
Besides the configuration services, the CallManager also provides basic connectivity and call-routing services to the devices on the network, acting as a proxy among these devices, including both the Selsius equipment and H.323 nodes. Whenever a call between two devices is placed, the call setup is negotiated through the CallManager. Calling systems pass dial strings and commands to the CallManager, which locates the destination and then passes along the request(s). Once the call has been negotiated with both end points, all subsequent voice traffic passes directly between the two systems, though subsequent command traffic is passed through the CallManager.
This proxy design works quite well, and adds to the overall functionality of the system. For example, if an H.323 client (such as Microsoft's NetMeeting) wants to call a user with a Selsius phone, it passes the destination system's extension number to the CallManager, which then attempts to contact the destination on behalf of the calling system. If the destination device is busy, the CallManager could return a "busy" signal, or could forward the call to the voicemail inbox associated with the destination, depending on how the destination device is configured. Without this middleman service, none of these advanced features would work.
However, not everything is rosy, mostly because of problems with the state of most H.323 implementations. For example, while a Selsius phone can issue a "hold" request to the gateway, which will then forward the request to the destination system, the end node may not understand the request and will likely drop the call. This was typical of the interoperability issues that we found in our tests, with NetMeeting clients and Cisco gateways dropping calls whenever any sort of advanced functionality was attempted. Although we were able to integrate the H.323 devices into the CallManager's extension pools and routing plans, with non-Selsius products we were unable to do much beyond placing calls, as any greater level of interoperability between H.323 systems is virtually nonexistent.
The Phones Perhaps the most interesting part of the Selsius IP-PBX suite is the phones. These units have the same look and feel as regular multiline handsets, but have 10BASE-T Ethernet ports instead of RJ-11 phone jacks. A variety of units is available, ranging from a 30-button console to a plain 12-button handset. Selsius also has a software-based phone that can be used in lieu of NetMeeting (or in conjunction with it) for tight integration at the desktop. Button assignments include features such as "hold" and "transfer," as well as programmable speed-dial entries, message checking and other typical business-class telephony services.
Each of the hardware-based phones has a full-blown IP stack, a dedicated processor and on-board codecs, which provides exceptional voice quality, very low latency and high reliability. However, these features come at a fairly steep price, ranging from $550 for the 30-button console to $375 for the base 12-button unit. Although these prices are in line with the digital multiline feature phones we've seen ship with most PBX systems, they are considerably higher than the analog handsets you can get for $20 at a flea market.
Also, there are some issues that arise if you put an Ethernet phone on every desk, requiring multiple Ethernet drops to each office. If you do not have two sets of CAT-5 wiring running into every office, you will need to do some planning. Although the phones come with two Ethernet ports--providing the ability to downstream a PC's Ethernet connection off the back of the handset--those ports are fixed at 10 Mbps, meaning you'll need a separate Ethernet run if your desktop PCs have 100-Mbps adapters. Furthermore, the phones do not support 802.1Q VLAN tagging and prioritization services, restricting their ability to provide these services to downstream Ethernet devices.
Our experience indicates that these phones should be used on a separate LAN infrastructure from the PCs, as large amounts of traffic from either camp can cause problems for the other. If the PC user begins a large download, it is very likely that the RTP voice traffic will be bumped or delayed, or that the TCP-based client/server application will time out from excessive collisions and retries caused by the RTP traffic. In the best design, you'd want to dedicate ports and queues on your switching infrastructure for the phones and PCs, keeping the traffic separate from end-to-end across your network.
Hanging Up Overall, our experience with the Selsius IP-PBX suite of products left us with very little to complain about. The only significant technical issues that arose during our tests--loss of signal from the overzealous noise-suppression and lack of support for G.729 codecs--should be addressed and resolved within the next few months.
That leaves the cost issues. The phones are not cheap, and neither are the gateways (an eight-port analog gateway costs $4,795). Separate licensing costs for the CallManager software ($25,000 for a 100-user installation) makes the solution more expensive.
The complete cost for a 100-user installation, including licenses, gateways and phones, runs close to $70,000. Although this compares favorably with a PBX system of similar scale, we could expect to get much more functionality from a true PBX.
In the end, though, having the ability to manage our telephony equipment using the same tools and services that we use to manage our PCs makes the extra up-front cost negligible. All in all, we found these units to be more than useful. Anybody interested in deploying first-generation VoIP technology will likely find this to be an excellent choice.
Eric A. Hall is president of EHS Co., a network technology research and testing firm in San Mateo, Calif. Send comments on this article to him at ehall@ehsco.com.
|