November 1998

IP-Based PBXs -- Facts And Fictions

BY MIKE KATZ

Much has been said over the past year regarding the impending demise of the PBX. This is due in part to the arrival of the next generation of Internet telephony products -- IP PBXs. This article is the first installment in a two-part series that examines the facts and fictions surrounding IP-based PBXs.

With the arrival of the next generation of Communications Servers and LAN-based PBX products, a new set of challengers is vying for future market dominance, and they're going up against a few formidable opponents. Few analysts believe that the large entrenched PBX companies will simply roll over and allow their market to be eroded right before their eyes. However, an IP-based PBX (or IP PBX) has many important benefits for customers. It should provide end customers with the ability to scale their phone systems as easily as they scale their corporate intranet, all the while delivering voice quality and applications support beyond what a traditional PBX solution can provide, and at price points that make sense to mainstream corporate America.

DEFINING THE IP PBX
Let us begin with a definition of an IP PBX. Basically, it is any IP telephony-based device that can be connected to an internal LAN or corporate intranet/VPN (virtual private network) that provides basic PBX features over the network. For example, the PBX feature set might include call answering, auto attendant, hold, conference, transfer, and perhaps some form of voice messaging. It may or may not deliver the voice component of the call over the network to proprietary Ethernet phones or IP appliances, however it must in all cases use IP as the protocol to transmit voice. A wide range of products, mostly from new market entrants, falls into this category (Table 1).

There are two major classes of IP PBX products that have the potential to deliver on the promises mentioned above. Centralized hybrid solutions that support analog phones (also known as 2500 sets) are one type, and pure distributed solutions that support endpoint-to-endpoint communications (i.e., Ethernet phones that use a proprietary protocol to connect and send speech back and forth) are another.

Within the centralized hybrid solutions, desktop delivery of voice is handled traditionally with IP telephony methods employed for integration of multiple units. For example, they typically employ two-wire phone lines with 2500 sets for desktop voice support while the IP telephony mechanisms enable tie trunk expansion among multiple boxes. At the other extreme, IP PBXs use a central controller software for call setup and connection to the PSTN, but carry the voice part of the call totally on the in-house LAN, delivering the call to the desktop via a proprietary Ethernet phone. These two delivery mechanisms really define the two solutions.

CHARACTERIZING THE PAST
To better understand the requirements and desired operational characteristics of this next generation of PBXs, it is a good idea to look at what we have today. Today, we get dial tone when a phone goes off hook. This is the first (and usually the best) indicator that the PBX is working and a call can be completed. Dial tone is supplied 99.97 percent of the time, even when the power is off in the building, (that is, of course, if you believe the statistics provided from Bellcore).

Next, calls are dialed through the PBX to the Central Office and then routed on to the called party. Calls go through -- meaning the person dialing the call hears ringing, a busy signal, or some other call progress tone -- 97 percent of the time on average. This adds up to a very high reliability score for call placement and delivery - no matter whether the call is internal or external to a corporation.

When the called party answers, the quality of the speech you hear is at least in the range of 3.6 to 3.8 on the Mean Opinion Scale (MOS) score. This is referred to in the industry as "toll" quality voice. There are no perceptible delays or echo characteristics in the speech and the pace of the conversation is natural and regular. So, for an IP PBX solution to be successful, it must perform at least up to the standards set by the existing PSTN network. Other characteristics and capabilities such as integration with applications that can control the phone and total cost of ownership are important, but not necessarily more so than the basics mentioned above.

KEYS TO SUCCESS
So, it appears that our two alternatives must meet and exceed the operational characteristics of the existing solution to enable market acceptance. The focus here is on evaluation of reliability and fault tolerance, not on open applications capabilities or cost, which do factor into the success equation.

Centralized Hybrid Approach
For the centralized hybrid approach, the delivery of dial tone, call progress information, and applications integration is at the same level as in the traditional PBX environment. The 2500 set phones are connected over two-wire copper in a star configuration back to the centralized hybrid. Linking multiple distributed boxes together provides scalability and growth to the solution so that larger enterprises can be covered. Should the intranet suffer a partial outage, then only the PBX stations the are connected via that segment lose service -- not the entire PBX. There are some solutions that can also automatically fall back to PSTN connectivity during such an event.

By limiting the requirements for intranet connectivity to only those calls that go between boxes means that a centralized hybrid approach has reasonable reliability. Most centralized hybrid solutions can be "battery backed up," therefore they can withstand building power failures. The limitations are similar to existing PBX solutions in that adding phones means adding more cards, wires, and the like, and that moves, adds, and changes require some amount of work. But, with centralized hybrids, mere mortals can now perform this work.

Distributed Solutions Approach
For purely distributed solutions, running calls across the intranet will work just fine if the following is and stays true: The LAN connecting all of the devices is up and running and is designed to avoid performance bottlenecks. The distributed approach offers the benefit of location independence. If the intranet passes by a location within the building, then the phone system does as well. This is often referred to as a single cable plant or single wire to the desktop. Calls are set up via a centralized call control engine that has the knowledge of which IP address is assigned to which phone. So, when you call extension 202 from 201, the call control engine is telling 201 what IP address 202 is at. This is an elegant and scalable solution that is also used by the centralized hybrid approach. It simplifies the end stations' intelligence requirements and makes moving and adding phones simple.

However, there are limitations. It appears that intranet outages can be catastrophic to the intranet-connected telephone devices. This is especially true if the common call control engine is not reachable -- this creates a problem of how to get the end device to locate another telephony device or to route a call to the PSTN if the network between them is not functioning. Random intranet traffic coupled with time of day usage can congest an intranet, so much so that any IP-based voice call will have its voice quality affected perhaps enough to wreck the call. Even if it's going from one phone to another on the next floor within the building. Can you remember when the last LAN outage occurred in your business? Can you remember when the last phone system outage occurred? Can you remember how long one was down versus the other? Proper LAN architecture and design matters.

ARCHITECTURE OF THE LAN
In this article, we have been using Intranet and LAN architectures interchangeably, but there is a difference. An intranet may -- and most likely does -- extend out of your building to multiple corporate sites. The LAN, by its nature, is likely to be physically limited to your building. As indicated above, to get either the centralized hybrid version or the purely distributed version of an IP PBX to work requires a "managed network" for Internet telephony traffic. In past issues, I've defined a managed network as one that has predictable performance measured in terms of Quality of Service or Quality of Transmission guarantees. These guarantees are made usually around latency (that is, the time it takes to pass an IP packet from one point in the network to another) and the ability of the network to move large amounts of packet traffic without dropping or losing packets.

So, does corporate America possess the necessary LAN architecture to make either solution a success? Not yet. A critical piece of the LAN architecture is deciding whether the distributed telephone devices or centralized hybrids are deployed on a shared or switched Ethernet LAN. A shared Ethernet hub provides a non-predictable bandwidth medium to run IP telephony over. So, if your LAN segment has some heavy Internet browser or database lookup traffic, and is served by a shared hub, then it's likely that the available bandwidth for IP telephony calls and hence the end user voice quality will be affected during a call.

Some statistics are available on shared Ethernet hub deployment versus switched, with most sources indicating that 80 percent of corporate LANs are run on shared hubs versus 20 percent switched. This means that the 80 percent of typical corporate customers will have to upgrade either all or part of their network to deploy the purely distributed approach. The centralized hybrid approach needs only to be connected to a segment that can reach all other like devices. The volume of LAN traffic on any given segment is less of an impediment to the centralized hybrid approach due to the smaller volume requirements for "off-box" calls. Thus, there is less of a need to upgrade the in-house LAN.

CONCLUSION
Is the IP PBX headed for success? Most certainly. It's really an issue of when and how it will be successfully deployable in the corporate market. With the criteria stated above and with currently deployed LAN architectures, the near-term picture favors the centralized hybrid approach. However, as we all know, corporate bandwidth needs continue to grow as more and more intranet/Internet applications are deployed. These applications will drive up the need for more bandwidth over the next 2 to 3 years and will demand a corporate infrastructure change. This change will bring about the deployment of newer Ethernet switches with, among other advancements, embedded prioritization schemes for voice data packets. This changeover will help to enable the purely distributed approach.

In next month's installment, we will expand on the potential of the IP PBX and discuss the necessary telephone terminal or IP-based telephony appliances that will play a big role in this new and exciting future.

Mike Katz is vice president of marketing and business development at NetPhone, Inc. Headquartered in Marlborough, Massachusetts, NetPhone is a leading provider of computer telephony solutions for small business environments. For more information, contact the author at [email protected].