2.1. VoIP or IP Telephony
Are "VoIP" and "IP telephony" two different technologies, or do they both describe the same thing? Well, it really depends on whom you ask. Some vendors prefer IP telephony when referring to their IP-based voice offerings, arguing that VoIP refers to the specific act of transmitting digitized sound data on an IP network and IP telephony refers to the overall technology family. Others give VoIP the broader definition, identifying it as inclusive of IP telephony, and referring to IP telephony only as the act of mimicking traditional telephony applications.
For the purposes of this book, we'll take the latter tack: VoIP refers to the overall technology family, while IP telephony means specific application functions such as call signaling and voice mail. So when we talk about conference calling, we might call it telephony, but when we talk about conference calling, call-waiting, and voice encoding, we will refer to them collectively as VoIP. In general conversation, though, VoIP and IP telephony can be used interchangeably.
2.1.1. VoIP's Pros and Cons
VoIP certainly has a few disadvantages when compared to old-school phone hardware. High-utilization service guarantees are harder to deliver with VoIP than with an old-fashioned PBX. The same scalability characteristics that attract people to VoIP can ultimately be the reasons their implementations fail: a VoIP network can be so extensible that service-level guarantees are hard to make, whereas a traditional circuit-switched voice network has hard capacity limits, around which levels of service tolerance can be guaranteed easily. Certain broadcast audio applications, like overhead paging, can be difficult with VoIP, too.
The gains VoIP brings to the table far exceed the few difficulties it imposes, though. There's nothing that old PBX can do that a VoIP telephony system can't, even if VoIP makes a few things tougher.
One thing VoIP makes easier is physical provisioning. While a PBX requires a network of electrical, usually copper wire, loops , VoIP requires an IP network. Since IP networks are a staple of every modern business, the logistics of building a network for voice is largely simplified because the required physical elements are already in place for other common business applications: databases, messaging, Internet access, and so on. VoIP is carried on the network the same way those are.
If you're an Internet user (and who isn't these days), then you know TCP/IP is the core protocol that defines the architecture of the Internet. In most organizations, and even in many homes , a TCP/IP local area network is an important interpersonal communications tool, used for email, web surfing, and instant messaging. When VoIP replaces the traditional telephone using TCP/IP, the local area network becomes the key piece of telecommunications infrastructure.
Once that key piece is standardized within the enterprise, VoIP administrators have only one network to maintainthe one that supports TCP/IP. This means supporting a single network cabling system, rather than separate ones for voice and data. If you use wireless Ethernet, you don't need local area cabling at allVoIP will still work. Meanwhile, old-school PBX administrators still have to maintain a separate local area cabling plant that serves only the PBX system.
But that key piece of telecom can be a key failure point, too. When the voice and data networks are separated, as they are in traditional telephony, their physical paths lie separately, protecting the voice system from failures isolated on the data network, and vice versa.
But with VoIP, these paths converge. When the path is broken by an equipment failure, a power failure, or a construction crew accidentally slicing underground cables, the data network fails. When a computer virus swamps your data network, VoIP phone calls may no longer be possible. When data fails, voice fails, too.
Even in the home, where you might rely on a cable or DSL Internet connection, your VoIP calling capability will swiftly disappear when your broadband provider's service fails or your power goes out.
2.1.2. VoIP Network Fundamentals
Since VoIP is layered on top of TCP/IP, you must have some form of TCP/IP network in order to use it. For small VoIP experiments, any Ethernet LAN will do, even a hub-based or wireless one. But for larger, critical VoIP implementations, your choice of network infrastructure will be critical. For starters, using broadcast Ethernet devices like hubs is a poor choice, as is using early-generation Ethernet switches that lack quality-of-service features. Wide area equipment, like routers, will need to support these features, too. ( Quality-of-service features are covered in detail in Chapter 9.)
Generally speaking, the faster your switches, routers, and network links are, the better your VoIP network will perform. Nothing beats good ol' speed for increasing the performance of a wide area network, but sometimes a slow network link is an economic or geographic necessity. VoIP is a speed-sensitive business, as you'll find out.
2.1.3. The Layers of a VoIP Network
Like other networks, VoIP networking can be described using the Open Systems Interconnect (OSI) reference model, a standardized way of describing the different parts of the data communications process. The OSI model has seven layers that represent each part: physical, data link, network, transport, session, presentation, and application. The purpose of the OSI model is to simplify connections between different types of networks and to allow engineers who design network applications to assume a standardized platform upon which to build.
220.127.116.11 The physical layer
The OSI physical layer is the most fundamental part of the datacom process. It's the layer that provides for the electrical, mechanical, radiant, or optical signaling pathways that are required in order to move data across any data network. In an IP network, the physical layer can include twisted-pair LAN cabling, plugs, cross-connects and patch panels, power sources, V.35 cables like those often used with serial interfaces on routers, and so on.
Though the physical layer is itself intended to be permanent and stable, its assortment of connective technologies (copper twisted-pair, fiber- optic cables, etc.) are prone to noise and distortion, two problems that cause data transmission errors. The physical layer has no way of dealing with these problems, and that's why certain guidelines related to distance and interference exist at the physical layer. For example, a 100BaseT Ethernet connection on twisted-pair copper cable cannot be longer than 100 meters .
18.104.22.168 The data link layer
Since the physical layer is not immune to the laws of physics and the signal degradation they incur, the data link layer provides a medium for detecting errors in data transmission. Error detection at the data link layer works on behalf of a single physical link, such as an Ethernet segment or a single T1 circuit.
The data link layer " frames " the continuous stream of signals flowing across a link. Framing means delimiting that signal into manageable pieces, called frames (imagine that). For error detection, each frame can be subject to a CRC, or cyclic redundancy check. With certain types of connections, error correction can be attempted.
The data link and physical layers are often viewed as one and the same, and in many network substrates, such as Ethernet, their functions tend to be inseparable. That is to say, you can't build an Ethernet physical layer without building its data link layer tooboth layers are facilitated by the same device, which is usually an Ethernet interface, a hub, an Ethernet switch, or an Ethernet coax bus. The data link layer is the lowest layer that VoIP applications can reference, and usually only in an indirect manner (only quality-of-service functions interact with the data link layermore on this in Chapter 9).
22.214.171.124 The network layer
While the data link layer provides data framing over a single physical connection, such as an Ethernet segment, the network layer provides the logistical intelligence required for a single network to exist across many physical connectionssay, two Ethernet segments connected by a wide area connection. The network layer is more visible to applications than the data link and physical layers. Its job is to provide:
The network layer isn't the lowest layer that is relevant to VoIP, but it is the lowest layer that VoIP applications must reference in order to function. For example, datagrams and addressesthings implemented at the network layerare critical to the functioning of VoIP applications.
The addressing scheme used by VoIP is inherited from IP. Each device on an IP network has an IP address, so each VoIP endpoint has one too. An IP address consists of 32 bits, commonly presented by four 8-bit figures separated by dots:
Each figure in the address allows for 256 values, so the overall address space (32 bits) of IP's addressing scheme allows for about 4.3 billion addresses. The newer version of IP, Version 6, allows for a 128-bit address space, but adoption of IP Version 6 has been slow, and this book deals strictly with the 32-bit address scheme of IP Version 4the protocol that today's Internet runs on. In the context of the Internet and IP networks, the network layer is sometimes referred to as the Internet layer .
Using IP addresses, the network layer can facilitate wide area networking over dozens, hundreds, thousands, or millions of physical links. Consider the Internet, which uses IP to connect millions of disparate networks. Individually, each of these networks tends to share a group of related addresses. Each group is what IP calls a subnet .
Every datagram sent across an IP network contains a source and destination address so that the devices responsible for maintaining the network layer know where to route the datagram. However, the network layer isn't responsible for any kind of error controlthat's the job of the next layer up.
126.96.36.199 The transport layer
Even though the data link layer provides error detection on an individual network link, this alone isn't enough to satisfy the needs of a large, application- intensive network. That's why the transport layer provides error control across the entire networkfrom sender to receiverregardless of the number of physical links between them. Transport layer error control operates independently of the measures provided by the data link layer, which tend to be specific to the type of link they are responsible for.
On the transport layer, protocols have been designed for two kinds of service:
The kind of service elected depends on the needs of the application. Some applications don't need a high degree of reliability (video gaming, for example), while others must have absolute reliability (bank transactions). Within the transport layer, IP provides protocolsUDP and TCPthat handle both needs.
188.8.131.52 To connect or not to connect
Within IP, datagrams can be delivered using a "best effort" approachthat is, the host transmitting the datagram will not know whether it was received by the intended recipient. Also called connectionless networking , this method is employed by the User Datagram Protocol (UDP).
If you've ever played the Quake series of video games over a network, you've used the UDP Protocol. UDP excels in situations in which very fast delivery of data is a requirement, and reliability features, like confirming that the data has been delivered, would be a waste. In a multiplayer network game such as Quake, you and the other players each control an armed character that is trying to kill the others in a virtual 3D world. Real-time delivery of characters ' location and trajectory data within the virtual world is critical to the game play. Even a slight delay in delivery of these datagrams could mean life or death for your Quake warrior . Delivery guarantees impose too much overheadbecause dozens of UDP datagrams can be used by Quake in a second.
The same is true of the traffic carried over the network during VoIP phone calls. This traffic is carried across the network at a rate of between 30 and 50 datagrams per second. To verify delivery of each one would introduce a performance bottleneck that is unacceptable in a voice application. Therefore, almost all voice data flowing across a VoIP network is considered connectionless and carried by UDP.
The more reliable protocol for data transmission in an IP network is called Transmission Control Protocol, or TCP. Like UDP, TCP is encapsulated within IP. TCP's distinguishing characteristic is that transmitters using TCP must set up a transmission channel, or connection, before they send data to their receivers. For this reason, TCP is considered a connection-oriented protocol.
Error control takes place during a TCP transmission. At the end of the transmission, the sender and receiver agree to end their conversation, and the connection is closed. TCP guarantees that packets will arrive in the correct order, too. Because TCP is so cautious compared to UDP, it isn't normally used to carry voice data, but it can be used to carry call-signaling data: the bits of information that a VoIP network uses to establish, monitor, and end calls. TCP datagrams are called packets, though you often hear people refer to UDP datagrams as packets too.
IP provides both connection-oriented (TCP) and connectionless (UDP) network protocols at the transport layer, which allows it to replace both functions of the PSTN: voice transmission and call signaling.
184.108.40.206 The session, presentation, and application layers
Operating systems, end user applications, application services (like DNS), and user interfaces are provided at the topmost layers of the OSI reference model. Your interaction with a computer system or a network is most directly affected by the systems running at the application layer. The application layer's job is to take input and drive underlying functionality down through the other six layers without you, the user, having to know the details of what's going on down there.
In a VoIP network, the user interface to the telephony functionsoften just a telephone receiver with a 12-key dial-padis provided at the application layer. A VoIP- adapted OSI model is shown in Figure 2-1.
A VoIP network is a set of networked applications and endpoints ( agents that allow humans to use the applications), just as the World Wide Web is a set of networked applications and endpoints. In the case of the World Wide Web, the applications are web sites, and the endpoints are web browsers that request and display web pages. But in the case of a VoIP network, the applications are telephone calls, conference calls, voice mail, automated attendants, and even video conferencing or text messaging, while the endpoints are traditional telephones, IP phones, and software phones (softphones) that run on PCs.
Figure 2-1. OSI reference model layers
On the WWW, web sites are hosted using software like Apache, a web server. This software communicates with the endpoints (the browsers) in order to facilitate user interaction with the application (web sites).
This model works much the same with Voice over IP. In a VoIP network, specialized servers, which we'll call VoIP servers for now, communicate with the IP or traditional phones (endpoints) in order to facilitate calling (the application).
2.1.4. VoIP Servers
VoIP servers, which are software-based devices that direct or participate in Voice over IP data conversations in order to facilitate calling and other VoIP applications, are usually connected to the network using Ethernet. An ATA is itself a highly specialized VoIP server and so is a VoIP-enabled voice mail system.
Carrier-class users of VoIP might have reasons to connect VoIP servers to a different type of data link, such as ATM (asynchronous transfer mode), but most enterprise implementers will use Ethernet.
VoIP servers fulfill many telephony roles:
When VoIP endpoints and servers are connected to the same IP network, VoIP becomes the call-switching and voice-transmission mechanism, replacing the traditional PBX.
What differentiates VoIP servers from voice endpoints is whether they provide a user interface for the telephony application. Phones do, so they are endpoints. Switches, ATAs, PSTN gateway devices, and other specialized VoIP devices don't, so they are VoIP servers. Another differentiator between endpoints and servers is their abundance on the network. Like the WWW, there are more endpoints than servers in a VoIP system, sometimes even in thousands-to-one ratios.
2.1.5. Voice Endpoints
Endpoints that are TCP/IP aware (that is, they are valid hosts on the IP network) and connect directly to a data link that carries TCP/IP (such as Ethernet) are usually called IP phones. IP phones resemble feature-enriched business telephones but differ in that they usually have an RJ45 twisted-pair Ethernet connection rather than an analog or digital loop connection. IP phones can be plugged directly into an Ethernet hub or switch using an Ethernet patch cable or through a cabling distribution frame like those found in many offices. Usually, IP phones have a 10/100BaseT autonegotiating interface, much like a desktop PC Ethernet adapter.
The voice applications that run on an IP phone facilitate calling in a manner similar to a traditional phone, but the mechanics of call signaling and voice transmission are worlds apart from the old-school telephony world, as you'll discover.
Even though they don't use RJ45 twisted-pair connections or have Ethernet smarts onboard, traditional analog telephones can be connected to Ethernet, through the use of an ATA (analog telephone adapter). The ATA is a device that converts the single-pair RJ11 analog connection into a four-pair 10/100BaseT Ethernet interface, as in Figure 2-2. ATA devices tend to be less expensive than IP phones. They provide fewer telephony features, tooafter all, that old analog phone can't really run sophisticated VoIP applications, even with an ATA attached, because it doesn't contain integrated circuits or programmable components . In some cases, the limited functionality provided by a traditional analog telephone is enough.
IP telephones and ATAs are both hosts on an IP network. Like other TCP/IP hosts, they must have IP addresses and be compliant with the design of your IP network.
Figure 2-2. Analog phones are voice endpoints, and they can be used with VoIP networks in tandem with an ATA device
220.127.116.11 IP phones or traditional phones
The phones you choose are often determined by your project budgettraditional phones are often already there; it's just a matter of VoIP-enabling them. This can mean less capital investment during your transition to VoIP. But your choice of endpoints is also affected by your feature requirements: IP phones tend to have greater feature sets and programmability than do traditional phones.
Another big advantage of IP phones is that they can be completely software-based, so that they run as an application on a TCP/IP-aware Windows, Mac, or Linux PC. This can be appealing to mobile users who want to maintain a familiar look and feel on their telephone wherever they go. As long as the user has access to the Internet, the "softphone" can be made to function just like an IP "hardphone."
If you choose to integrate existing traditional phones into your VoIP network, then you must continue to support the wiring that is required for those phones. Often, this wiring is very simple, like the one-pair copper wiring used to power a simple analog phone, but not always. Most traditional PBX-type phone systems use telephones that require two pairs of copper and a digital signal busthese are called, simply enough, digital phones.
Wireless IP phones can solve the challenges of wiring, but they introduce some challenges of their own. Since wireless Ethernet does not yet offer quality-of-service mechanisms or a high yield for simultaneous use, the abundance ceiling of wireless IP phones is much lower than their wired cousins. Wiring aside, using traditional cordless phones is one way to maintain mobility without the restrictions of wireless Ethernet.
If you choose to use wired IP phones, then your Ethernet wiring system must be sufficient to support them. 100BaseT Ethernet and Category 5e cabling are considered the minimum for connecting IP phones. For this reason, exclusively using IP may not always be possible. Not all sites where you want to use IP phones have Ethernet wiring in place.
If you had an old 10Base2 Ethernet segment at a certain site, you wouldn't be able to connect any IP phones at that site, because there are no IP phones on the market that support the physical layer interfaces (BNC coax connectors) required by 10Base2 Ethernet. Moreover, 100BaseT is the only Ethernet spec that allows for the right quality-of-service required to support a large rollout of IP phones.
2.1.6. Project 2.1. Configure an IP Hardphone and the VoIP Test Network
18.104.22.168.1 What you need for this project:
IP phones are really just software applications that speak VoIP's protocols: SIP, SCCP, H.323, or MGCP for call signaling; RTP for audio transmission; and sometimes LDAP for directory integration. They may also include XML or Java services so that their displays and buttons can be used to further enhance the end user's telephony experience. When bundled together, this suite of protocol software that comprises an IP phone can run either on a PC, what we call a softphone, or on a specialized chassis whose enclosure looks like a traditional telephone, what we call a hardphone (see the previous section for more details).
Hard or soft, IP phones all require a TCP/IP stack to support data networking as well as a physical interface to the network. In a softphone, these are provided by the PC's operating system and networking hardware. In a hardphone, they are embedded more tightly and are less visible to the user.
Like a PC with a 10/100 Ethernet interface, an IP hardphone has an RJ45-compatible jack, so the first step in getting a hardphone online is connecting a patch cable between that jack and an Ethernet switch. This step will be largely the same regardless of the make and model of IP phone you're using.
Next, the IP phone must be given a TCP/IP host configuration that is workable on the network to which it's connected. To configure an IP phone for the network, you'll need:
The IP address used by the phone can be static, or it can be dynamically assigned using DHCP (Dynamic Host Configuration Protocol) if you have a DHCP server on this Ethernet segment. DHCP is not necessary in a small environment with very few IP phones to keep track of. It becomes a necessity in larger environments where an administrator mistakenly assigning the same address to two different phones can cause a disruption, just as in PC networking. For now, we'll use static addresses.
The specific steps required to configure each make of IP phone varies depending on the administrator features and firmware of each. Most permit rudimentary network configuration using the buttons on the phone itself. The Grandstream Budgetone 101 telephone is an entry-level SIP-based IP phone, and its initial configuration is done in this manner.
22.214.171.124 Configuring a Grandstream Budgetone 101 IP phone
The Budgetone 101 phone has a Menu key, two arrow keys, and an LCD display, which are used to navigate its configuration menu options: DHCP, IP Address, Subnet Mask, Router Address, DNS Server Address, TFTP Server Address, Codec Selection Order, SIP Server Address, and Firmware Versions (called Code Rel on the phone's screen). When you get to the option you want, you press the Menu key to select it, and then enter the numeric data required for each option using the keypad. Use this menu only to set up the IP address, subnet mask, and router (default gateway) address.
To get the phone enabled for the next configuration step, turn DHCP off, and assign an IP address, subnet mask, and router address.
More advanced configuration is performed using the Budgetone's built-in web configuration tool. When you access the IP address you assigned to the phone using your web browser, you'll be prompted to log in to the phone, as in Figure 2-3. The default password is "admin."
Then, you'll be confronted with a big page of configuration options like the one in Figure 2-4. Many of the options are available only through this interface, not from the phone's keypad menu. For this project, the only settings we're concerned about are the codec selection ones. Configure the first (highest-priority) codec to be "PCMU" if you're in North America or "PCMA" if you're elsewhere in the world. That's all we're going to cover about codecs for now. After applying any configuration changes, the Budgetone needs to be power cycled.
Figure 2-3. The Budgetone's web configuration login page
Figure 2-4. The Budgetone's main configuration page
Some IP phones offer a Telnet interface rather than a web-based one. To use these tools, one must connect to the phone with a Telnet client instead of a web browser. In any event, once the network configuration is set on the IP phone, ping its address from another host on the same network subnet to make sure it's speaking TCP/IP.
126.96.36.199 A simple VoIP test network
Throughout this book, a TCP/IP network is used to illustrate Voice over IP concepts through projects and hacks. This network, illustrated in Figure 2-5, is structured as follows :
Many VoIP devices need access to a time clock. The NTP (network time protocol) server we've chosen is time.nist.gov . More NTP servers are available from the list at http://www.nist.gov.
2.1.7. Project 2.2. Make an IP-to-IP Phone Call
188.8.131.52.1 What you need for this project:
With both IP phones connected to the same Ethernet switch or directly connected (to each other) using a crossover patch cable, make a note of the IP address you've
Figure 2-5. The VoIP test network at the end of Project 2.1
established for each. In this example, we'll use 10.1.1.103 for the receiver and 10.1.1.104 for the caller. If you have your phones configured for DHCP, give them this static configuration instead.
The Budgetone can place IP telephone calls from one IP endpoint directly to another without the need for a VoIP call-management server. This is called IP-to-IP calling . Since each IP phone has a uniquely identifying characteristic within the scope of the networkan IP addressone phone can call the other by IP address as if it was a phone number.
To do this, first make sure there is nothing set for User Name or SIP User in the Budgetone's configuration page. That is, make sure they are both blank, apply the changes if necessary, and then power cycle the phone.
Now, to dial by IP address. All IP addresses are 12 decimal digits long, even if leading zeros aren't written out. Conversely, the dots normally included in an IP address are not dialed. So, on the Budgetone phones, 10.1.1.103 is dialed as:
010 001 001 103
To dial, take the phone off the hook so you hear a dial-tone. Then, press the Menu key, dial the address of your second phone according to the convention shown earlier, and press the Send or Redial button. Of course, nobody would want to dial 12-digit IP addresses in order to place phone calls all the time; call-management servers, like SIP registrars, provide more elegant dialing conventions. Dialing by IP address, in this case, allows you to circumvent call management and make a direct VoIP connection between two endpoints.
When the receiving phone rings, have somebody answer the call. If you can hear the other person talk through your IP phone's handset, you've just made your first successful VoIP phone callsort of the IP equivalent of Bell's and Watson's first phone call back in 1876.
If the receiving phone doesn't ring, then you should check the IP address you dialed, check the phone's configuration to make sure it is listening on the default port for SIP (5060), and make sure SIP registration is turned off. These options, which are accessed in the Budgetone's web configuration page, will be covered in greater detail later.