VoIP 101 - Useful information to assist you in understanding VoIP

Internet protocol telephony (IPT) provides the foundation for what, without question, will become a major driver for enterprise productivity improvement – Unified Communications (UC). The basic premise of UC is that if a means of communication is available to two or more parties, then they should be able use it intuitively.

In order to provide you with some detailed information about the various considerations which need to be addressed when evaluating, implementing and supporting a VoIP UC system, we have compiled an extensive overview which can be found here on our website. This guide is available in hardcopy. Please contact Gregg Communications by phone at (630) 571–7000, or at sales@greggcomm.com to reserve your copy. Gregg Communications’ personnel also are available to assist you in working through the details of what VoIP and UC technology can do for your business, and best practices for implementation and on-going support.

We hope the following information is helpful!

Visit weekly for more VoIP information!

Visit next week to see Chapter 3 (3.2) of 8!!

Chapter 3 (3.1):

3. DATA NETWORKING
The fundamental building blocks of a typical enterprise data network are Ethernet, switching, IP and routing.

3.1 LAN INFRASTRUCTURE
Today, a majority of LANs are based on Ethernet technology. Increasingly, IP runs over Ethernet, replacing protocols like Systems Network Architecture (SNA), IPX and AppleTalk.
Ethernet has moved from shared bus to shared hub, and today, switched Ethernet dominates the enterprise. Speeds have also improved from 10 Mbps to 10,000 Mbps (10 Gbps IEEE 802.3ae). One of the earliest evolutions was the shift from shared coaxial cable to twisted pair cable. Ethernet is standardized as IEEE 802.3. Each device on the network has a unique six-byte media access control (MAC) address. Three bytes identify the vendor and a different three bytes identify the specific device. Information is sent between network devices using a predefined format known as a frame (see Figure 2). Frame formats continue to evolve but are backward compatible.

The NIC device resides on all networked devices in one form or another. It sends and receives all the signals to and from the device, and is responsible for packaging raw information produced by network devices into frames, before the data is sent to the cable that connects the device to the network. If the target address of a specific communication does not match that of the NIC, then it simply ignores the frame.

To communicate with another device, the sending device first listens for a quiet period, then begins transmitting. It listens to make sure that its transmission has been correctly sent, i.e.: the checksum matches the transmitted data. To prevent two devices from communicating at precisely the same time, Ethernet employs a scheme known as Carrier Sense, Multiple Access/Collision Detection (CSMA/CD).

Here’s how CSMA/CD works. Imagine two computers that hear silence on the media and determine that it is safe to transmit. They both transmit and listen at the same time, so if another device heard silence and started transmitting at exactly the same time, they would immediately recognize it, because the information they detect coming back on the network would not match what they sent. They have detected a collision.

As a result, each computer then backs off – quickly flooding bits onto the cable and ceasing transmission for a random amount of time. They resume listening for a quiet slot and the cycle begins again – though this time, hopefully without a collision. This simple mechanism has been found to be fairly scalable and works well on LANs with a small number of users. Over the years, Ethernet’s CSMA/CD has outperformed competing approaches, such as IBM’s Token Ring.

As you design a network infrastructure, keep in mind that
Ethernet has some important constraints in terms of cable lengths:

• The maximum length of twisted pair cables that connect Ethernet switches to devices (other switches, computers, IP phones, and so on) is 100 meters.
• The maximum fiber cable length is 420 meters.

Today’s cabling is typically UTP type 5 or 5e. A test certificate should be obtained from the cable contractor to ensure the cable conforms to Ethernet requirements. (RJ-45 pin layouts are defined in TIA 568B.) When you connect an Ethernet switch or hub to another switch, you need a crossover cable (also defined in TIA 568B), unless the uplink port of the switch is used.

Although the hub and spoke topology (one of a number of different topologies that can be used with Ethernet) created by a LAN switch (the hub) and several NICs (the spokes) is superficially similar to the PBX and twisted pair cable that connects telephone handsets, there are fundamental differences between the two systems. These differences include:

1. Unlike PBX systems, LAN devices can be easily interconnected and daisy-chained to extend the network’s capacity.
2. Unlike PBX systems, Ethernet devices are backward compatible, so older NICs continue to work with newer switch ports.
3. Unlike PBX systems, Ethernet is an open standard (IEEE 802.3), and any compliant device can be added to the network, irrespective of vendor.
4. Unlike PBX systems, addressing schemes with Ethernet are relatively easy to implement, because Ethernet-compliant devices have unique MAC addresses built into the hardware, enabling network managers to deploy Ethernet without having to manage the addressing scheme. Devices simply “declare themselves” on the network. IP addresses above this layer must still be managed, but even these can be allocated automatically using a scheme like Dynamic Host Configuration Protocol (DHCP).

3.1.1 ETHERNET SWITCHING
Over the last 25 years, Ethernet has evolved. Today’s Ethernet networks are built from both chassis-based and stackable switches, rather than shared media hubs. A switch provides each device connected to one of its ports with a dedicated bi-directional (full duplex) connection. This means that a device connected to a switch port communicates at the maximum speed supported by that device. This differs from shared Ethernet topologies (such as hubs or, more often today, wireless), where the bandwidth is shared.

To achieve this improvement, Ethernet switches must know the addresses of as many of the devices connected to them as possible, and identify the ports used to reach these addresses. Switches automatically do this using a protocol defined in IEEE 802.1, transparent bridging. The switch stores the source address and switch port of every frame it receives in a table, and finds the destination devices by flooding its other ports with a request for the destination device. When the destination device responds, its address and port number are added to the table. After source/destination addresses are known, the switch uses that information to begin forwarding frames. Shared hubs forward frames to every device connected to them, reducing overall throughput for every device.

In some circumstances, it makes sense to segment traffic either by department or by application using a virtual local area network (VLAN) in order to enhance security or optimize bandwidth. Given that voice and data are sharing the same switch infrastructure, it may make sense to segment the LAN into smaller groups of users to protect real-time voice traffic from unpredictable data traffic (which can create spikes of high-volume traffic over brief time periods). One method of ensuring optimum voice quality is to run voice traffic on a separate VLAN. This virtual segmentation allows voice traffic to share the same physical infrastructure as bursty data traffic, but voice traffic is protected at a logical level from interacting with data traffic.

Ethernet switch architecture can also be designed to eliminate points of failure – uplinks, specific switch ports – that could impact everyone in a department or office floor. Redundant links can be built between switches, but this introduces the problem of a logical loop, where switches keep claiming they are responsible for devices that are, in fact, connected to some other part of the network. Or worse, the redundancy could lead to broadcast storms, where switches continue forwarding broadcasts and network devices respond to those broadcasts, until the responses feedback on themselves causing a network meltdown.

The spanning tree algorithm IEEE 802.1d provides a way to benefit from the redundancy, while avoiding the problems described above. Each link is weighted. For any path, a switch uses only the lowest path for a link, ignoring the others. It should be noted that spanning tree information takes time to update in the event of catastrophic failure, and this process of updating spanning tree information can impact a call in progress. Switch manufacturers have developed proprietary solutions for providing rapid spanning tree updates, and these solutions largely address the challenge of maintaining call quality when the network is experiencing technical issues.

When designing enterprise networks, it is important to recognize that in spite of efforts to segment traffic, much of the traffic still transits certain links, resulting in bottlenecks. The IEEE 802.3ad standard addresses this bottleneck by providing a standard mechanism for aggregating multiple links between switches.

In concluding this section on LAN infrastructure, we would like to point out that while Ethernet’s plug-and-play design makes it easy to implement, as the leader of your organization’s migration to a fully converged voice/data network, you are seeking to implement advanced networking capabilities, such as redundancy, link aggregation and quality of service (QoS), which require careful planning and fine tuning.

3.1.2 POWER OVER ETHERNET
One of the advantages of an IP-based PBX system is that it enables the use of a converged network (as opposed to maintaining two separate networks for data and voice). IP telephones plug directly into the Ethernet network, and interact with a media gateway controller (MGCP) or a gatekeeper (H.323) for call control over the LAN.

However, the challenge of a single, converged network is that the phone (which is seen as a lifeline to emergency services) may not be available during a power outage. With users expecting dial tone no matter what else is going on around them, this can create problems.

Even though in many cases, digital sets were not line powered, the legacy PBX vendors—seeing an opportunity to hold back the inevitable— accused the IP-PBX community of cutting corners on fundamentals. This led to the myth that data networks need to be upgraded so that low voltage devices like wireless hubs (see next section) and IP telephones could be powered through the LAN.

The IEEE 802.3af standard defines two ways to provide power
to IP phones:

1. End Span – Replace Ethernet switches with new devices that utilize DC current over the pairs used for data 1/2 and 3/6 (on the RJ-45 jacks). This approach is most appropriate for a new building or as part of a major network upgrade, because it requires new Ethernet switches.
2. Mid Span – This device inserts power onto the unused 4/5 and 7/8 pairs on the RJ-45 jacks. The device has two ports and sits between the Ethernet switch and the device it is powering. It is a less expensive option than upgrading the Ethernet infrastructure.

Note: In both cases, the system is non-destructive, because a noncompliant device can be plugged into the powered line without damaging the device.

3.1.3 WIRELESS LANS
In the USA, wireless premises for voice has remained vendor specific, in contrast with Europe, where regulated spectrum was established and a standard approach has been widely adopted – the European Telecommunications Standards Institute (ETSI) Digital Enhanced Cordless Telecommunications (DECT) technology. In 2005, the Federal Communications Commission (FCC) opened up a spectrum in the 1900 MHz band, which effectively means DECT can now be employed in the U.S. Whatever air interface is selected, the challenge with wireless has always been money. The cost of providing full building coverage, campus-wide roaming and establishing the necessary access points or base stations is often too high. It is recommended that a careful financial analysis accompany any large-scale wireless project, because technology decisions can lead to vastly different price points.

WiFi or 802.11 technologies are the current leaders in mindshare for wireless premises in the U.S. market. Broad market adoption has helped drive access point prices down, and many enterprises are intrigued by the potential to provide employees with a single mobile device for applications, as well as voice.

The 802.11 standards continue to evolve, and the 802.11n standard is expected to be ratified in late 2008. It improves on previous 802.11a and g standards (54Mbit/s) to deliver a maximum data rate of 248 Mbit/s.

Key factors to keep in mind when planning a wireless implementation:

1. QoS – Delays for enterprise voice should not exceed 150ms. Given that WiFi is a contention protocol (like the original shared Ethernet), when a particular access point is heavily used, voice quality suffers. Ratified in 2005, IEEE 802.11e defines traffic classes, assigning time-sensitive voice traffic to a higher class relative to other traffic types. Not all wireless solutions implement 802.11e. We recommend favoring solutions that include this standard when designing voice over WiFi solutions.
2. Reliability – Except for 802.11a, which specifies the 5GHz band, 802.11 standards use the 2.4 GHz band and suffer from interference with other wireless devices.
3. Security – A voice-specific encryption standard, IEEE 802.11i, was developed. However, data encryption can slow down voice delivery.
4. Standards – Multiple signaling standards (802.11a, 11b, 11g and the emerging 11n), should be compared based on distance, capacity and frequency interference. For international markets, keep in mind that not all standards or in some countries, channel slots within the standards, are available.
5. Handsets – Running voice through personal digital assistant (PDA) devices may be desirable, but there are implications for battery life, overheating and emissions that should be carefully evaluated prior to an organizational adoption. Traditional wireless premises phones are more successful.
6. Coverage – At the high frequencies used by wireless local area networks (WiFi), establishing coverage throughout a building is not a trivial task, and should be carried out by experienced wireless network designers