VoIP: Lets Make it Happen

 

Kevin McMahon

Final Term Paper for:

MSIT 640

Instructor: Jack Ligon

Due: April 28, 2002

 


Abstract

            Voice over Internet Protocol (VoIP) is one of the more exciting features that could become available on the Internet.  It allows the Internet infrastructure to replace the Public Switched Telephone Network (PSTN) for the use of telephone-based communications.  Currently, the technology required for VoIP has been created but is not yet readily available and in place for VoIP to reach widespread use.  We will examine why VoIP is a good idea as well as what is necessary to make it viable and available for everyone; so that everyone can take advantage of the many benefits.

 

Introduction

            Currently there are two major networks, with some overlap, in the United States.  These are the PSTN (Public Switched Telephone Network) and the large collection of networks known as “The Internet”.  The PSTN is a circuit-based network and the Internet is a packet-based network.  One of the connections between them is known as “The Last Mile.”

 

The Public Switched Telephone Network

            The PSTN has 3 functions.  Its original design was to facilitate telephone communications between 2 individuals; this it does and it does well.  Also, parts of the Internet are interconnected using leased lines that are part of the PSTN; this is the overlap mentioned previously.  Finally, for many people, the PSTN still provides the “Last Mile” connection from their home to the Internet. 

            Telephone calls are made up of two consistent data flows, one in each direction.  The data, in this case, is the voice (and other background noise) of the two people involved in the call.  The nature of voice data is such that it requires a constant amount of bandwidth to transmit; from the start of the call to the finish of the call, and from one call to another, the data rate requirements do not change.  This consistency is why the PSTN was designed as a circuit based system.  Here, a connection is set up from point A to point B that connects the two users, providing exactly the data requirements.  As soon as the connection is no longer needed, it is terminated and the resources that were allotted to the connection are put back on the available stack for other callers to use.  The underlying network knows the requirements for each call, and it knows what resources it has available.  Whenever a phone number is dialed, the required resources are compared to the available resources and if there is enough available, the call goes through; if the resources are not available, a “circuits busy” message is returned.

Originally the voice data was totally analog; variations in current traveled over wires where it was reconverted into sound.  In today’s world, the voice data leaves the user’s house in analog form and is sampled and quantized into digital format at the telephone office.  It is then multiplexed and transmitted over high-speed data lines where it is de-multiplexed, reconverted to an analog signal and transmitted over the wires to the end user.  The multiplexing process involves breaking the continuous call data into packets; these packets are then placed on the high-speed lines at specific intervals.  As mentioned above, the data rate required for voice data is consistent; since we are now breaking our stream of data into packets, we are producing a constant amount of packets every second.  Since we know ahead of time how many packets per second to expect we can allot the correct amount of “slots” on our multiplexed conveyor belt.  The number of packets per seconds will neither increase nor decrease.

 

The Internet

The Internet is an example of a Packet-Switched network.  In a packet-switched network, unlike the PSTN, there is no physical circuit set up between two users that wish to communicate (although a virtual circuit may be setup).  The Internet is comprised of many different proprietary networks, some large and some small.  These networks, often of different types, communicate using Internet Protocol (IP).  The users in this type of network communicate using a system similar to that of the postal service.  Each message to be sent is first broken up into many smaller packets.  Each packet is then labeled with a “to” and a “from” address.  Each packet traverses the network alone, and is then re-assembled at the receiving end.

Postal addresses are broken into several parts: the recipient, the street address, and the city, state and zip code.  Each part is successively narrower in scope and thus plays a different role in getting the letter to its destination.  First, the zip code (which is really a numeric translation of the city and state) is examined so that the closest post office to the address can be determined and the letter is sent there.  Once the letter is at the local post office, the street address is examined and it is placed on a truck and delivered to the destination address.  Once the letter has arrived at its destination address, the recipient name designation determines to which occupant the letter belongs (of the many possible occupants at that address). 

Much like mailboxes, all connections to an IP network are designated by their IP address.  The IP address works in a similar manner as a postal address; it starts off general and gets successively more specific until it identifies the specific user.  When one computer needs to send data to another computer over the Internet, it must encapsulate that data into an IP packet (other network specific packets may be inside the IP Packet to be unpacked by the destination network).  Once this IP packet reaches the nearest router (like a post office) the router examines the network ID portion of the destination IP address to determine if that router is directly attached to the network to which the packet is destined.  If it is, then the rest of the address can be resolved to determine which local address the packet needs to be forwarded to.  If it is not directly attached to the destination host, then a chart must be consulted to determine the next “hop” on the packet’s journey.  The packet will then be forwarded to the next hop, which is simply one router closer to a router connected to the destination host (Leon-Garcia & Widjaja, 2000).

            Unlike the data traveling over the PSTN, the data traveling across the Internet is not consistent.  Packets are generally not created in a consistent manner; a user will usually request a web page, and then read that page before requesting another.  This allows for downtime between transmissions when no data belonging to that user is on the network; with the PSTN even when the users are not talking, the same number of packets are still traversing the network to be translated at the other end as dead air.  The PSTN, traditionally, (there is now three-way calling) only allows two users to occupy the circuit at one time, but since the Internet is not circuit based, packets from one user can be traversing the network to an almost unlimited number of destinations. 

 

The Last Mile

The “last mile” is the connection between an Internet Service Provider (ISP) and the user’s home.  Currently the most common connection for access to the Internet is a 56K modem, which transmits data over the PSTN until it reaches the ISP and is then on the Internet.  The two main problems with this last mile connection are wasted bandwidth and data rate.  Even though the backbone of the PSTN is comprised of very high-speed fiber connections, the lines between the user’s home phone and the phone company office are still largely unchanged; it is still comprised of 2 copper wires.  These wires form a circuit connection with a constant conveyor belt for data.  Unfortunately since the copper wires are not high quality this provides a very small data rate.  With today’s emphasis on multimedia this creates a problem.  Also, the constant conveyor wastes bandwidth because Internet data is very bursty; there are periods of no data transfer followed by periods of rapid data transfer.  During the slow periods, the user is still paying for the whole bandwidth even though it is not being used.

 

Advantages of VoIP

One Network

            As we have seen above, the two major networks both serve their function well.  Also, as we can see from the description of the last mile, many people still use the PSTN as an access point, or gateway, to the Internet, but this is changing.  Cable providers and satellite providers are now offering high-speed access to the Internet.  Eventually, it is conceivable if not inevitable, that everyone will have some form of high-speed connection to the Internet that does not rely on the PSTN.  This will result in the emergence of two autonomous networks, where neither one is relying on the other.  If and when this becomes the case, the question will be: Why do we need two networks?  Can one of these networks, and its associated costs, be eliminated?  The answer is yes.

            We have already seen the Internet grow up alongside the PSTN.  If the PSTN could have performed the services of the Internet well, the Internet would not have been formed.  It is clear, then, that if one network were to replace the other, it would have to be the Internet.  The main function of the PSTN is to provide real-time voice communications; this is what Voice over Internet Protocol (VoIP) technology provides utilizing the Internet. 

            Having only one integrated network would have many advantages.  Right off the top, the infrastructure requirement is reduced.  Instead of having a separate voice and data connection to every house or business we could manage by having only one high-speed connection that served all purposes.  Assuming the same connection carried the user’s television (cable or satellite) it would be the only communication wire(s) that would need to be laid.  This would make it easier to build new apartment buildings and office buildings; if the end user desired a satellite connection, like DirecTV, no communications wires at all would need to be connected to their residence. 

            Having one unified network could also improve e-commerce.  Currently when you are looking at an online catalog and have some question you have to email customer service, then wait.  The return time for a question like this, though much better than traditional postal mail, can be painfully slow.  Once you finally get a response, if you are still interested, you have to return to the site and find the item again; this is terribly inefficient.  If there was a convergence of voice and web applications, you could simply click a link on the website and be instantly connected with someone who can quickly resolve your question and complete the sale (Cyganski & Orr, 2001).

            Although the one network scheme is not currently available, some private companies have combined their data and voice networks into a single network to take advantage of VoIP within their proprietary network.  This can provide significant cost savings.  Most medium to large companies have a phone network outside of the PSTN on their property; they will have some extension based system to ease calling within the company and to direct external calls.  One company, Bayley Construction, which has two main offices in Washington and California, is a good example.  They found it necessary to replace their old phone system, which no longer suited their needs; they decided to combine their voice and data into one network using VoIP technology.  By utilizing the leased T-1 lines that interconnect their two offices they achieved “significant cost savings” (VoIP builds, 2000).  Inter-office communications formerly required a long distance call; with the new system they are saving between two and three thousand dollars per month in long distance alone.  They have also reduced the costly overhead of maintaining two separate networks.  This system also allows them to more easily communicate with customers since a single 800 number can connect to both offices as if they were in the same building (VoIP builds, 2000).

           

Long Distance Charges

            As we saw in the Bayley example, by using their proprietary network to communicate, they eliminated inter-office long distance charges.  They managed to save over two thousand dollars per month on long distance; the larger the company is, the more potential savings they have by combining their networks.  VoIP has even greater potential over the public Internet.  With a network convergence you would no longer have to pay the phone company as well as an ISP to keep you connected to both networks; although it’s possible that the price for the single network connection will be a little higher than either separate connection is now.  Once you are connected though, whether your call is to the computer next door, or to a computer in Japan, there is no long distance charge.  When you are surfing the web, it does not cost you more to download a web page from a server in Australia than it does to download a page from a server in New York.  This same network feature could make keeping in touch with loved ones a much easier and less costly experience.

 

Requirements of Voice Data

            Voice data does not have the same requirements as the normal packet data that the Internet grew around; web pages and photographs do not have a real-time component.  The Internet was designed as a “best-effort service” (Leon-Garcia & Widjaja 2000).  In this type of system all packets are treated equally and each packet is attempted, but not guaranteed, to be promptly delivered.  This is not the best system for voice data because the requirements for quality voice communications are very stringent.

 

Lost Packets

Lost packets are not generally a problem for web page data because a missing packet can easily be retransmitted.  There is no real-time requirement for most of the data found on the Internet.  Voice communication does have real-time requirements; delays in transmission, such as waiting for a retransmitted packet, can make it very difficult to have a conversation.  A delay as short as 1/6th of a second can confuse both parties and distort the intended meaning of what was said (Cyganski & Orr, 2001).  Every packet of data, then, has 1/6th of a second to arrive and be decoded at the receiver from the time that it was created.  This, generally, does not allow for retransmission of lost packets.  Luckily voice data can survive with reasonably good quality even in the face of some lost packets.  Though if the number of lost packets gets too high, then it will distort the reproduced sound.  As a best effort service, IP treats all packets equally; this can cause packet loss problems.  When one router gets too busy along the route between the users, it will begin dropping packets so that it can handle others; depending on how busy the router is, too many of the voice packets could be lost.

 

Packet Delay

            There are two types of delay to which packet data is subject: variable rate and fixed rate.  The fixed delay is the minimum amount of time it takes for a packet to move from one point to another on the network.  This consists of the minimum processing time at each router, plus the transmission time along the physical medium (which is limited to the speed of light).  The processing time at each router is required because each router must examine every packet for its addressing information and then determine where the packet needs to be forwarded. 

            The variable delay is due to the load on each router.  When each router is only handling a small amount of packets, then each packet can be quickly examined and forwarded; unfortunately the routers get busy sometimes.  When a router is heavily loaded it has to store some of the packets in a buffer until it has time to examine them and forward them.  This can lead to additional delay (which will depend on how busy each router along the path is) and can also lead to the packet loss described above.

 

Constant Rate

            As mentioned earlier, voice data is created at a constant rate; therefore it must be decoded at the same constant rate otherwise it will not sound correct.  There are ways that we can minimize variable delay, but there will always be some amount of variable delay when using the Internet.  Since our voice data needs to be decoded at a constant rate we must employ a technique to ensure that this occurs.  The method most used is called buffering. 

When we want to create a buffer, the first X number of packets received are not processed immediately.  Once we have a buffer of X packets to draw from, we draw packets at a constant rate and decode the data to recreate the transmitted sound.  Sometimes the data arriving in the buffer will arrive fast and sometimes it will arrive slowly, but as long as we remove packets from the buffer at a constant rate we have met the requirement for constancy.  The challenge here is choosing a buffer size.  If the buffer is not large enough, then it might empty completely while the data is arriving slowly and then we will have no packets to process; packets that arrive too late are treated as lost packets.  On the other hand, if we make the buffer too large we are still hurting ourselves because the buffer adds an additional fixed amount of delay.  If this delay is too large, as seen above, the speech can become confusing.

 

How can we meet these requirements?

IPv4

As we have seen, packet loss and variable delay are inevitable in the current Internet system.  When discussing IP so far we have been referring to Internet Protocol version 4 (IPv4), which is the foundation of the Internet, as we know it today.  This protocol treats all packets as equal so it will indiscriminately buffer and/or discard packets when a router gets overloaded.  When a packet is buffered due to network congestion, it creates additional delay that may degrade the integrity of the voice communication.  If the router is forced to drop a packet, a retransmission will likely not be possible given the time requirements and these lost packets also can degrade our voice communication.

 

IPv6

IPv4, the current standard, clearly creates some problems for VoIP in the way that it handles packets.  Internet Protocol version 6 (IPv6) is designed, among other things, to provide the quality of service (QoS) protocols that will pave the way for widespread VoIP; QoS refers to parameters such as packet loss, packet delay and error rate.  IPv6 employs a more complex header system than does IPv4.  The header includes a 24-bit flow label that, combined with the source and destination addresses, provides for a unique label for every flow on the Internet.  According to Robert Hinden, “A flow is a sequence of packets sent from a particular source to a particular (unicast or multicast) destination for which the source desires special handling by the intervening routers” (Hinden, 1995). 

Each packet originating from a given source can also give a priority to its packets using a 4-bit priority field.  The priority field tells each router how sensitive to delay the packet is and therefore, how much preference it should be given over other packets from the same source (Hinden, 1995).  For example, a low priority packet belonging to an email message should be discarded when the router gets overloaded, while the high priority VoIP packet should be protected if at all possible. 

 

RSVP

The flow label in IPv6 allows other protocols to be used to enhance the reliability of IPv6.  One example would be a resource reservation protocol such as RSVP.  RSVP is a protocol that negotiates with the routers along the path between two users.  It first announces what characteristics the data flow has, such as data rate, tolerance to delay and tolerance to lost packets.  It then requests that the routers along the path “reserve” enough resources to support the flow.  If all of the routers have enough resources left to support the flow then the connection will be allowed and the reservations will be made; if there is not enough resources left at any one router a message similar to a PSTN “all circuits busy” message will be sent back to the requester (Leon-Garcia & Widjaja, 2000). 

Once a flow has been established each router must monitor the flow to determine if it is staying within the negotiated parameters.  If the flow is creating packets at a rate higher than it originally requested, that flow can be flagged.  As long as network resources are plentiful, that flow will still receive its requested QoS.  When a router becomes overloaded, though, packets belonging to a flow that has been flagged are the first packets to be discarded. 

 

Faster Routing (decreased fixed delay)

The flow designator can also act to speed up the routing of packets.  The IP addressing scheme is very robust.  It is similar to mail addressing in that “if we plucked a packet from anywhere in the Internet, by examining this address we could determine its source and destination. And, if we dropped this packet back into the Internet anywhere else, it would still be routed to the correct destination” (Cyganski & Orr, 2001).  This robustness, though, also acts to create some of the packet delivery delay; every packet must be examined at every router to determine its destination and then processed to determine the next hop in its journey.  This creates overhead because the router has to compare the destination address to a rather large routing table to decide which next hop is the most optimal.  Using the flow descriptor and source address can speed up the delivery process because the router can store in cache (very fast memory) the specifics for each flow to which it has committed.  When a packet arrives with a flow descriptor it can compare this packet to its smaller, faster cache chart to determine where to forward the packet; this can be considerably faster than comparing the IP addresses to its routing tables (Hinden, 1995).

 

Why Aren’t We All Using VoIP Now?

Impractical use of IPv4

            Although IPv4 does not allow for QoS specifications it is still possible to use VoIP on the Internet.  Unfortunately the phone (or computer) call that is made will not have the same clarity that people have become accustomed to when using the PSTN. This is due to the packet loss and the delays that are caused by network congestion.  A simple solution to this would be to simply increase the bandwidth of the backbone of the Internet; given an unlimited bandwidth VoIP will work even without QoS because there will be no congestion.  Unfortunately there are two problems with this idea.  The first is money; it is very expensive to add extra routers, especially when they are in excess of what is needed.  These extra routers and bandwidth would only be used during the peak Internet times; the rest of the time they would be wasted.  The other problem lies in the practicality of trying to stay way ahead of network usage.  No matter how many routers are added, eventually it will become congested again; it is a vicious cycle that can doubtfully be won.

 

Slow Transition to IPv6

It appears to be impractical to attempt widespread VoIP over the current Internet, which relies on IPv4.  Our solution then is clear; we need to move to the next generation protocol, IPv6.  Why then have we not moved over to the newer system yet?  The easiest answer is: money.  It is expensive to update to a new system such as IPv6.  There are a lot of routers and other physical equipment that would have to be replaced.  Also, a lot of equipment that does not have to be replaced would need to be reconfigured; this will require a very large number of man-hours to implement.  One of the other features of IPv6 is that it allows for a much larger address space by lengthening the bits allotted per address; this is part of the problem with reconfiguring to an IPv6 network (Demaria, 2002). 

Software is another reason that IPv6 has not yet taken over.  There are innumerable applications in use today that access the Internet.  All of these applications are looking for the standard 32-bit IP address that is defined in IPv4.  All of these programs will have to be rewritten to expect the new 128-bit IP address.  This also can be a huge undertaking; for example, changing a single constant in a header file “requires finding the source code, rewriting, recompiling, testing and distributing that code” (Demaria, 2002). 

Although the standard for IPv6 has been finalized, there is still not an excess of equipment ready to use it.  There are routers out that are designed for IPv6 but not to the same extent as there are for IPv4.  In fact, Microsoft and Cisco (two of the largest computer giants) have yet to release a stable production of IPv6 stacks.  Also, many vendors do not fully support IPv6 because of the programming issues mentioned above (Demaria, 2002). 

 

Conclusions

Create a Demand

            A good VoIP program has a chance to be the “killer app” of the new millennium, much like the word processor and the spreadsheet were near the dawn of the Personal Computer.  There is a demand for this type of service, though currently not large enough to force a quick change.  If a company were to produce a quality VoIP product and advertise it well, it could greatly increase the demand for this service.  If the demand reaches a high enough point, then the “powers that be” will have no choice but to go through the hassle of transitioning to IPv6; the market dictates what the business environment must do. 

The Last Mile

            One of the problems with creating a demand at this point is that not everyone has access to a high speed last mile connection.  Many areas of the country have no alternative to 56k dial-up access.  Without universally available high-speed options, we will never have the ability to replace the PSTN with the Internet.  As a side note, not everyone in the country has a PC but inevitably the day will come when everyone will have some device with Internet connectivity, even if it is only an Internet “phone.”  Having everyone connected to the Internet will also be required if we are to replace the PSTN with the Internet.

 

Let’s Get on the Ball

            VoIP is a technology whose time has come; it has been on the verge for years and it is time to implement it.  The same thing applies to IPv6, which we need in order to widely implement VoIP; it has been ready for years we need to get started.  IPv6 is not only needed for VoIP but also for the increased address space that it provides; in fact the Japanese government has decided on a deadline of 2005 by which time Japanese companies need to be upgraded to IPv6 (Demaria, 2002).  Once everyone has equal access to a high-speed Internet that is run on IPv6, then we can all enjoy a new world of Internet voice communication.  And maybe, eventually, we will eliminate the PSTN entirely and have one unified Internet and truly a world wide web through which everyone is connected. 


References

 

Cyganski, D. Orr, J. & Vaz, R. (2001). Information technology: inside and out. Upper Saddle River: Prentice-Hall.

 

Demaria, M. (2002). Slow road to IPv6 – More address space is on the horizon, but migration and vendor support is still lagging. Network Computing, p83.

 

Hinden, R. (1995). IP next generation overview. [Online]. Available: http://playground.sun.com/pub/ipng/html/INET-IPng-Paper.html#CH9. April 26, 2002.

 

Leon-Garcia, A., & Widjaja, I. (2000). Communication networks: fundamental concepts and key architectures. Boston: McGraw Hill.

 

VoIP builds a better bridge” (2000). Communications News, p20.

 

 

Home Outdoors  Various School Projects Rat Babies Wacky Test Results Contact Kevin McMahon