Hundreds of Columns Sorted by Topic
The Future IC/UC Net Will Be Federated
by Seamus Hourihan, Vice President, Marketing & Product Management
11/5/2007
VoIP is dead, over and "out!" IP interactive communications (IC) or unified communications (UC) is "in!" IC or UC is much more than voice. Even the acronyms suggest so - "I see" and "You see." It's more than just video. It's insight gained from the availability of your family's, friend's or colleague's presence information and the ability to collaborate for business, learning or pleasure with the simultaneous use of data applications. Sounds like network nirvana!
But how will this really work? IC and UC services and applications will only become valuable when we can use them to reach anyone, anywhere, anytime. To paraphrase Metcalfe's Law: the usefulness, or utility, of interactive communication equals the square of the number of users. Consequently, IC/UC must span multiple IP networks -- business, residential and mobile; wireline, wireless and cable. Today's consumers and businesses will be satisfied with--and pay money for--nothing less.
Our only options for delivering this network nirvana are the Internet or the Federnet - a federation of managed IP networks. There are seven fundamental IP network precepts driving the emergence of the Federnet for IC/UC.
1. In IP, We Trust No One
Service providers and enterprises delivering interactive, unified communications need to protect their service and application infrastructure against DoS and DDoS attacks. These attacks might be malicious or non-malicious. Malicious attacks are pre-meditated by hackers, the black hats. For most of them, their goal is no longer just the personal gratification that they could do it, but to benefit financially. For example, shorting the stock of their service provider target. Terrorist cyber attacks are another type of malicious attack. Their goal is destroying the infrastructure used by a democratic, capitalistic society. Non-malicious attacks might result from recently upgraded endpoints that start registering every single second or an overload to an American Idol voting telephone number. Regardless of the type of attack, a successful attack will conclude in a variety of losses including SLA promises, customers, a reputation, a complete business, and perhaps even human lives.
Subscribers are not capable of protecting themselves from everything. They need and want to trust their service provider. They need to ensure that their communications are actually established with their intended callee whether the callee is personally known to them or not. They don't want their IP PBXs, voice servers or PCs to be maliciously attacked or personally besieged or harassed by calls from any anonymous user on the Internet. They demand guarantees relative to the identity of callers and the ability to block unknown callers. For some communications, they will demand privacy or confidentiality in their communications.
2. Addresses will Perpetually be a Collection of Heterogeneous Schemes
Despite what the Internet purist want and want you to believe, the Internet is not open end-to-end. For security reasons data firewall/NAT devices are everywhere protecting large web sites, enterprise networks and even PCs at home. Developed only for IP data applications, they only allow data into a network if it has been requested from someone or something inside the protected network. But voice is different. I need to be able to call you from outside your network. I can't, and you can't call me -- a proverbial Catch 22.
Additionally, more and more enterprises are using managed network-based MPLS VPNs (RFC 2547), private networks in simpler words, for secure managed network connections between locations. But these network islands do not currently have secure bridges to those of their service provider or other enterprises capable of supporting IP interactive communications.
Many service provider VoIP trunking and IP PSTN termination network islands have been built using private address spaces. So interconnection with others also using overlapping private address spaces is impossible. In developing countries like China that are starved for IP address space, "public" networks are being built in a way to conserve what little IPv4 address space they have using NAT devices everywhere.
The promise of IPv6, while eliminating our thirst for address space, will only increase the address space babel. It will be years, maybe decades, maybe never before everything uses an IPv6 address. Even then firewalls/NAT devices will not be eliminated in light of their important security role.
The problem of address space mediation doesn't only exist at the IP layer. Another type of address space problem relates to telephone number incompatibilities. SIP URIs like shourihan@acmepacket.com are the ultimate solution, but we aren't even close to that being a reality any time soon. Believe it or not, even in the world of VoIP there is a requirement to add or strip number -- add a "1", strip the "011" - before passing them on to another VoIP network that has its own, often myopic view of the world.
3. SIP is Not the Only nor a Single Signaling Protocol
The reality today is that we live in a multiprotocol signaling world. Since all next-generation services architectures including 3GPP IMS, ETSI TISPAN, ATIS, the Multi-Service Forum, and PacketCable have embraced SIP, SIP will ultimately be THE protocol for new wireline, wireless and cable services. However, SIP will NOT be the only protocol for sometime. Today H.323 exists in many different types of networks. These networks include international PSTN trunking and termination networks. Most new IP PBX offerings being deployed today are still using H.323. New H.323 voice and video services are being deployed by service providers in countries as different as Italy and China. MGCP is also being deployed for new voice services by several US ILECs. PacketCable today uses NCS, an MGCP derivative. Even H.248 aka Megaco is also being deployed for new services.
It would be bad enough if we just had these high-level differences in signaling protocol "languages." But we also have different "dialects" of each language. H.323 is notorious for its multiple versions (1 through 4), annex, service and configuration options, and incompatible vendor implementations such as Cisco, Clarent, VocalTec and others. In the world of SIP, while most implementations are compliant with the current RFC 3261 standard, there are still products that adhere only to the old standard RFC 2543. Even RFC 3261 is not "single" standard. It offers multiple options for transport protocols -- UDP, TCP and SCTP; multiple options for signaling security -- none, TLS, MTLS, DTLS, IPSec; multiple options for media security - none, SRTP, IPSec. The transport of DTMF digits may be carried in-band within the media flow using RFC 2833 or out-of-band within the SIP signaling messages. Another very obscure level of signaling protocol incompatibility relates to response or cause codes. Within some deployments there are requirements to translate SIP cause codes such as "404 - Server Not Found" to "503 -- Service Unavailable" before passing them into another network to precipitate the correct "network busy signal" vs. incorrect "dead air" service behavior. While all these different options provide tremendous choice and flexibility, they also guarantee incompatibility and lack of interoperability between networks.
4. Codecs will Never Converge to a Couple - One for Audio, One for Video
While the world will ultimately standardize on SIP for new deployments, codecs will never converge to only a couple -- one for voice and one for video. Codecs are the algorithms for digitizing analog voice and video so they can be transported in IP packets. There is an even greater existing selection of standards to choose from today with more new codecs being invented. Within the traditional wireline world, voice codecs are typically the ITU G.7xx series and video codecs are the ITU H series. In addition to the type of codec, there are also different options for frame sizes -- 10, 20 or 30 ms, for example, which add to the complexity. In mobile wireless, an entirely different set of voice and video codecs are used that feature adaptive dynamic support for multiple bit rates to optimize bandwidth utilization over the radio access network. Lastly, new codecs are being developed to further improve quality while minimizing bandwidth utilization. Some of these new codecs include iLBC, iSAC, Speex and Microsoft's new OCS codecs. Every call between endpoints not supporting at least one common codec, wireless and wireline phones for example, will require transcoding.
5. Infinite Bandwidth, QoS and Signaling Resources End-to-End Will be a Myth Forever
Today's IP networks are constructed using a selection of different QoS mechanisms, networks links with different bandwidth and different-sized call, application and media servers. Voice can't tolerate excessive delay or jitter, so all these resources must have enough capacity and performance to support a new call. The choices in QoS mechanisms include IEEE 802.1 p & q, ToS, DiffServ and MPLS. Unfortunately, these deployed mechanisms don't extend beyond the domain of single IP network domain. They don't operate on the transit links between providers or on the access links (T1/E1, DSL, frame, etc.) connecting enterprise or residential locations to the service provider backbone. These links, regardless of their size or bandwidth throughput, do have a finite capacity. If a link is at capacity, and more traffic - just one more call -- is placed on the link, the quality of all active calls will deteriorate, not just that last call. Similarly, the servers delivering interactive communication services - SIP proxies, H.323 gatekeepers, MGCP call agents, NCS call management servers (CMS), 3G IMS CSCFs, softswitches, application servers and media servers - also have finite capacities for call handling. Consequently, they also face the same potential overload issues.
6. Some Sessions are more Valuable than Others
On the Internet all packets are equal and are delivered on a best efforts basis. Any packet has the same probability of getting dropped or delayed as any other. In the voice world, we need the ability to provide special handling for particular calls or sessions. In the presence of voice server overload, service providers need the ability to gracefully reject the low value American Idol televoting call to support the high-value enterprise video conference spanning multiple locations. Emergency calls, E9-1-1 calls in the US, also need special handling in terms of prioritization and possibly pre-emption if service resources are oversubscribed. Lawful intercept capabilities supporting government regulations like CALEA in the US must always work to support law enforcement in their pursuit of pedophiles, drug cartels and terrorists.
7. Business Models will never be Homogeneous
Interactive communication services are potentially so functionally rich that business models will never be homogeneous. It's not just voice. It's video and multimedia sessions such as truly interactive, collaborative white boarding. There are person-person communications and multi-party conference sessions. Distance learning will push "conferencing" to a new level of interactivity and control. Sessions will be available with or without QoS support. QoS can be invoked on a per session basis or even mid-session via a "turbo" button. And don't forget the highly-profitable options like directory assistance. Service providers today compete for this business on the basis of cumbersome, difficult-to-use automation vs. the ability to talk to a real human. Consequently, any subscriber "on-net" service (where on-net means the network of a single provider or federation of providers) will likely have some combination flat rate - $ per month per person or residence, and variable rate - $ per session, per minute or maybe even per packet. There might even be a model where you buy interactive communications services from an ITSP and QoS from your facilities-based transport provider.
The business relationship between providers for sessions that need "off-net" support is still a big unanswered question. How will revenues or costs will be allocated and traffic exchanged. Will it work like the Internet where big backbone providers "peer" and exchange equal volumes of traffic and smaller ISPs pay for backbone connections? If transcoding is required, which service provider pays for that? How will calls to the PSTN that are still locked into the fundamental money per minute model of that world be handled?
IC and UC, the Future - the Federnet
Our network nirvana can only be realized by connecting IP networks together in a way that enables the end-to-end delivery of trusted, first class IC/UC to anyone, anywhere, at anytime. Each of these seven precepts is driving the need for more intelligence, not less, into our IP networks at their borders. The best-effort, insecure Internet will never hack it. IC the future is the Federnet. UC?
About the Author
Seamus Hourihan, who first coined the term "session border control," has driven industry recognition of the critical importance of the new product category created by Acme Packet. He is a prolific speaker on the conference and tradeshow circuit and is the author of several networking guidebooks and numerous articles. Recently, Seamus was recognized by Internet Telephony magazine as one of the "Top 100 Voices of IP Communications." Seamus brings over 25 years of experience to Acme Packet in executive management, marketing, product management and business development roles at voice over IP, IP networking, web infrastructure and computer companies. Seamus was vice president of marketing for internetworking leader Wellfleet and, after its merger with Synoptics, Bay Networks (now Nortel) for nearly seven years. During his tenure, annual revenue grew from $10 million to $2 billion and Inc. recognized Wellfleet as "America's Fastest Growing Company" for two consecutive years. More recently, Seamus was vice president of marketing for Pingtel, widely recognized as a leader in SIP products and technology. He also held management positions at Data General, MASSCOMP and Bright Tiger and has operated his own consulting company. Seamus holds an AB degree from Dartmouth College and an MBA from Babson College.
About Acme Packet
Acme Packet (NASDAQ: APKT), the leader in session border control solutions, enables the delivery of trusted, first class interactive communications--voice, video and multimedia sessions--across IP network borders. Our Net-Net family of session border controllers supports multiple applications in service provider, large enterprise and contact center networks--from VoIP trunking to hosted enterprise and residential services to fixed-mobile convergence. They satisfy critical security, service assurance and regulatory requirements in wireline, cable and wireless networks; and support multiple protocols--SIP, H.323, MGCP/NCS and H.248--and multiple border points--interconnect, access and data center. Our products have been selected by over 420 service providers in 81 countries, including 23 of the top 25, and 76 of the top 100 service providers in the world.
Send us your response to this article.
Learn How to Get Your Column Published on this Site
