that people would have a reference to point at on their chests. (Comment -- Sounds really geeky).
Lots of complicated functions in a converged environment for media. He gave as an example transcoding of multimedia for support of emergency services (like being able to watch someone’s messenger supported web cam on a 911 call?)
Lots of new services (come from convergence of media and functions (chart of stuff like Pulver’s purple services).
Lots of new user devices (projectors, etc.)
User preferences
Multiple languages
Social requirements.
Intel Vision – everything wants to be wireless, people want to be able to everything everywhere. IMS architecture is the best current bet for a solution. (Comment – surprised to see Intel jumping on this bandwagon, and IMS is beginning to sound suspiciously like yet another bandwagon to me. It’s nice, but I don’t believe in one architecture for everything solutions).
How does IMS help make money – hiding the access network, pushing solutions to common problems down in the protocol stack, and hiding differences in each user’s bundle of services. IMS does allow more than one solution to each problem. (e.g. International roaming protocol allows use of multiple authentication, Radius/Diameter or SIM based).
He showed Next GEN IMS architecture diagram from ETSI TiSpan – (it looked more OSI-ish than previous ones which were more “network of boxes” than layers). Media gateway has interfaces across the architecture though so that it can handle authentication, content delivery, and other functions.
IMS is only one of several parallel service architectures in the TiSpan NGN architecture (Content and PSTN emulation are others).
Afternoon session –
Application servers
This was the section that Personeta and several others spoke in focussed on service delivery in IP and converged networks. The room was pretty full room, 100-200 at least. A good number of service providers.
Paul Nussbaum (CGI Logica) – Addressed VoiceXML application servers like the web application server architecture – same architecture, creates pages from back end dynamic data, the VoiceXML server is just like a web browser. You can reuse all the back-end you had for Web. (comment -- I think I gave a talk like this for a VON in about 2000).
There are other models – old IVR services that add VoiceXML, gateway based VoiceXML interpreters, commodity servers running software for VoiceXML.
Basic message: App servers are just like the web application server – with a different front end to do voice. All the management and execution control are like that. LogicaCMG is primarily a mobile company, #1 in text messaging and #2 in multimedia (3 of 4 text messages go through their equipment!) $3B company.
Paul Singh (Veraz)
Paul is a regular VON speaker and started an application server/softswitch company (IPVerse) which was acquired by Veraz.
The way it was – TDM switches plus service nodes dedicated to services. Each one had its own App server, TDM fabric, signaling gateway and media server. Wasted resources, multiple slands/smokestacks, etc. etc. etc. This is the reason not a lot of new services get launched. (Comment – I’d disagree with this, in that one of the big limits is in fact interconnecting with shared infrastructure, especially in operations. If you could do a new service truly as a stand alone it would be easier. The trouble is that carriers want the services integrated in billing, provisioning, and management and that’s what makes the service node model tough to extend, plus the fact that SS7/circuit is a tough shared infrastructure to integrate with.)
The way it should be – Shared network and resources, brokering through a softswitch to lots and lots of application servers.
SIP has currently won as the softswitch to app server interface (Comment – I think so, but I wonder whether we will see XML/Web Services interfaces for call control to a softswitch re-emerge as SIP gets more complex and implementing it becomes an entry barrier). Back end includes CDR (doesn’t matter which), SNMP and SQL, with SOAP/XML as the next thing.
He gave an example of a multi-continent network (
Bob Wise (Continuous
Computing)
Continuous computing provides Trillium protocol stacks. His talk was on ATCA, a standard blade architecture. ATCA is a next generation of cPCI
Bigger boards (more cooling and power – 200 W/boards) More and faster backplane options
Higher density (bigger boards) means smaller stacks. It’s midplane based, cabling interfaces in the back and processors in the front.
Also includes additional management software (insertion, removal, alarming) (Comment – that’s fine, problem often is that this stuff is bundled with fault recovery/relliance which doesn’t necessarily work for your application).
Typical boards – Dual processors, 16G memory, two base gigabit Ethernets, four fabric gigabit Ethernets.
Can accept mezzanine (PMC) cards.
You can build all the basic components for next gen networking out of this technology
(Comment – having had
experience in building fault tolerant systems out of this kind of technology I
believe it is the right thing to do, but achieving low downtime requires that
all the components work well together and that can be a real challenge in an
open environment where no one vendor has control of all the pieces as they do
in a vendor specific architecture)
Tellme – Hosted/outsourced provider of VoiceXML based services.
(Comment – Tellme is an old company in this industry, one of the first to really embrace VoiceXML).
What’s beyond VoiceXML – call control. VoiceXML is way too primitive – calls are implicitly answered, and you can either hang up or transfer calls, that’s it.
Call and Voice Dialog are one thing in VoiceXML. All calls have IVR dialogs. Not all calls need IVR, and not all IVR applications need it for the whole call.
He showed a network diagram with VoiceXML, bridging, and announcements all connected to SIP softswitch and under control of a B2BUA call control application server.
Call control application servers have lots of choices – multiple languages (C++, Java, Python, XML), and call model (JAIN or SIP (Servlets, CC/XML, and others)
(He cited Bay Packets and jNetx for JAIN call control). Iptel Application Agent (extension of a proxy server into programmable B2BUA, uses Python). Hughes has a C++ one
Convedia (Graland Sharratt) -- Another prior VON
speaker, Convedia has a big media server
Of course he gave a talk on media servers, not application servers.
You need a media server to do any non-trivial service.
There were a lot of interfaces to control media servers, SIP is the current winner.
He presented a lot of interfaces for interworking between media and application servers. His preferred solution was using VoiceXML and CCXML in the application servers, using MSML and MCML as the interface to direct the media server.
Personeta (Mark Mcilvane) –
Mark talked about the needs for application servers. I won’t go into great details since those of us with Personeta know the material.
Message – the next generation central office is an operating system, not a place. Applications have to be abstracted from the details handled by the operating system just as IT applications do. . Applications have to be independent of what the network and “operating system” because a lot of those elements will continue to vary and what you want to survive and be usable are your applications. Personeta’s product is a service delivery platform that addresses this problem.
Lucent (Jack Kozik) – architectures for application delivery.
He built up a picture of network architecture based on SIP (i.e. no circuit/IN connectivity to existing networks).
What do you do when you need a new server – have to add new vendors and solutions – may be able to grow in place, but the infrastructure needs to support adding partners and parties.
One problem – adding new servers may be an overlay – each
one may have it’s own client, registration, routing,
optimization,
SIP world solution – proxy servers – Proxy servers basically
allow clients and servers to share a dispatching layer that handles load
balancing, authentication, discovery. Not bad.
(Comment – Guess what, Lucent
happens to have one of the better proxy servers out there),
How to go even faster – IMS. That adds infrastructure for subscriber data, billing, SIP extensions, etc.
Question for Jack (where does Vo WiFi come in) – common service using WiFi to reach extensions. Some vendors make multi-mode phones. Both WiFi and 3G plug into IMS.
Question for Paul – Where are the App servers located? (Could be anywhere you want and they could be owned by anyone. His sample deployment was focused in one place)
One of the other speakers said App servers have to be close to their data because unlike the Web world a voice customer assumes they are disconnected if they don’t get a response within 2 seconds. We don’t have ways to prioritize access for “real time LDAP”)
Media servers (Convedia person) go in the center of the network for economy of scale, or near the access points for economy of transport (probably not as important)
Question – is there a standard benchmark for applications (no, everyone has their own, “vendors use benchmarks like a drunk uses a lamp post – more for support than illumination”).
Jack – recurring question from his customers is “how many customers can I support.
Question to Panel – what’s going to drive carriers to VoIP applications
Mark – Cost, and a new set of “network applications” (Good)
Jack – Convergence – carriers want it and bundling wireless/wireline is easier in a next gen architecture. (“Convergence on your phone bill only goes so far, you need a converged architecture to really deliver something that looks converged to the customer”)
TellMe – Really good voice quality – IP can deliver better voice quality than TDM. (Comment – I tried to sell this one for years, but I never got anyone to take it). The moderator asked if anyone was doing wideband voice – Skype was the only answer. (I think this is sad, because I think it’s a major opportunity that the industry is missing. There’s nothing sacred about 3Khz audio).
Another answer – Agility – “Phone companies” are competing with Google – you can’t do this with an old architecture. It’s really about divergence – everyone has a favorite device and they want to keep it. If “phone companies” want subscribers, they have to learn to serve the user’s favorite device and access method and there will be more and more over time.
Someone else in the audience (Japanese) – service portability is the motivator, users want services anywhere any time.
Audience comment – it’s all about the user interface – have to have very friendly user interfaces to motivate the users (One of the panelists talked about this as the iPod model – easy interfaces are the key to selling to broad markets).
Mark – have to solve the user’s real problem. Customers don’t know or care whether the pipe is circuits or packets, they care about the interface and the cost and ease of use – do IP centrex without IP Voice.
AOL – announced they were getting into the consumer VoIP business. (I didn’t hear the talk) This was clearly the biggest news of the week here, but AOL had virtually no presence at VON (Comment -- and that’s consistent with all other events I’ve attended, AOL does things its way and isn’t a big participant in industry forums)
Jim Crowe, Level3.
Not a retail company. They partner with people for the retail opportunities.
Basic message – look at the mobile market. Voice service is the razor, features are the
blades. (Comment – presumably the message
is that they give away the voice service and make money on features? I don’t know that for sure)
Need to build ecosystem of small innovative partners to deliver services.
Ring tones – a $3B market today. 10 times the market for music downloads. This will no doubt expand to the IP phone market and expand even faster.
How about this service: sound like the actor of your choice (for a fee.) (Comment – I actually discussed this kind of service with someone in Lucent/AT&T onc. The speech technology to do this has been around for a while.. We called it the “False Voice” service).
Question (Carl Ford) What’s the interconnection model? -- SIP (No surprise, Level3 was one of the key supporters of SIP in the early days.)
Question – what’s your relationship with AOL? -- Level 3 is supporting AOL, but it’s not exclusive. They deliver services, especially operator services to many including all the Cable companies.
Chris Fine – Goldman Sachs, the view from Wall street.
What does Wall street vote with it’s money – (showed a chart of stock prices) Wireless carriers and big tech (multi-industry) players have held up, but not at the level of the S&P. Telecom vendors are way back.
RBOCs – performed well for a while, but now lagging.
VoIP equipment outperforms legacy
But – the market leaders are energy (Oil) and chemicals. Tech is no longer the market leader.
What’s becoming more favorable?
What’s become less favorable?
Wireless is critical, Wireless revenue growth is $50 Billion, Data is 11, broadband is 8, traditional is down by $16 Billion.
Data capability has become the number 2 criterion for picking a wireless carrier (Voice quality was first, price was about 4th.)
Wireless also has the greatest growth in margin.
RBOCs are capturing back the lead on adding broadband subscribers.
Effect of the telecom act of 1996 – premium of P/E for RBOCS was above the rest of the market. After 1996 took a big hit, but with consolidation it has come back up near the average for the S&P.
Drivers – reduce cost and increase revenues, Opex is more important than CAPEX. The carrier business is like the transportation business. You want to buy the carrier that becomes FedEX (high value added), not the Post office (low value, high volume).
Interesting chart on equipment purchases – the total has a HUGE peak in 2000, with a drop afterwards (looked like a classic mountain profile) The projection is for “next gen” purchases to cross the line of legacy equipment purchases next year. The total is still declining though.
Acme Packet.
(They make session border controllers, probably the first company to do so)
Network Nirvana (NetVona) is about personalized services for every user. That means crossing multiple networks and bridging differences in addressing, signaling, Codecs, QoS, Security, and Business relationships
Addressing – Many problems with overlapping address spaces, IPv6 helps, but it’s not the whole answer.
Protocols – SIP won’t be the only one. H.323 will survive, but there are a lot of others too. Session protocols are only part of the picture.
Codecs – lots out there. Strategies include stripping (dropping out codings not needed), routing, or transcoding.
QoS – no infinite bandwidth, but a variety of different ways of doing QoS, and they only exist within a single network.
He gave an example from Telecom Italia. The network has SIP and 2 versions of H.323. It interconnects with enterprise networks and other providers (e.g. fastweb).
Question (from carrier) – so, I need to have peering relationships with everyone in the service delivery path to get QoS and security (yes)
Question how does opensource help or hurt? (I didn’t understand the answer, I don’t think really gave a direct answer to this question).
Lucent (Stef Von Arle) – he’s VP of services for lucent. ( This is their system integration unit not associated with any particular product)
Driver for industry has been lower cost – that’s a lousy situation., declining revenues and equipment sales. To get out of this you have to be able to generate value for the end user.
Competition isn’t over – it’s just beginning (RBOC vs Cable).
End of UNE/P drives VoIP because equal access doesn’t matter.
Personal and business users merge. I want to do personal stuff with my business devices and business stuff from home.
People don’t want 5 devices with 5 profiles with 5 passwords, etc.
They have surveys that people will pay for services.
Service proposal – you are watching a video stream, want to forward it to someone so you check their presence and either stream it to them (with a delay so they see it from the start, or send it to a mailbox.
Another service – you need a pricing decision, use presence to find the product manager who is available and handle it now, vs mail tag. (Comment – hmm, another one I can remember trying to push 5+ years back.)
Enterprises are moving from being strictly cost focused for IT and communications to value based – good news, you can sell based on what you save the business in effort.
Convergence types:
Pieces of the architecture answer.
Services have to be access and terminal independent. They don’t have to work equally well. (and of course IMS is the answer – surprise).
Have 30 trials, 7 deployments (including Sprint). Lucent doesn’t deliver the whole thing, they partner. (Comment – but, they have their critical piece that has to be there.)
The future is about multi vendor integrations – so they have LOTS of vendor relationships and expertise in integration.
Services are always on, access and device independent, presence aware.
End user demand is driving needs and that demand is changing. VoIP is not only about lowest cost.
How do you pick “best of breed”? -- It’s tough, but sometimes the customer
picks. They are the integrator for HOT
telecom (
Lucent has relationships with 200 vendors, they have the products in their lab and know how to integrate them into their architecture. That is a powerful differentiator for them.
Sonus (Hassan Achmed)
Early work was about convergence of infrastructure and about costs. It’s time now to talk about Services.
Early deployment was in network core, but interworking between networks was basically in circuit. Network interconnection is being addressed and gave us a new element – session border controller.
Next step in the evolution is packets at the edge.
The changing character of the last mile (now growing in broadband) has created a big motivation for VoIP. It didn’t make sense to replace circuit equipment in incumbent carriers in an era of declining subscribers, but with the last mile packet and IP already, it creates an opportunity to reach customers with packet voice by competitive competitors. RBOCs have to deploy broadband to deliver multimedia. Once it’s there, it doesn’t make sense to maintain a separate infrastructure for Voice.
He talked a lot about the broadband enabling of Mobility,
and the fact that the evolution of
I came in at the end of this panel and only heard parts of
the Time Warner talk and the discussion.
Time Warner
220K customers for their service. 70% take the triple play ($38.95 unlimited voice). Interesting comment on services – they keep it simple because otherwise consumer feel they are paying for things they don’t use!. (Partners – IP Unity, Cisco, Arris, Motorola (Comment I don’t see one for applications)
Lots of what enables them to grow (>11K/week) is simplicity, and having the 35,000 people out there selling, provisioning, etc.
The service is transparent to the end user – they use Analog
phones and wiring. Most users keep their
same phone number. They want 911, and
listed or unlisted numbers. They do all
the installation.
Core features for 90%, calling name, Call waiting,
Anonymous call screening, Voice Messaging,
(Interesting comment – my discussions with others on the cable business have been along the lines of why bother recreating the PSTN over cable when IP can do so much more. The Cable Labs standards are very “old world” model in some ways. The answer may be that as a business strategy these guys are going after the naďve consumer, who wants a plug compatible replacement for their PSTN that reuses the existing phones, and that’s the way you get to. I don’t know though, would you want to be in the business of people who want legacy services and equipment with a new protocol?)
Question – how do you match the feature capability of the “Over the top” SIP players (e.g. players who don’t provide the broadband access and just ride on it like VONage), and do you need to?
Answer – not their strategy, they want to offer POTS and be a more user friendly phone company, not a different form of communications. The Cox guy said they have smart engineers working on this, but the demand isn’t there yet. They are comfortable putting services out with a low initial rate, but we aren’t there yet. Cited the iPod model again. (Comment – sure, but the iPod isn’t just a CD player you don’t need CDs for, it’s a different paradigm for people.) The 3rd guy on the panel basically said the same thing. The driver is cheap POTS, maybe doing some thing a little bit more like TV caller ID, but demand isn’t there for more.
Question – what’s this like providing lifeline service. Time/Warner – 100% of their customers are primary line. Today the cable modems don’t have batter backup, but the next generation will. Cox is doing full installs with battery backup today. He made an interesting point – cordless phones don’t work with the power out, and almost all new phones bought today are cordless. (Comment -- Yup, I had a hard time last time I bought a phone getting something reasonable that wasn’t cordless, which I don’t want because it interferes with my 802.11 network). Consumers don’t really care about the installation and equipment – the only visible difference is whether the interface is inside or outside, but the average user has no clue what that interface box does anyway.
Question – what’s the real price after taxes and fees – a few dollars more ($7-$9 – not sure exactly what this included) Some charge more fees than others.
Question – how many subscribers are switching per month. Everyone is in a different stage. Comcast hasn’t launched yet, but is the biggest. Fair to assume that their base is growing an order of magnitude faster than Vonage. Cox is mainly circuit switched. Cox – circuit switch business is declining, but still important. Time/Warner is 100% VoIP.
Question – how does FMC play out (Time Warner is a sprint MVNO) Time/Warner is really just doing this as a market trial. They haven’t thought the technology through yet.
Avaya Speaker
He started with a lot of generalities on priorities of CIO and CEO. No surprises.
He gave the following growth figures: GDP 4%, IT, 5%, IP Telephony – 88%. (Comment, of course IP telephony is small and doubling something small is still small. 17 Million lines of IP telephony shipped in 2004 (worldwide, don’t ask me how they define “ports”)). The projection was for 40M a year by 2007 (which is well below the growth rate of the last 3 years). The total enterprise telephony lines are 400-500 million.
He also projected a cross over in ports, IP exceeding TDM in 2005. The interesting thing to me is that the total is pretty flat over the decade at 45-50M lines. (Comment – this doesn’t look like a revolutionary trend line to me, just normal replacement of old technology with new when the old wears out. If IP telephony is actually giving businesses a competitive edge I would expect to see a big peak as businesses junk their TDM gear early to take advantage of IP.)
Basic message from him was that communication needs to be integrated into business processes, and that by doing this the business can achieve great savings and improvement in customer services (Comment – I believe this, but of course if/when it really works I’d expect a stampede of businesses to this, and the projections I saw didn’t look like a stampede)
Question – what economic impact does VoIP
have on Avaya?
Answer – price per user doesn’t change much from circuit. (Comment – I think this is correct and it’s
interesting, because so much of the discussion is about VoIP
being cheaper. It may be cheaper to
manage for the enterprise user, but other statistics don’t suggest that – IP
networks aren’t cheap to maintain. The
real driver must be features, not cost)
Stuart Alsop (New
Perspective – The digital life has finally arrived – 1980 digital work (PCs) – 1990 Digital data (internet) 2000 Digital media (music, pictures, TV) next up Digital communications? (Comment – now this is a really weird one. Digital communications started in the 1960s! He really means Packet communications, but it’s an indication that the terms aren’t used very consistently)
Interesting comment – TiVo isn’t really digital, it’s only digital internally, the interfaces are all analog.
Impact of VoIP:
What does making a phone call mean? “Voice is a commodity?” -- Only to this industry, customers think that making phone calls is a critical service. He gave a personal perspective where talking on the phone is the second biggest part of the way he spends his time. (Comment – I believe this, and I think it’s more true for younger people, but it’s also why we as engineers get this wrong so often, since I’d bet that if you map phone use among engineers and IT people vs the general public we would come up as using phones much less.)
Digital phones make life better, that’s why they are important (email notification, unified messaging, etc. – though he gave 2008 as the year where unified messaging really happens.
Who wins, RBOC vs Vonage, etc.? Some feel that the ability to own the network and handle the QoS will allow them to relegate the others to a niche. The prevailing view is that these companies don’t understand how to serve customers and that’s the real issue – RBOCs won’t offer services in some ways.
Interesting comment on Comcast – Price for internet+TV is $50, for Internet
only is $60 – you mean that they pay you $10/month to watch TV? (Comment
– I’ll bet the $50 is a teaser rate of some sort – I personally hate these
things. Actually this was just one instance of where I think the speakers
either have their economic facts wrong or are living in a different world. Internet+TV for me is $95. Basic phone another $28.
Direct TV? -- They disabled some of the most popular TiVO features to save money. How can this make sense?
Message is that none of these companies know how to serve the customer.
Who wins, Wireless vs startup?
30% of customers switch wireless carriers every year (customers really hate their wireless carrier)
The conclusion he draws is that bypassing the wireless carriers with WiFi is attractive, but reliability is not adequate to meet the vital telephone need.)
WiFi helps the wireless carriers: WiFi won’t have 100% coverage so it’s not a real competitor, but WiFi as an alternative makes sense, even to the carrier because it frees up spectrum and expands coverage into buildings.
His conclusion is that the Cellular carriers have the opportunity to be the new “RBOCs”, providing universal service, but they have to get the WiFi integration right.
VoIP is not an add on to presence – phone calls matter more, Presence is an enhancement to phone calls, not vice versa.
VoIP is the hottest thing (as a VC investment) since Search.
Money is looking for the next Google,
and what’s attractive includes
Question – will we really have a future without regulation, how can that be? The answer was really that regulation is based on the wrong notion that the incumbents who own the network have power over the customer. Look instead at TiVo – People are religious about it and insist on buying TiVo versus generic DVRs. That kind of power over the customer is based on delivering value, not on owning the network. As a followup of course the questioner said that the cable companies can kill TiVo by using their knowledge of the network to deliver an even better service, but of course this requires the Cable companies to get it right.
I went to this because I was curious about IAX after hearing about it and visiting a vendor booth on it (with a sales rep who seemed clueless about the technical details. I won’t go through the talks here since I got to this one late. IAX is a lightweight VoIP protocol (pronounced like “icks”) that was done as part of an open source call control development (Asterisk) which is oriented towards IP PBXs, but probably more broadly applicable. It’s a 25 page spec, vs about 2,000 pages now covering SIP and all the critical extensions (Comment – I wonder how much of that 2000 pages of babble is really comparable, since unfortunately IETF RFCs are becoming nearly as verbose as ITU Specs and the IAX spec is not a formal standards document)
IAX mixes the audio and control streams, and uses binary formats for the messages. It uses a P2P model. Addressing is based on “phone numbers”. Each IAX “switch” maintains a dialing plan so you contact a switch with a phone number, and it routes it.
There are a limited number of products that currently support it, a few soft phones and a few hardware phones, plus a couple of gateways. Asterisk is the only server that implements call control in it, but Asterisk interworks with SIP. Several configurations were described: Asterisk supporting IAX phones locally, Asterisk supporting SIP phones and networked to a gateway via IAX. IAX will go through firewalls and NATs eaiser than SIP
Each speaker talked about the problem of building towards a “private” standard. Current protocol is IAX2, and the feeling is that if there is an IAX3 there may be substantial rework, making some reluctant to interwork with it. Yet another Protocol isn’t necessarily what people need.
What’s interesting is that there were many people in the audience who were using Asterisk and AIX for various things.
Question (Verizon) – What happens
when you connect AIX to SIP – he thought it would be transcoding
of the media stream (and delay.) No transcoding, just reformatting of the packets to create an
RTP stream using the information in the Aix packets. (Comment
– yes, but if this means going through a server store and forward the delay in
doing that may be substantial unless this is done by an embedded box which does
it without operating system overhead.)
Question -- How are things addressed? Phone numbers. You can use any numbering plan you want.
Question – (actually comments) – Someone took exception with the last speaker who was negative on SIP. He said that after 8 years, we still have disagreements in how to interwork in SIP. We have a big industry in session border controllers trying to straighten things out. Lots of people use Skype end to end because it works. He made an interesting observation – MGCP was master/slave, and as such it was easier to implement the dumb endpoints consistently. The problem with SIP is you have to put all the feature intelligence in the endpoints, and that’s hard. His message – the 1.8 Million Skype users are telling us there’s something wrong with the SIP model. Cisco has a 70% share of the business endpoints, and they speak a Cisco proprietary extension of SIP. Nortel has their own. We don’t have one standard for VoIP today and won’t. Comment from the panel (From the guy who seems to be the architect of Asterisk – Asterisk supports 7 protocols today, and that’s not going away. Skype has to have been an embarrassment to the VoIP community and to SIP in particular.
Another testimonial for IAX from a VoIP
provider in
The moderator’s talk
(An analyst)
He showed the same kind of charts that others have, a flat equipment market but with a major switch from Circuit to “Next Gen over 2004-2006 period. He also forecast consumer VoIP at 17.4 Million by 2008. (Comment – higher than many others presented) That’s about 29% penetration of the number of subscribers with high speed IP capability. The answer is of course also relative to 142 Million VoIP lines.
In 2004, RBOCs have almost 12M DSL subscribers. There are roughly 1 million VoIP customers, split equally between Cable and everyone else (I think Cable is defined as VoIP supplied by a cable operator.
Questions for the panel:
· Is it “The empire strikes back”, or “The phantom menace”?
· What about all the new operational processes involved in VoIP in the existing RBOC operations?
· What are the new services?
· What about Mobility, and what do you REALLY think about IMS – IMS was driven by Vendors?
Paul Perry (Verizon)
His opening
statement was a response to the moderator, what’s his first experience with VoIP – doesn’t know, probably Netmeeting. One comment he made was that he believes that
the IAD is the reason that VoIP has become
interesting – by enabling a simple and familiar user interface and re-use of
the good old telephone. (Comment – how absolutely strange. What most of us in the VoIP
community believe is that the telephone user interface is the problem we are
trying to solve and the thing the users would just as soon get rid of, while
nobody cares about the transport technology.
Here’s about the 100th speaker I’ve heard here saying users
want the plain old telephone).
He talked about VoiceWing, the consumer version of IOBI and their directory service. VoiceWing has reviewed very well. It’s easy to set up and it works. (Comment – Okay, VoiceWing is NOT simple telephony, it’s a better control interface, so what’s important about re-using your telephones in promoting VoIP?)
He spent a lot of time talking about Iobi – They feel it’s a unique differentiator. They don’t see this from other providers (Comment – it’s something Personeta and probably others with an IN interface in addition to IP could easily provide for carriers)
There are consumer and professional versions of IoBi. They have a new phone that they announced (with a screen) that goes with this and basically integrates the client.
They created IOBI themselves, based on work started in 2000 merging Bell Heads and Net Heads. (Comment – yes, I talked to Verizon about these services in the spring of 2001, of course the concept goes back a LOT farther)
They see IOBI as the unifying theme in their offerings for the future including when they get Fiber to the home.
SBC (Nancy Lambros) She is an SBC Labs person and a frequent VON speakers.
She went through a lot of issues surrounding VoIP and what the status/progress is. Her focus seemed to be more on the industry and the SBC labs. They have done some major trials of IP services and reselling IP services. The HIPPS service just started to be deployed late last year.
What they want is a service that is accessible through any device.
She went through a lot of the requirements for VoIP services, and emphasized that they are being asked to do 911 Calea, and a national emergency service on their VoIP lines. She described briefly their rollout – first a VoIP backplane, then HIPPS, which is their own implementation of a hosted IP PBX service oriented towards business, and growing towards consumer offerings. The AT&T acquisition obviously has put some new uncertainties in as AT&T brings it’s own CallVantage service.
Qwest (Daniel Jaycox) He is another regular participant in VON.
The Qwest discussion was a bit sketchy. Again going through all the things about VoIP. Customers don’t ask for VoIP, they want low cost and sometimes features. Qwest has announced its service – not clear how they are implementing it.
BellSouth (Sharolyn Farmer) She works for the Science and technology organization on services.
She described their offerings in several categories – Wholesale offerings where they provide both call termination for VoIP carriers and a VoIP to circuit translation where the circuit is delivered back to the carrier. (Comment – this is an interesting model It saves the VoIP carrier from dealing with gateways, but leaves them the responsibility of buying circuit connectivity to their peers.) For businesses they have an IP Centrex type offering (Based on Siemens and Broadsoft I believe). She talked only a little about the consumer offering.
Again much of the focus of the presentation was on requirements on the service from regulators and others.
There was a fairly lively question session asking about plans for IP TV (everyone said they were, though I don’t think the Qwest and BellSouth answers were really specific to IPTV but to TV over DSL/Fiber in general). Also questions about verifying some of those projections of the number of lines (not a lot of new information).
(Comment – After listening to the service provider sessions from cable and from the incumbents, I am struck by the fact that I still don’t understand the target customer for the offerings. What people seem to be saying is that something that reuses the PSTN wiring and phones in the premesis is essential, and that their target customers want cheap service with a simple interface. That’s fine and I’m sure that customer exists, but it’s not consistent with where the mobile world is going (more complex phones, more capable interfaces, and a new phone every year), and I have to wonder whether this whole set of players is chasing the wrong audience. What’s going to happen when more people get exposed to enhanced user interface services like IOBI, IP Centrex, SKYPE. I really believe that the meaning of telephony to people might change radically.
Nortel (Same guy who always gives the Nortel VON talks)
He went through a reality check on various technologies. This was on a curve of visibility, which has a big peak of “hype”, then a fall off and then a climb towards success:
VoIP has gone past the valley of despair and is climbing towards success – it’s real and it’s going to make it.
“Voice is an IT Application” – It’s still climbing towards the Hype peak. Problem is we aren’t prepared to deliver IT applications to meet the reliability assumptions for Voice.
A business PC costs $30/Day , Total IT spending per employee is about $70/day. In comparison, Office phone is $1/day, Office LAN is $1/day, Mobile phone is $2/day. (IT has a way to go in cost. ) (Comment – this really depends on where you cover LAN administration and support and presumably administration of the phone system as well I’ve seen widely varying estimates of these costs, and the LAN cost here is definitely on the low end)
Voice is available as a medium for IT apps, but we aren’t yet to the point where IT can deliver Voice with the efficiency of the current providers.
Presence – peak of Hype. “Presence is the most overhyped thing today”. (Comment – another shocker. This speaker’s earlier talks were ALL about presence and how it was the real killer technology in VON).
His point really was that
relatively few of the people you deal with will trust you with their
presence. “Value of Presence is
inversely proportional to the preferred length of the desired communications”.
(Idea is that if you just want to trade a quick message with someone it’s
likely to work, it won’t work as a way to schedule a half hour with
someone. It’s also inversely proportional to the
business of the person being contacted.
(Busy people are seldom available).
(Comment Yes, I can see this as a basic flaw.
One of the problems with most “availability” schemes is that they rely
on the user to manually update the information, and few do a good job at
it. To me the whole IM thing seems odd,
since I first used email in an environment where email was equivalent to IM –
your email was always on and you received instant notification that you got
mail (networked workstations or timesharing environments) The Client/server
world, in which clients often worked off line and even when on line had no
active notification from the server of an new message created a problem with
email that IM solved)
Location -- This is still very low on the rise towards hype – under appreciated. Location is a useful screen for whether or not communication is relevant to someone. He presented something from a company he saw using WiFi to locate people in the same area to initiate a face to face meeting.
What has value to people: Simplifying communication. Enabling people to manage multiple personalities, minimizing the number of devices people have to carry and mobility. (He commented on how many people carry multiple cell phones, not because of coverage but because one is for work and one is personal. That’s silly. We should be able to manage two personalities on one device (Comment – now that’s what I heard from him before, and a theme I’ve been on for years. The network can make one phone handle multiple personalities each with it’s own screen rules. I’ve given talks at VON, IN/IP World Forum, ICIN, and I think Supercom on this!)
He gave as an example the problem he has where he gives out his phone number to people, and then has hundreds of people who carry his personal phone number and he can no longer leave the phone open 24 hours a day for his family. Phone networks make this worse by giving out his personal ID (phone number) every time he uses his phone. What he really wants to be able to do is give each person a token that they can use to contact him, but the token won’t enable the holder to find out anything about you, and the availability status can be different for every token. (Comment – yes, this is quite doable with network technology and it doesn’t even require VoIP!)
IETF update (Live via video with Henry Sinnrich)
It took a few minutes to set up the
link (
John Peterson – He talked about the ECRIT work group (Emergency Context Resolution – this is aimed at supporting 911 type services). It’s the first meeting and an international effort with requirements from all over the globe (Comments – this looks to me like a recipe for a very complex solution)
Dean Willits (Sipping) (He used to be with Dynamicsoft, not sure where he is now) A lot of his update was also on emergency response work (pre-emption). They also worked on identity management. They met with the OMA (organization that specifies mobile standards for Push to Talk, Presence, etc. among others)
Robert Spots (Presence) He reported about some of the Presence activities – device and mode selectivity for availability
(And in support of the fact that “Voice is an IT Application” isn’t ready for prime time yet – some piece of software on the PC displaying the video decided it was time to apply an update and splatted a popup all over the video which the session chair had to figure out how to eliminate and then reset the conference).
Johnathan Rosenberg (Now Cisco) He talked about a new effort on NAT traversal (AIX) IETF is working this as a high priority. It’s tied into the transition to IPv6. (Comment – this is one of the things that makes me suspect that SIP is getting bogged down. NATS and IPv6 are old problems and we haven’t solved them yet. This was going on when SIP started!)
ENUM – Most of the problems and
limits are politics, not technical. We
are finally going to start trialing ENUM/north American numbering plan interworking. Jeff
(presumably Pulver) has a block of “500” numbers for
testing this. Sweeden
and
XCON (centralized conferencing) -- goal is to develop protocols between participants and the central focus in the conference. (One of my projects a couple of years ago involved conferencing in SIP, and I was amazed to discover that while SIP started out from a group that was working on conferencing, there were no real standards in architecture or protocols to support it, just a few sketchy drafts). They reported that much progress has been made basically in the area of control of a conference through a central point.
The Exhibit hall for VON was huge in comparison with most Pulver events I’ve attended. Not quite at the scale of Supercom but getting close. All the major telecom vendors were there (Lucent, Nortel, Siemens, Alcatel, etc.) as were all the major IP equipment vendors, plus a large number of small companies selling application servers, session border controllers, PBXs, and other equipment. There were displays from several carriers presenting their wholesale services or integration services. Microsoft showed their enterprise solutions and I think occasionally some IP TV. The most noteable absence was of course Sun, always a regular at these meetings and no surprise not to see them after learning that Sun had just laid off the JAIN team.
As I noted above in the summary, I didn’t see anything really revolutionary or surprising, but I did see a lot more maturity in what was being demonstrated. Lots of session border controllers (I remember 2-3 years back ACME Packet had a small booth and spent most of it’s time explaining why you needed such a box, now people seek those solutions). All the application server companies were there (Bay Packets, Personeta, SS8, Telcordia, etc.) I saw several demonstrations/presentations on converged desktop services, mostly all IP, and a few demonstrating WiFi and dual mode WiFi/Mobile communication. Lucent had an interesting demo of a unified profile technology which mapped data from a large number of sources into a single editable user profile. The technology is part of their large stand alone HLR offering for mobile networks (the basis I think of their HSS in the IMS architecture).
A side note here is that another indication of the health of the industry is that there were more and better handouts from vendors this year than in recent years, and lots of vendors were giving away iPODs or other MP3 devices in raffles. Several booths had shows of various sorts to bring people in. Also no lack of vendor sponsored parties and receptions.
It was interesting seeing VON become more like a big trade show, but at the same time I think it lost something. The scheduling of VON on top of the IETF meeting was clearly evident in that a lot of the technical gurus who attended past VONs were elsewhere. Just unfortunately scheduling I think, but the conference had a less technical and more marketing feel to it than the other VONs I have attended.
The party was one of the nicer ones I’ve seen at a VON. Partly it was nice to be within walking
distance rather than dealing with busses, but it was also fun to hear a band
I’ve heard and enjoyed for years (Heart) still doing their classic stuff, and
recognizing just about everything they played.