Obtaining the Source Files | Installing Asterisk

The very first step we must undertake is to obtain the source files. When obtaining the source code, we have two major choices. We can either download the latest version through Digium's web site from http://www.asterisk.org/downloads or use svn to obtain the latest stable release. The maintainers of Asterisk have been doing a good job of keeping the stable releases available on the FTP servers, so we will use this method.

The commands we issue to download Asterisk's source files are:

# mkdir -p /usr/src/asterisk

# cd /usr/src/asterisk

# wget http://downloads.asterisk.org/pub/telephony/asterisk/asterisk-
1.6.1-current.tar.gz

# wget http://downloads.asterisk.org/pub/telephony/asterisk/asterisk-
addons -1.6.1-current.tar.gz

# wget http://downloads.asterisk.org/pub/telephony/dahdi-linux-complete/
dahdi-1inux-complete-current.tar.gz

# wget http://downloads.digium.com/pub/libpri/libpri-1.4-current.tar.gz

The download may take anywhere from about one minute on an extremely fast connection, to a couple of hours on a slow connection. When the download is complete, we will need to unpack the tarballs.

For unpacking the tarballs, you will want to issue a tar -xvzf [name of file] (without the brackets):

# tar -zxf dahdi-linux-complete-current.tar.gz
# tar -zxf asterisk-1.6.1-current.tar.gz
# tar -zxf libpri-1.4-current.tar.gz
# tar -zxf asterisk-addons-1.6.1-current.tar.gz

In the next three sections, we'll compile and install the source distributions we've just downloaded. Note that we should install DAHDI first, then LibPRI, and finally Asterisk.

Preparing to Install Asterisk

In order to install Asterisk, we will need a computer with Linux installed. It's a good idea to ensure your system is up-to-date, for instance, by using the yum tool. Once we have installed our distribution of choice, we need to make sure we have a few additional packages installed. The required extra packages over a base installation are:

• Bison, and associated -devel (1.0.X only) gcc
• Kernel-source
• Libtermcap-devel
• ncurses, and associated -developenssl, and associated -develzlib, and associated -devel

Once we have installed these packages, we are ready to install Asterisk. We should not run an X server or any windowing software on our Asterisk machine, as the resources it consumes are almost guaranteed to delay our voice processing, and therefore negatively impact our sound quality. So, you may save a little time and disk space by choosing not to install any such frontend.

One note here—we should prepare to manage our server. We must keep in mind that we will not be able to rely on graphical tools on the server to manage users, file systems, and other aspects of the day-to-day maintenance that all the systems will need. Unless particularly comfortable with command-line configuration, you should probably consider installing a web-based set of tools such as Webmin, available at www.webmin.com in order to configure Linux. The graphical configuration options for Asterisk that are available are mostly web based. So, we may at some point decide to install these under a web server too, in order to enable graphical configuration.

The commands used to install Webmin are as follows:

# wget http://voxel.dl.sourceforge.net/sourceforge/webadmin/webmin-1.470-
1.noarch.rpm

# rpm -U webmin-1.470-1.noarch.rpm

Once webmin is installed, the default address to access webmin from your browser is https://<IP of Asterisk server>:10000.

Choosing the Extension Length

While creating our phone system, we will need to create a set of extensions. Although Asterisk has no such requirement, all these extensions should probably have the same length to give comfort to our users. We must determine the length that we will use for all of our extensions.

When creating extensions, it is often advantageous to group certain extensions together. For example, all sales extensions could be in 200s, support in 300s, management in 100s, and so on. Or we could go further and say that all first-tier support will be in 3100s, the second-tier support in 3200s, third-tier support in 3300s, and so on.

We should keep in mind that it is easier to add extensions when there is an available number than it is to renumber all extensions in a building, because we have filled up all of our available dial strings. For instance, suppose we chose 1-digit extensions and have the following phone list:

0—Operator

1—Reception desk

2—Break room

3—Conference room

4—John

5—Sally

6—Jennifer

7—Fax machine

8—Voicemail Access

9— Outgoing calls

This system will work fine until we add another extension. When we add another extension, we will have to give new extension numbers to all of our users.

Now consider the following phone list:

1000000—Kitchen

2000000—Bedroom

3000000—Office

8000000—Voicemail

In this house, for someone in the kitchen to call the office (think "Dinner's ready, will you please leave that computer and come eat?!"), the user has to dial seven digits to accomplish what could have been done with one.

Therefore, we need to be smart about how long we make our extensions. Often, if we are replacing a phone system, we should just adopt the numbering already in place to make the transition a little easier for our users. Some phone systems may not have had extension numbers before, such as old analog key systems. All lines were simply visible from all stations. In such instances, we should be sensitive to the new learning that will have to take place and make the length of the extension number as small as possible.

We also need to consider some special instances. First, most people do not want an extension that begins with a 0. Simply put, nobody likes to be a nothing and having a leading 0 for anybody but the operator makes them feel emotionally put down. Also, we should reserve all extensions beginning with a 9 as outgoing telephone calls. Add to that the need to provide services such as call recording, conferencing, and voicemail access. We will give all such services a prefix, such as 8. Thus, we see that we have already lost 30 percent of all of the available extensions.

A good rule of thumb in computing is to take what we believe we will use and triple it, and then round up. Thus, if we believe that at the height of our system, we will have 100 users, we should assume that we will have 300 users. If we believe we will never have more than 10 users, we should assume 30.

With this in mind, here is a table of what we will need:

Expected number of extensions	Our assumed number of extensions	Length to use for extensions
2	7	1
22	70	2
222	700	3
2222	7000	4

Keep in mind when reading this chart that it is much easier to have people dial an extra digit than it is to make them learn all new extension numbers. Thus, if we are a border case, we should go ahead and move on up to the next extension length.

Another idea that we can take advantage of is using an extension that gives a lot of information about the destination. Take for instance, a corporation with seven locations. The first digit in the extension could designate the location. Then the second digit can designate the department, and the remaining digit(s) can designate which member of the group is sought. Thus, knowing the structure and an extension can give an idea of where that person is and what he or she does.

In some environments, such information is not desirable. For instance, in a college campus, some employees work very late at night. If the extension gives their precise location, stalking and threats of physical harm can prove problematic. Therefore, we need to be sensitive to such concerns.

One alternative to these layouts is to use the last few digits of a phone number to refer to each extension. This can work very well if all of such digit strings are unique. However, it can cause problems. Suppose we chose a 4-digit extension and have the phone numbers 555-1234 and 777-1234. Which one is extension 1234? Or suppose we use 7 digit extensions and have (800) 555-1234 and (866) 555-1234. Which one is extension 5551234? Thus, some organizations have moved to a full 10-digit extension length. While this allows 10¹⁰ extensions, it can cause some users to complain about usability and convenience.

With the flexibility of Asterisk, we can choose many different ways to allocate extensions, all of which will influence our decision on extension length. We must balance our users’ expectations along with our desire to leave room to grow. By doing so, we can create extensions that are easy to maintain and user friendly.

How Much Hardware do I Need? | Asterisk 1.6

This is probably one of the questions most frequently asked by those who are new to the world of Asterisk. The answer depends largely on what we are going to do with our system.

Conversations that bridge between codecs (called transcoding) take maximum power to handle. Voice over IP conversations seem to take a little more processing power than straight Time-Division Multiple-Access (TDM) calls. Having our server run scripts to find information will take more power than if we define everything statically. How many different conversations we have going at a time will affect how much horsepower we need our server to have as well as the features we use.

Do you see the complexity of answering this question? We have to figure out what we are going to use before we can figure out how big a server we will need. That said, there are some good rules of thumb we can start off with.

First, while we can run an Asterisk server on an old Pentium 90 with 64 MB of RAM, why would we want to? We are creating a robust phone system. We do not have to pay to license the use of the software and we do not have to pay per extension. We can go spend some of the money we saved and buy a decently powerful server. Most would recommend that a small deployment should have a CPU of at least 2.4 GHz and 512 MB RAM; the hard drive space is not as important but typically 120 GB would suffice. The hard drive space greatly depends on how many voicemail messages you want to allow users to store on the server as well as whether you want to record incoming or outgoing conversations. Voicemails and recordings are stored on the server, and without limitation or careful planning you can run out of space. For larger deployments you might want to get a Dual CPU solution as well as an increase in RAM (that is, 2 to 3 GB RAM). As we select the components for our server, we need to remember that we are not building an email server or a web server. We are creating a PBX that people are going to expect to be running all the time. We should select a stable chipset, with an up-to-date BIOS, and match it with other current high-quality components. By using high-quality components, we increase the likelihood of ending up with a high-availability phone system.

On another note, we should select a server with as much redundancy as possible. A RAID-1 controller could save our phone system in the event of a hard drive failure. A pair of RAID-1 controllers that are mirrored could save our phone system in the event of a controller failure or a PCI slot failure. A server with redundant power supplies will help us in the event of power failure or a power supply failure. Of course, our phone system should be on an Uninterrupted Power Supply (UPS). This is not only for protection from power failures; it will also protect from spikes, and often even lightning.

Depending on the reliability requirements, we might need a redundant server. There are hardware devices that will detect if a PRI is down, and automatically failover. Then again, for most installations, this is overkill.

The most important lesson to keep in mind is that people have grown to depend on phone systems. We should not skimp on hardware, as doing so could cost us dearly in the long run. With the unique pricing structure of Asterisk, all we will have to pay for is any additional hardware to get increased reliability and capacity.

Along with hardware, the question often asked is—which distribution of Linux should I use? If you already have experience with some distribution of Linux, you should be able to make Asterisk work with that distribution. Asterisk is very flexible and has been built with commonly available dependencies, and any distribution of Linux should work. That said, some distributions will require more effort to enable some features such as automatically starting Asterisk when the server boots. As each distribution treats startup scripts differently, most distributions will require a minor amount of tweaking.

Also check the wiki at http://www.voip-info.org for more information on the distribution you intend to use. It has up-to-date notes on compatibility problems, caveats, known issues, and often workarounds for those issues.

Choosing a Device | Asterisk 1.6

Now that we have seen the broad offerings of terminal devices, we will see how difficult it can be to choose one to meet our needs. After choosing a type of device, we have to choose a manufacturer and model. This task can be daunting. Let's take a few minutes and discuss how we will make the best decision based on the available information.

Features, Features, and More Features…

As we review available phone handsets, we will be inundated with all the features that manufacturers can throw at us. These lists are overwhelming, even to the most seasoned experts. It is very difficult to compare two handsets solely on features, as some features have different names.

Determining the usability of a particular phone handset should be a straightforward process. This process has four major steps—requirement elicitation, prioritization, and documentation, followed by handset testing.

Requirement Elicitation

This is the brainstorming step. We should go to each user and determine what his or her needs are. We ask the user what features he or she uses on the current phone. We observe the person working for a period of time to get a good sampling of what he or she actually does.

We should then go to the user's manager and see what a person in that position is expected to do. We add these features to our list. While this list will be unique to each user, many will be very similar. We should see patterns of usage emerge between groups of employees.

Requirement Prioritization

In this step, we take our requirements list from the previous step and, working with the user and manager, determine which features are used most, which are most important to that user's role in the organization, and which features are simply nice to have. We should also attempt to recognize any deficiencies in the current technology. Changes are often embraced if the change adds value to the user by making a task easier or in some cases removing a task entirely. It's important that we recognize all nuances of the current system in order to provide the user with a replacement that will suit them.

We should then create a quantitative scale for each feature. For example, if we were working with an operator, a Transfer button would be a value of 10, while a Do Not Disturb button will probably be a value of 1. If we had a phone handset with both of these abilities, then we add these scores together and it would score an 11. By putting numbers on the required features, we can come up with a quantitative answer to a very subjective issue.

Requirement Documentation

This step is the most important of all of the steps so far, especially for consultants. We take the list of requirements and their weights, and write them in a short document. We then have the user and the manager sign it to indicate their agreement.

This may seem a little formal for picking a telephone handset, but it is an effective method of communicating expectations and plans between you, the implementer, and the users. This can help in preventing surprises or differing recollections of what was promised.

Phone Testing

This is the final step. After comparing the available handsets against the document we created in the previous step, we choose the highest scoring handset. We then take a handset of that type to the user and have him or her use it (if we have a test system installed by this point) or at least sign off on it conceptually.

Again, this is an opportunity to ensure our users’ expectations are reasonable, that commitments are clearly defined, and that our users are kept informed during the decision-making process. It can also help us get a buy-in from the users as we make the major adjustments that will invariably accompany a new phone system.

Determining True Cost

When we look at which handsets to compare our requirements document with, the issue of cost also will have to be looked at. Before we offer a handset that would not be possible under our project budget, we should determine that the handset meets all of the requirements of the business, which includes the element of cost.

The issue of cost is not as simple as looking at the retail price of a handset. Each type of phone will have multiple types of cost. These costs will usually fall into one of the following categories:

• Handset cost: This is the easiest cost to determine. It is the actual amount of money that will have to be spent to acquire the telephone.
• Port cost: This cost is the element of what the phone connects to on the other end. For instance, on a VoIP phone this could be a portion of the cost of a new network switch that supports Quality of Service (QoS) to enable reliable voice communications.
• Headset cost: If a phone will require a headset, then we should consider the cost of that headset as we choose the phone. Different connectors are available depending on the model.
• Software license cost: Some phones will require the purchase of G.729 license. Other phones may require a license for the software on the phone (usually referred to as firmware). We should not fail to consider this cost while computing the cost of the phone.
• Installation cost: The time required to install different phones differs. This time translates into cost.

By considering each of these factors for each different handset, we get an idea about the true cost of each particular phone. With all of these costs defined, we can see which phones are within our budget and which are simply too expensive.

Compatibility with Asterisk

Not all handsets interoperate equally with Asterisk. Referring to the Asterisk Users Mailing List archive, we can ensure that no serious incompatibilities have been discovered. Also, a Wiki is available at http://www.voip-info.org. A vast array of useful information about Asterisk is available there. This site is searchable and is constantly updated.

We do not have to select a single protocol for all VoIP phones. Instead, we can mix and match protocols to our best advantage, thanks to the flexibility and power of Asterisk.

Sound Quality Analysis

Sound quality is a very subjective thing. Each user will have a personal threshold between acceptable and unacceptable.

Each phone will have a varying sound quality. The variables that can affect the quality of a call are staggering. Network latency can significantly affect sound quality, but so can configurations of the phone. Determining what the cause of low sound quality is can be difficult to do.

Build quality from a manufacturer can also affect quality. When wide variations are allowed from one phone to the next, the result is usually inconsistent from handset to handset. Thus, we have to choose a manufacturer we can trust.

While there is no absolute, the quality of sound on telephone handsets, from highest to lowest, is usually as follows—analog hard phone, VoIP hard phone, analog soft phone, VoIP soft phone. If you are doing a comparison between different handsets, the main things to pay attention to are the amount of background noise (or hiss), distortion, drop outs, popping, and highly digitized voice. If we have users who are extremely sensitive to sound quality, analog will probably be our best bet. For those users who are a little more forgiving, VoIP allows us to use one network for our phones and computers.

When determining what terminal equipment to use, we need to consider the sound quality of each device and match it against the needs and expectations of our users, and weigh that with the cost of that device as compared to the budget.

Usability Issues

The world's most advanced VoIP handset is absolutely useless if our users cannot figure out how to use it. As we decide what equipment to provide for our users, we should consider where they are at in the continuum of technological awareness. While VoIP hard phones with context-sensitive buttons are useful for most users, some people find the interface confusing and frustrating.

This is one big issue that we need to address in the handset testing that we do after eliciting the requirements that our user has for a new phone. We have a duty to ensure that our users can use the handsets we choose. We must be careful not to assume that they will figure it out, as doing so often causes hurt feelings and resistance to change. The success of Asterisk will be largely measured by the response of our users.

Analog Adapters | Types of Terminal Devices

Dedicated communication devices such as modems and fax machines are still very prevalent in business today. Although these devices could be replaced with more modern, reliable, and faster technologies, the new technologies have not yet been embraced.

Therefore, in order to use these existing technologies, analog adapters allow companies to continue using legacy devices such as traditional fax machines. An analog adapter usually consists of an Ethernet jack with which the device will connect to your network and an FXS port (telephone jack) which will connect to your traditional communication device (such as a fax machine).

Another common use of an analog adapter is to connect a cordless phone for those requiring mobility around the office. Granted there are WIFI SIP phones in the market, often users will find that the WIFI signals interfere with other frequency emitting devices. Such interference can cause distortion or the calls to drop.

One issue with analog adapters is their use with traditional fax machines. For years the reliability of faxing over an adapter was poor at best. Today analog adapters are becoming equipped with T.38 capability. This protocol allows regular faxes to be sent over UDP. The use of T.38 dramatically improves the reliability of faxing. However, keep in mind—your VSP as well as your PBX must also support T.38 in order for the fax to be transmitted using this protocol. As of version 1.4, Asterisk now supports T.38 negotiation for SIP users and the related pass through of UDPTL T.38 data. Please note that Asterisk currently cannot terminate T.38 calls or act as a T.38 PSTN gateway without external support.

One extra note about faxing—Asterisk supports receiving and sending faxes through an add-on called SpanDSP. With this Asterisk can receive a fax and turn it into a TIFF file. This TIFF file can then be further processed to become a PostScript or PDF file and be emailed to the proper recipient. Another notable fax detection solution is NVFaxDetect. The installation of this add-on is not covered here, as it is changing rapidly.

These communication devices are usually supported for legacy reasons. We should continually strive to reduce outdated technologies and replace them with up-to-date solutions.

Another PBX

We can connect PBXs together to provide services to users hosted on another PBX. We can use SIP, PRI, T1, H.323, or IAX to connect the PBXs.

If we are connecting multiple Asterisk PBXs, we should use IAX. The IAX protocol has a number of features with this specific use in mind, such as the ability to have multiple conversations trunked into the same UDP stream yielding greater efficiency.

Soft Phones | Types of Terminal Devices

Much as hard phones are phones implemented in hardware, soft phones are phones implemented in software. Using all the same protocols available to hard phones, soft phones are far less expensive to implement. By using the general-purpose computing resources of a personal computer, the expensive proposition of replacing all telephones in a building can be avoided.

Before going further, we should recognize that most hard phones are in reality soft phones combined with bit of special-purpose hardware. The computing power of a hard phone is not as vast as that of a PC, and unlike a PC, is specially tuned for carrying voice. Thus, we should not dismiss the use of hard phones immediately.

The sound quality experienced on a soft phone will depend greatly on the available resources on the PC, the quality of the software used, and the quality of the data network between the PC and our Asterisk server.

Soft phones will have a hard time being accepted by some users. In addition to the political issue of having people use their computer to talk on the phone, we also have to address disaster planning. If we lose power, keeping a computer up that draws in excess of 400 watts will be far more difficult and costly than keeping power to a hard phone that draws 15 watts, especially for prolonged outages.

The most significant advantage of the soft phone is cost and portability. In most businesses, desks contain a computer and phone at least. If you can remove the phone there is an obvious reduction in hardware costs. There are a variety of soft phone products available and most operating systems come with a basic soft phone package by default. Also the portability aspect of a softphone can be very appealing to companies that have employees who travel a lot. Imagine you're staying in a hotel that offers high speed access. Simply open your laptop, put on your headset, and you're able to make as well as receive calls as if you were in the office. There are also a variety of open source products available. Some companies such as Counterpath (www.counterpath.com) and Zoiper (www.zoiper.com) offer free clients for Windows, Linux, and Mac.

If you decide to go with a softphone also be sure to invest in a decent headset with noise cancellation. Headsets that have their own DSP and are USB driven are a good choice. This removes most audio processing resources from the PC so that other work done on the PC does not affect voice quality. The choice of product, soft or hard, is equally important as the PBX. You must be sure that the users will use the device and be sure that it will be reliable and supportable.