Authors: Les Cottrell. Created: April 19,2000
Next big wave on the Internet is voice/video/data convergence. Cisco says end of this year, others say 4-5 years. Drivers are lower cost of administration (costs of moves, adds, changes) and new applications (e.g. voice & data collaborations). Gartner says > 12% of installed PBX systems are now over 10 years old. Over 250K VoIP ports shipped since July 1999, over 50K shipped in previous year. Cisco says it is on 2nd generation, others are on 1st generation. Cisco has > 50% of VoIP market today. PBX price per port dropped $720 to $650 1995-2001, while 10/100 Enet ports drop from 1000 to $150/port.
AVVID (Architecture Voice, Video & Intrgrated Data) is a long term strategy (architecture) keep adding layers. Clients are PC (soft phone), IP phones, video, GSM wireless phone to PBX. Applications include call managers, telephony application servers, directory servers, content servers. The infrastructure includes switches, routers, gateways, QoS, security, voice, video, multicast, VPN, load balancing, caching, address management. Cisco is able to cover all levels. At the same time they are standards based and it is distributed (vs centralized PBX) so can by products from other vendors. Also Cisco claim scalability up to 10K users per cluster and they have 5 Call managers/cluster today.
They can use a single LDAP directory for call by name (today the use Sun/Netscape Alliance), for PBX information. If centralized database (e.g. SID) can offer an LDAP interface then Cisco can use it.
The Cisco campus has 2500 IP phones today, they are going to 8000.
Can do automatic rollover to PSTN when all IP capacity is in use. Have backup call manager (have lose 3 call managers before lose dial tone. High availability from Telcos does not count outages of < 10 seconds. Introducing powering phone from Enet (cost 48v DC, $10/phone, requires either new cards in the Catalyst or a new patch panel) not standardized yet, going through IETF (today's phones will work with whichever of two proposed standards wins), also require SNMP manageable UPS (and MG backup for major critical centers) on core routers & switches and large closet switches. They have location information using CDP (Cisco Discovery Protocol) to discover port on closet (i.e. building area) that a phone is connected to. In new Catalysts will be able to run supervisor port on different IOS levels so can test, upgrade more easily. They have a new end to end system test lab to do better quality testing of inter-working on new code releases. The higher availability requirement for the voice network will enable the data network to have higher availability.
The improvements in communications that come with the easier to use mixed applications of voice, video and data are what will make VoIP valuable. Can assist help desk/consultant taking over the screen of a user, sharing an Excel spreadsheet or a Powerpoint presentation or a Word document (e.g. for joint authoring), or whiteboard. It will leverage what we have in NetMeeting today, in particular single button to setup/join a meeting, i.e. they are improving the ease of use of existing technology. More GUI interfaces (drag & drop) to add people to conferences, to forward calls, callback without *69, can send & display text on phone, configuring voicemail. IP phone has 10/100 built in hub. Cost of phone is $495 for full-duplex speaker phone with large screen, they will also have a $295 & $145 but without full duplex speakers. They are pushing the standards so one many vendors will make phones and will be able to buy such phones at say Frys. Tandy and Samsung have signed up to produce such standard phones. In the long run John Chambers has publicly stated he wants to get out of making phones. Catalysts will have 24 analog port blades for analog FAXes, modems etc.
Less than 72% of calls get through so voicemail is very important. Management expects to leverage investment in data infrastructure. Two IP telephony solutions: VoIP (transport); voice in IP (LAN telephony/LAN PBX). Both are centered/based on H.323.
VoIP 1st step is to connect PBX to IP network to eliminate dependence on a carrier. Voice in IP removes PBX (with all its state information) and is more radical. Then connect IP phones to switches, and allow interfaces to PSTN but use IP as the driver. Can engineer the same level of reliability in data network as can in phone network. Can daisy chain PC off phone (switched on new phones).
IP phone type is sensed and the configuration is downloaded with an address from DNS/DHCP. Thus the phone can be moved about. Can use the private IP address space (i.e. 10.x.x.x) to save addresses. They cut CDRs (one CDR repository per cluster) and will use an SQL database to store/retrieve. The media convergence server is NT based with RAID disk mirroring, redundancy, remote maintenance, Cisco Smartnet, Pentium III 500MHz, 256 MB ECC SDRAM.
The GSM phone gateway is 802.11 but uses a different access point. They have an FXS product that provides an analog phone interface to IP gateway.
Cisco uOne Enterprise Edition target availability 2H CY2000. Today worldwide there are about 900M voicemails/day, compared to 7.2B emails/day. 30% of long distance call costs is on FAX. In the past there were 3 vertical approaches: voicemail, email, FAX. This requires separate administration, separate infrastructure, separate directories, separate storage, separate training, separate organizations and budgets. Users have to go multiple administrators, increased operational costs. Unified messaging merges these infrastructures, administration, directory, message store. The employee can get "messages" from anywhere, i.e. from Internet and phone. In the past Unified Messaging boxes were very expensive, were proprietary and did not scale well. The new generation is to apply IP, use open protocols (IMAP/SMTP, LDAP, and VPIM (Voice Processing for Internet Messaging and SMTP/MIME type for email), AMIS-A (used for interfacing voicemail)), leverage existing resources (directories and message stores), and to make it more distributed (e.g. break the UM application from the message store and from the directory). The distribution also provides resiliency, e.g. can replicate applications server. Also the telephone user interface is written in XML and so is more open and easier to modify or develop for. Also the hardware components is largely off the shelf (e.g. PCs) as opposed to proprietary boxes from Octel, Lucent etc. The Cisco call manager is the MCS-7830 (based on a Compaq PC & includes 100 voice mail boxes and 4 simulatneous sessions). There is a choice of back-end message stores and directories from Microsoft (Back office), Lotus and Sun/Netscape Alliance, mail clients from Outlook and Netscape. This will all tie in with the 3G cell phones to tie into stock quotes, calendars, email, voicemail etc. They have 3 levels, small office (one MCS box), enterprise (separate message store) and corporate (multiple boxes MCS, GateServers, message store, directories). The GateServers are stateless and so can fail over. They also bought Calista that allows going into a PBX and tapping off to access the voicemail. Nortel is participating in the interworking with VPIM, but not clear whether if they can do it with the Meridian mail. Cisco is a huge Octel/Definity campus. New employees get IP phones. Will install the new voicemail infrastructure on San Jose campus this summer. They have beta tests with 3 big companies and it is in use at Selsius in Texas.
They are working on integrating text to speech. They have a personal assistant effort that will include natural speech recognition technology from Nuance. This will be similar to "Wildfire". Look at http://www.webley.com.
Overall it was a very useful visit. There was quite a lot of redundancy in the presentations. We still have concerns over the reliability (e.g. how to re-engineer the data network, how to do upgrades, life-cycle of instruments) which we need further follow up on. We would like to pursue how to understand the technology and explore a pilot to understand the extra capabilities, the costs, the migration and integration with the legacy equipment (PBX, voicemail), how do things scale, what are the power requirements (to phonesets and UPS), get experience on reliability, and all the nasty details. A real pilot could probably not start until the PBX is upgraded to provide an ISDN/PRI interface to the Cisco gateway. We would need a much deeper level of Cisco support (installation, maintenance, trouble-shooting) to provide the reliability needed. Cisco is going full bore on VoIP and appears to lead the field in some areas. However in others they lag, for example their unified messaging solution is not released yet, while Lucent is shipping its product. Despite this Cisco appears to have a good architecture, however, implementing it etc. may depend on how well they can harness the skills of the many companies they have bought up recently.
They will be getting rid of their Lucent PBX at the San Jose campus this fall. We are still very early in the timeline for convergence of the various technologies. One interesting anecdote: The largest installation in the Bay Area is at Treasure Island. They had 3 installation engineers on site. It took about a day to configure the gateways, call manager, but then took 3 days to print out the detail paper stick on labels for the instruments. The new phone has no physical labels the labels are on the screen.