Some good rules of thumb to follow when determining where and how many network switches may be required for the project are:

The following diagram illustrates how simplifying the network connectivity can reduce delay in the network:

Many network administrators wrongly assume that because a network doesn't have much network traffic, or because they are running Gigabit switches, they don't need to worry about the small amount of network traffic generated by a phone conversation.

If voice packets do not leave one device and arrive at the second device as quickly as possible and in the same order, system users will experience a poor-quality call. This poor quality will take the form of delay or jitter. Delay is the amount of time it takes the person listening to begin or end hearing what the talking party is saying. Jitter is broken or jumbled parts of a conversation such as typically experienced in a poor quality cell phone call. Echo and static are examples of other quality problems, but these are not created by poor network design.

The ability to identify certain traffic on the network as being of higher importance is referred to as Quality of Service (QoS). Network traffic can be flagged with a priority value, which is in turn honored by network switches. QoS helps to ensure that packets traverse the network and arrive at their destination in an orderly and timely fashion.

There are two types of QoS that are most common across switching products: Differentiated Services Code Point (DiffServ or DSCP) and Class of Service (CoS). DSCP operates at the IP protocol level (Layer 3 of the Open Systems Interconnection or OSI model—http://www.itu.int/rec/T-REC-X.200-199407-I/en) and utilizes a field in the header of each IP packet to identify the priority of the IP packet. DSCP is defined in IETF's RFC 2474 (http://tools.ietf.org/html/rfc2474).

CoS uses a 3-bit field in the Ethernet frame header to identify the priority of the packet. This standard is referred to as the IEEE 802.1p standard (http://ieee802.org/1/). Since CoS operates at the Ethernet frame level, it is considered to operate at OSI Layer 2 and is not routable as such. DSCP is routable because it operates within the IP packet.

Virtual Local Area Networks (VLANs) allow network administrators to segment their networks into separate logical pieces. By segmenting the phone network from the data network, the system designer is able to separate IP addressing, DHCP, and DNS services. Separating services can make integration easier than attempting to integrate with existing services. This segmentation can also make the phone system more reliable because it is not dependent on anything on the data network in order to operate.

The preceding network diagram illustrates the difference between how network elements might be physically and virtually connected.

Additionally, if the phone network is segmented from the data network by a VlAN, it is possible to firewall the two networks from each other. This is advantageous if there are applications on the data network that cause problems with the communications systems, or if there is some sort of denial of service attack that originates from the data network.

VLANs operate at the Ethernet frame layer (Layer 2) and their interoperability is defined in the standard referred to as IEEE 802.1Q (http://ieee802.org/1/). To ensure interoperability between switches if different vendor products are selected, make sure that all of them support the IEEE 802.1Q standard.

How to power the IP phones is one of the most important factors while selecting network equipment. Convenience, cost, and reliability are all factors that need to be considered.

The most convenient way to power the phones is with Power over Ethernet (PoE). PoE is a network standard that detects if a device that is plugged into the network can accept power, and then provides DC power across the network cable.

Power injectors (also referred to as midspans) also can inject power in the network closet. They are inserted between the network switch and a PoE-capable phone. If the existing network equipment can support your network properly and handle Quality of Service, power injectors may make financial sense. However, power injectors introduce extra cables and extra equipment into the network closet. This may reduce reliability and is certainly not as convenient as knowing that every network port has power available on it.

Both PoE switches and power injectors allow the network administrator to power all phones from the convenience of the network closet. In a power outage situation, phones can stay operational for as long as uninterruptable power or generators allow. The most widely accepted standard for PoE is IEEE 802.3af (http://www.ieee802.org/3/). Some older Cisco equipment utilized a proprietary PoE standard that has been largely abandoned. Additionally, not all PoE switches are created equal. Some may not be able to supply enough power for all of the devices needed to connect to the switch. Verify the power consumption of the phones to be used and make sure the switch can provide enough power.

Optionally, AC to DC power adapters can be utilized at each phone to plug them into a wall outlet (or a small local UPS). Not all phones come with power adapters, so it is important to consider this potentially added cost at the time of purchase.

If Gigabit Ethernet and single network cable are requirements, network switch and IP phone choices become more limited and expensive. IP Phones typically have a two-port network switch built into them. The switch in the phone allows the phone to sit between a user's PC and the network switch. If the switch in the phone is only rated at 100 Mbps, the PC that connects to it will connect to the network at a maximum of 100 Mbps.

If Gigabit Ethernet to the desktop is required and a single network drop is to be utilized, phones with in-built Gigabit Ethernet switches must be utilized. If Gigabit Ethernet needs are limited and network equipment costs are a factor, consider running separate runs to those computers and using 100 Mbps switches for phones.

Most network administrators prefer to have homogeneous network equipment—not because network equipment won't interoperate, but simply for ease of management of the equipment. Consider what equipment you have in place and if any of it will meet the later needs of the communications system. Either DSCP or CoS is a "must have" for a quality deployment.

Analog gateways provide Foreign Exchange Office (FXO) and/or Foreign Exchange Station (FXS) interfaces. The FXO lines connect to Plain Old Telephone Service (POTS) lines provided by the telecommunications provider, whereas the FXS lines connect to analog stations (phones or fax machines).

Analog gateways come in many different port configurations from one to thirty-two ports. Some analog gateways will have a mix of FXO and FXS ports, and some may even allow ports to be configured as either FXO or FXS.

Most analog gateways will support caller ID service from the telecommunications provider if enabled on the phone line. The caller ID will be passed directly to the receiving telephone for display (if it is equipped with a display).

The most popular analog gateways are from AudioCodes and Patton Electronics. The AudioCodes gateways have the ability to be automatically configured in sipxconfig, whereas the Patton Electronics gateways are more configurable.

Digital gateways will provide interfaces to PRI, T1, E1, or BRI digital phone lines. Because the audio is sent in a digital form to the gateway, the call quality is much better than that of the typical analog phone line.

Additionally, the added feature of DID (which may be referred to as Direct Dial In or DDI in Europe) numbers is that the telecommunications provider will provide a block of phone numbers that are assigned to the circuit. The number of DIDs allocated to a circuit does not correspond to the number of channels on a circuit (for example, a 23-channel PRI may have 200 DIDs assigned to it). As a call begins, the number that was dialed is passed into the gateway before the ringing begins. This dialed number can then be used to route the call to a chosen extension on the phone system.

When selecting a digital gateway, it is important to verify what your telecommunications provider will be supplying for the digital circuit encoding and ISDN signaling on the digital circuit. Then make sure that the gateways you select can accommodate the provider.

For T1/PRI circuits, the line encoding is typically B8ZS (Bipolar with 8 Zeros Substitution) or ESF (Extended Super Frame). For E1 circuits, line encoding is typically HDB3 (High Density Bipolar three coding). For the signaling to the provider's phone switch, the provider will typically specify what signaling it will be utilizing. In North America, DMS-100 (Digital Multiplex System-100, a Nortel standard) or NI2 (US National ISDN Phase 2) signaling standards are the most prevalent.

Like the analog gateways, the most popular digital gateways are from AudioCodes and Patton Electronics. If you are going to have a mix of analog and digital gateways, consider staying with the same brand to ease management.

Hard phones are available from many different vendors. They come in many different shapes, sizes, and costs. Almost any SIP phone can be made to work with sipXecs by manually configuring it. However, manually configuring phones would become burdensome in all but the smallest system deployments.

To ease system administration, there is an ever-expanding list of phones and FXS gateways that can be automatically configured by sipXecs' sipxconfig server. The current list is as follows:

Not all of the above-listed phones support all of the phone system features. The most commonly utilized and most compatible phones with sipXecs are:

Spend some time researching the different phone vendors and the different models they offer. Take into consideration the information gathered earlier in the chapter about how the phones are used in the organization and what features are important for individual users.

If there is any question about whether a phone will be compatible with the sipXecs system, free interoperability testing is available at http://interop.pingtel.com/.

With as many hard phones that are available, it seems like there are twice the number of software-based phones available. Their functionality seems to vary across the board. The great advantage of software-based phones is their low cost. The disadvantage is that it is an interface that users are not used to. Consider the users of the system and how they may adapt to this change.

The most popular softphones are from Counterpath (http://www.counterpath.com). Their free X-Lite softphone is used by many. If more integration with desktop applications (such as Microsoft Outlook) is required, consider Counterpath's eyeBeam or Bria products.

An important consideration when using softphones is whether the existing computers in the organization can utilize them. Since all of the voice traffic travels through the PC, the PC needs to be fast enough to process that traffic as well as run any applications the user may be using when on the phone. If the computer is too slow, the quality of the call can suffer.

For sound input and output from the softphone, a headset is utilized. The best option for this is a headset that uses USB and has its own sound processing based in the headset instead of trying to utilize system-integrated sound cards. Poor quality audio may be experienced when utilizing the sound card integrated into the user's PC.

There are many different types of headsets available in the market. Long known for their phone system headsets, Plantronics and GN-Netcom have a wide selection of products for computers. The administrator should verify compatibility of any headset with the softphone that is being utilized.

There are two types of SIP cordless phones available for use in the market. The phones either use a standard Wi-Fi network (802.11b/g), or they rely on a separate wireless Digital Enhanced Cordless Telecommunications (DECT) network. It may seem like a no-brainer to use the same Wi-Fi network deployed for mobile computing use in the organization. However, Wi-Fi phones suffer from poor battery performance. Additionally, the Access Points ( AP) deployed may not support QoS and the phones may not roam from AP to AP smoothly.

DECT wireless handsets will cost more to implement because a separate wireless infrastructure will be required to support the phones. The payoff, however, will be with better voice quality and noticeably longer battery life in the handset.