This is a huge subject so let's follow Einstein's example and "Make everything as simple as possible, but not simpler."

The starting point is the status quo of the infrastructure, and the first decision is equally obvious: which employees will benefit from the additional functionality, and can the remainder continue as before? In most companies there will be a clear division, so a mix-and-match solution would be needed.

IP PBXs can interoperate with PBXs; i.e., TDM and IP phones call each other, and in a mix-and-match solution, the PBX acts as a gateway to the PSTN for the IP component. An elegant alternative is to deploy a hybrid platform that does both IP and TDM. It's elegant because you can retire the PBX and keep the phones, which represent the biggest part of the TDM investment. However, the hybrid platform would normally have to come from the same vendor since the phones are proprietary devices.

This is a huge subject so let's follow Einstein's example and "Make everything as simple as possible, but not simpler."

The starting point is the status quo of the infrastructure, and the first decision is equally obvious: which employees will benefit from the additional functionality, and can the remainder continue as before? In most companies there will be a clear division, so a mix-and-match solution would be needed.

IP PBXs can interoperate with PBXs; i.e., TDM and IP phones call each other, and in a mix-and-match solution, the PBX acts as a gateway to the PSTN for the IP component. An elegant alternative is to deploy a hybrid platform that does both IP and TDM. It's elegant because you can retire the PBX and keep the phones, which represent the biggest part of the TDM investment. However, the hybrid platform would normally have to come from the same vendor since the phones are proprietary devices.

Ideally, the IP phones would work with the legacy systems of different vendors. This is a tall order, but it's one that Mitel's 3300 can handle because, as illustrated in figure 1, the requisite functionality is embedded in silicon inside the platform.

The gateway requirement was foreseen at the design stage. A related issue that has to be considered is the functionality of the IP switch. PBXs are feature rich, and although nobody will use all the features, different individuals and user groups employ different subsets. When transitioning from TDM to VoIP, it is hard to overstate the importance of getting this first phase right. If some features are missing and work practices have to be changed, then the workforce will complain, and they may not make optimum use of the new IPC functionality.

Figure 1: Mitel's architecture enables anything-to-anything connectivity: TDM to TDM, IP to IP, and TDM to IP. Thus, the platform functions as a universal gateway.

Wireline or Wireless

PBX vendors made their money selling phones, not switches, and that model didn't change when they made the transition to VoIP. However, in many companies about 50 percent of the employees are mobile, a sector that includes so-called corridor warriors.

It's clear that there is no point in replacing one wireline phone with another when usage will be minimal. But replacing legacy mobiles with the new dual-mode (WLAN and cellular) smartphones and engineering the corporate infrastructure for wireless VoIP represents a significant investment as well as a major ICT task. So, what's the alternative?

Forget technology for the moment and consider the requirement. The combination of dual-mode phones and WLANs stops employees from calling each other on their mobile phones when they are in the office, which they do from habit, and it results in an unnecessarily high phone bill. But there would have to be a huge number of internal cellular calls before any return on the investment could be realized.

An alternative is to go for a solution that replaces regular phone numbers with a single "virtual" number. When a party calls this number, both legacy devices ring simultaneously; i.e., the TDM desktop phone and the 2G/2.5G mobile. If the called party is in the office, then he or she picks up the wireline device. Outside it's the mobile, but when entering the of- fice, an on-going call can be transferred to the regular phone.

PCs as Softphones

Dual-mode phones do VoIP in hot spots, but so do notebook PCs that have the requisite client software; e.g., Skype. Office PCs can be used in the same way. So this is another way of leveraging legacy investments.

The softphone is not a new development, but-and this is a personal opinion-IP PBX vendors downplayed softphones for the reason given earlier: the profit comes from the phones. In addition, this was regarded as a somewhat nerdy way of communicating in the early days of VoIP. However, that situation has changed. One reason for this is the use of Skype by business professionals on their home PCs.

Toward the end of 2006, Nokia introduced a software suite for desktop and notebook PCs, and the mobility bit comes via the service of an operator. No additional client software is needed in the phones, so the solution works with 2G/2.5G devices. This combination gives an end user the perception of a single network-one is broadband and the other cellular. In this way, there is seamless interoperability between services, devices, and networks.

Mobiles as PBX Extensions

So far we have taken a brief look at three pragmatic alternatives to dual-mode phones and WLANs: the virtual number solution; using PCs as VoIP phones; and Nokia's Communication Suite.

The rationale for these alternatives is mainly financial; i.e., the need to buy new phones and invest in a WLAN infrastructure. However, dual-mode phones have a lot going for them, with or without WLANs.

These devices have an open operating system; e.g., Symbian and Windows Mobile, a feature that enables client software to be downloaded (or embedded), which allows mobiles to operate in client-server mode. One example is the deployment of MS Office Communicator on the devices, which enables seamless interoperability with Office Communicator Server (OCS). The client IM software comes as part of a new release of the Office suite, and the real-time components, voice as well as video, come from third parties. This development is one of the ways that allows IT to manage wireline and wireless devices in the same way.

Siemens markets an interesting client-server solution that uses an application server on the enterprise's IP network. This controls the handover between fixed and cellular networks and gives mobile employees a unified mailbox, directory, and a single-number service. It currently runs on Nokia and HTC smartphones as well as Windows Mobile devices.

The IP PBX processes all calls-cellular as well as VoIP-and the mobility application server polls its clients in order to know which interfaces should be used (WiFi or GSM). Calls go to the server and then, as determined by the interface and location, they are redirected to the IP PBX, the cellular network, or a WiFi hot spot/home. In hot spots, an Internet channel will handle the call from the phone to the server.

CPE Versus Hosted Service

This is a key issue-arguably the key issue for large and very large enterprises. As IP communications continues to become more important to the efficient running of businesses both large and small, so does the complexity of the solution. In addition, workforces are becoming more knowledgeable and demanding as a result of developments in the consumer sector. This is particularly true in the case of knowledge workers. They will expect the equivalent functionality of Skype and Google to be available on their corporate desktops.

On one hand, CIOs and ICT management want to control the corporate infrastructure, which includes the mobile devices. On the other hand, the task of bringing hundreds of small offices and teleworkers into an enterprise-wide uni- fied network having a private four-digit numbering plan is a daunting task. One way to square this circle is to mix and match CPE systems with the services of global carriers. For example, deploy carrier-class systems in the data centers and use a service provider to unify the smaller site and bring them into the corporate environment. However, at the end of the day, while hosted services are an attractive option, there is a dependency that is close to a lock-in. Factors such as financial stability are therefore of paramount importance. (See the sidebar "Ten Tips from Verizon Business.")

Right now one factor that works well for service providers is the current mix of hybrid and pure IP lines. There are far more of the former than the latter, and while the growth of hybrid is relatively flat, pure IP won't catch up until the end of the decade. It works well because TDM is an alien technology to IT departments, but there is a clear and compelling need to bring legacy devices into an enterprise-wide infrastructure having a four-digit numbering plan. This is a relatively easy task for a global SP, and it represents an entry-level solution into the enterprise market. (See "Multiple Ways of Migrating to VoIP," VON Magazine, November 2006.)

This scenario could flip over when IP lines dominate corporate infrastructures around 2010-12. At that time service providers would only be needed for the MPLS-VPN since IT would be managing an all-IP environment. That indicates the need for SPs to offer desktop applications and thereby compete with Microsoft.

Now it Gets Interesting

Migration is a means to an end, and that end has to be considered when determining the route. For example, Unified Communications (UC) is the long-awaited "killer application," one that impacts on the cost of communicating, personal and workgroup productivity, and, last but by no means least, operational efficiency.

Proprietary UC solutions have been marketed for some time, but Microsoft's IM client, Office Communicator, is emerging as a de facto standard on which client-server UC solutions are being based. The server is MS OCS, a product formerly known as LCS. Right now the real-time components, video as well as voice, come from third parties-primarily the vendors of IP PBXs. However, many analysts and industry observers expect MS to offer a complete IPT solution and therefore compete with the current ecosystem of business partners. Companies may therefore decide to delay this part of their migration strategy until a clearer picture emerges.

UC's impact on operational efficiency comes via the integration of real-time communications with mainstream business processes. This is set to be the biggest of the three benefits-one that will raise the profile of application servers and lower that of IP PBXs. That is another development that should be factored into the long-term strategy.

Mesh Networks

Technology designed for one task often ends up doing something else. One obvious example of this is the Internet and VoIP. It's the same story with WLANs. Access Points (APs) are wireless extensions that were initially used at the departmental/ workgroup level. Large-scale deployment in enterprises is really an extrapolation of a single AP way of working. It's fine for data, but wireless VoIP issues are significant since they involve managing contention, interference, and QoS in a pervasive environment having both data and voice clients.

Mesh networks were designed for task. Meru (www.merunetworks.com) has developed technology that works at the 802.11 MAC layer to manage contention and exert more control over client access. Clients only contend for the channel at their designated time, and if the client has no data to transmit, then it doesn't get a slot. Meru states that a single AP can handle more than 100 active clients. Regular WLANs max out at around 10 to 15 calls per cell.

Next-Generation WLAN Architecture

First-generation WLANs employed the "fast AP" model, so called because each AP acts as a self-contained router. That was the original design. These APs have the intelligence needed to associate with client devices and manage traffic flow. It's a distributed model. It's easy (too easy) to add more APs, but the end result is a network that is difficult and expensive to manage.

The industry's response was the development of "thin APs" and a centralized architecture. This solved the management problem, but it's an inefficient way to handle traffic since these APs forward all traffic to the central WLAN appliance, which then forwards it to the required destination.

Trapeze Networks (www.trapezenetworks.com) has developed a next-generation architecture that merges both distributed and centralized architectures. The design is based on the idea that application requirements should dictate the architecture. As mentioned earlier, first-generation WLANs don't do VoIP particularly well, and the arrival of the 802.11n standard will boost traffic by an order of magnitude. This will overload today's wireless switches.

Trapeze's APs aren't fat or thin, but they are smart. There's enough intelligence to perform cryptography and policy enforcement at the point closest to the client device. They also do data forwarding through the distributed network. In addition, IT management can decide which data is forwarded centrally and which gets forwarded to the distributed base.

The latter is used for VoIP traffic since it is latency sensitive. The AP forwards the traffic so packets take the shortest possible path. This facilitates migration to 802.11n and eliminates the need to upgrade legacy WLAN controllers. Security-sensitive data traffic can be delivered directly to the controller, which then forwards it to the AP onto the client device.

Making it Reliable

Reliability is the single most important parameter of a business phone system. You pick up the phone and get dial tone. Period. That's what the PSTN and PBXs have delivered, and anything less is unacceptable.

Many IP PBXs achieve five-nines availability via redundancy, which is heavy-handed engineering that adds to the cost and complexity. It's heavy-handed because VoIP networks are intrinsically fault tolerant. Rerouting traffic is an intrinsic feature of IP. Without it the Internet wouldn't work. ShoreTel has leveraged this fact. The architecture is fully distributed; there's no single point of failure and no redundant hardware. Instead, ShoreTel employs the n + 1 redundancy technology whereby the embedded intelligence makes each box a peer.

This allows call control to be distributed around the network. Call control is provided by every softswitch in the system. In the regular centralized model, call control is implemented in a computer system that is typically embedded in the IP PBX.

Another interesting reliability feature is the use of flash memory instead of hard disks and a bullet-proof OS (VxWorks).

Bob Emmerson, our European Editor, co-authored with Jeff Pulver the book, Run Your Organization in Real Time. He can be reached at bemmerson@vonmag.com.