New technologies and new products are constantly being introduced. The usage of IT is increasing, all processes are going online.

Transcription

1 Voice over IP (VoIP) Your key to success We live in the age of communication. Access to knowledge and the exchange of information are an integral part of our daily lives - Especially in business. In the beginning the Internet was a technology for business, research, development and military institutions with limited bandwidth. Today, it has grown to be the main medium to obtain and exchange information. High bandwidth makes video and audio communication possible. The world is interconnected. New technologies and new products are constantly being introduced. The usage of IT is increasing, all processes are going online. It is almost incomprehensible that the influence of modern communication technologies like the Internet in our business and private lives has just emerged a decade ago. On the other hand, the key invention for long distance communication is already 125 years old: the telephone. It has almost not evolved over the years. And why should it have changed? Converging networks VoIP is a technology that makes the transport of voice over data networks, i.e. via the internet protocol, possible. The services of the traditional PBX are provided by computers. On these communication platforms the transport of data and voice are done simultaneously. Business processes can be optimized. Phone connections can be initiated with a mouse click, speech can be recorded, saved in files or attached to s. Electronic calendars can automatically reroute calls. Only one address is needed to deliver voice, or fax. Apart from all the future possibilities, using VoIP today has a lot of advantages: no expensive telephone systems or special servers are needed. Maintenance costs can be reduced, i.e. technical issues are resolved by one IT- department. Operation costs, e.g. calls within the data network are at no additional costs. VoIP versus PSTN This article describes why VoIP as a packet based technology has structural advantages over the alternative technology, circuit based or switched networks traditionally used in telephone systems. These advantages are lower cost and better service. What are "circuit based/switched networks"?

2 Over a hundred years ago, the first telephone was connected to a public network with a copper cable. This cable only had to transfer voice that was spoken between the public network and the phone. According to the requested number, the operator connected the parties that wanted to communicate. Later, these jobs were rationalized by conveying the dialed number thru pulses on the line and having the public network automatically switching the lines. Using touchtone later just speeded up the process but it is essentially the same thing, i.e. the calling phone is more or less directly connected to the called phone. The next hundred years that picture did not change too much. Instead of using dedicated lines between the post offices, engineers started to multiplex these expensive lines to use one cable for several connections. Digital standards like ISDN, T1 and E1 transferred the voice digitally, but the paradigm of transporting voice over a switched line, i.e. reserved or setup path did not change. What are "packet oriented networks"? In the meantime, the computer industry developed at an incredible speed. With increasing processing power the need to feed more input to the computer and to deliver the results to the user arised. A lot of IO and storage devices were conceived. The communication systems in the beginning only connected peripheral devices like terminals but soon were also used to connect computers. In comparison to computers the communication paths were very unreliable (in those days computer people were complaining about reliability to telecom engineers - how things change). Thus, the computers were programmed to put data into frames, i.e. packets with a checksum to ensure the data was transferred correctly. If an error was detected the packet was simply rejected and retransmitted. It was not important to do this with a stringent timing restriction which voice has but to ensure that the data arrived correctly. After a while, there were a lot of "standards" to choose from, since each computer vendor (who also sold the operating system, communication stacks, printers, terminals, applications, etc.) thought it was best not to be interoperable with the competition. This arrogance worked well in the 60s but nearly destroyed IBM later. The first step to break this monopoly was done by a group of researches that wanted to connect different kind of computer networks. By defining a protocol that inter(connected) net(works) regardless of their physical attributes the Internet Protocol (IP) was born. In order to get the research funded, they told the military they could conduct war more efficiently even though they probably wanted to play games themselves. It was then used mainly by students, soldiers and geeks. It only really became popular when another researcher in CERN wanted to improve the documentation system by writing a browser. With search engines helping find interesting sites gave way to an unprecedented boom of the Internet Protocol.

3 In contrast to a switched network, a packet based network such as the Internet splits a data stream up into several packets that are sent through the network. All of these packets have their own destination address and in theory (and sometimes in practice) these packets travel different ways to their destination. That implies several benefits and problems: The underlying network resources are used only when data has to be transported. This allows a much more efficient usage of the resources. That in turn makes data compression economically feasible. The worst-case bandwidth can be allocated, but statistically only a fraction of that bandwidth is used. The rest of the allocated bandwidth can be used for other services like and file transfer. The big problem with this technology is that it is hard to guarantee that a packet will be delivered reliably within a predefined, short period and in the correct sequence. The network elements (routers and switches) that determine which way a packet will travel, have to decide this very fast for many packets. For this problem, special protocols have been introduced which negotiate the routes before data is actually transmitted. The voice can be sent simultaneously with other services. This makes cost sharing with other Applications possible. The required network technology has become available in the last few years. Total Cost of Ownership Operators are interested in the "bottom line" of the cost for their infrastructure for a given period of time. These costs can be divided into Investment (network infrastructure, software and cables), Maintenance Cost for running service teams including training Operation Investment costs for computer network technology are much lower than the respective telecommunications technology. Analysis found a factor of 10 and more. In many cases there is no need for any extra investment in the computer communication infrastructure because it is there or will be anyway. The same applies for the maintenance costs in a computer network. The service Team for the computer network also maintains the telecommunication network. Computer network management tools have become sophisticated enough to enforce service level agreements.

4 Connection fees are usually flat rate in the Internet environment. Market forces drive more and more internet providers into this direction. Billing for internet usage is a lot cheaper than billing each connection separately. In sum, the total cost of ownership for VoIP is significantly lower than a comparable switched network, especially when the network is used for other services than voice. Media integration Media integration comes for free when the network that is used is the Internet. There is no communication barrier for a telephone device to talk to a web server or to a database. That makes it simple (and cheap) to integrate services. A telephone can be used to listen to radio, as well as for chatting. It is very feasible to use the telephone directory and calendar that is stored on a PDA (if that PDA is an internet appliance) and to remotely access the phone's mailbox with a standard web browser. The possibilities are numerous and the future will show which services best complement the core functionality of making one-to-one phone calls. Why Phones? Often we get to hear "what is and/or why VoIP?" and the question ranked a close second is "why a VoIP phone if I can use a computer?". Well, the difference is that we hear the first question from "normal" people, i.e. people doing business and running normal lives who know why the need a phone but not why it should be VoIP based. The other question originates from technically inclined people - I do not want to call them geeks being a computer scientist myself. But they tend to adopt technology because it is different and new not necessarily pragmatic or economical. We believe that the first group is the bigger one worldwide and that is why we focused our Company on VoIP phones. While PCs will be used there are some indications that PCs do not make good voice communication devices: Netmeeting is distributed at no cost with Windows. It must have cost about one billion dollars to develop and to promote and it is still not used a lot. Phones are specialized devices for voice communication und thus easy to use for this purpose. And very often they deliver a lot better voice quality than PCs. PCs (or the popular operating systems on them) have trouble browsing the web and playing music simultaneously. When I optimize my business I want to open applications on my computer without disconnecting my phone call.

5 One of the main arguments against VoIP in general is stability. A phone is seen by many to be the epitome of stability where as a non Linux or Unix based PC is the negative example. In short, we take it as a compliment when people say they like our design and state that they cannot see that our phone is VoIP based. Voice over IP (VoIP) Glossary Connection versus packet oriented Networks are generally categorized into packet or connection oriented. In a connection oriented network data is transferred by switching the network elements together to provide a virtual connection or to use a fixed connection (e.g. a telephone line). During the transfer the connection is reserved for this purpose only whether useful data is transmitted or not. Thus a lot of potential bandwidth is wasted. In a packet oriented network, data is sent only when it needs to be and grouped for efficiency into packets or frames. The destination address is attached to the beginning of the data and the network elements then route the packets. While useful data is only sent when needed, each packet needs to include at least the address information and depending on the network the order of the packets is not always guaranteed or even whether they arrive. Thus, traditionally voice was transmitted on connection and data on packet oriented networks. ENUM ENUM links DNS entries with the E.164 telephone numbering system, two well known numbering mechanisms for locating resources in networks. The well-known PSTN numbering mechanism uses numbers in the range 0-9 to indicate country, area, service, mobility and other things. For example people in South Africa know that numbers starting with 011 are located in the Johannesburg area and they also know what the cost of the service will be if they dial a number starting with these digits. Numbering plans explain the system that lies behind these numbers. DNS names use alphanumeric characters to make it easier to remember the address. While it is hard to remember " ", it is easier to remember "astrovoice.net". While it would be easy to access a telephone with the number "sip:info@astrovoice.net" from a computer with a QWERTY keyboard, dialling such a number from a normal telephone requires a finger-breaking procedure. ENUM solves this problem by providing a lookup mechanism that translates " " into "sip:info@astrovoice.net".

6 H.323 H.323 is an ITU standard for the usage of multimedia communication via packet-oriented networks that guarantees interoperability between different equipment vendors. The largest packet-oriented network is the Internet but also WAN, ISDN or dialup connections on which data is transported in packets (e.g. PPP) belong into this group. H.323 describes the general infrastructure and the utilization of different speech coders and protocol signalling stacks. The speech coders are defined in their respective sub standards, e.g. G.711 (Alaw and ulaw used in ISDN), G.722, G and G.729.A for speech encoding. H.323 is definitely a widely deployed and mature standard, but it is also criticized for being complicated to implement by vendors and uses a lot of resources which are not abundant (especially in terminals). H.450 The H.450 Supplementary Services is a series of standards that define extended functionality and distribution thereof in a H.323 infrastructure. These services are called supplementary since they extend the basic services of H.323 which essentially boil down to being able to establish and release a connection. Examples of such Supplementary Services are Hold (local and remote), Call Waiting (an indication that a person is trying to reach someone who busy talking to someone else), Call Diverting (call is transferred when busy), Call Redirect (a call is transferred to a mobile after working hours), Pickup, Parking, etc features that a classical PBX and ISDN offer. In addition, mechanisms are provided that enable vendors to tunnel proprietary supplementary services if need be of course this is not intended to become a standard but is a workaround until these features are interesting enough to be integrated. Jitter Jitter is the variance of latency (i.e. delay) in a connection. The problem is that audio devices or connection-oriented systems (e.g. ISDN or PSTN) need a continuous stream of data. In order to compensate for this, VoIP terminals and gateways implement a jitter buffer that collect the packets before relaying them onto their audio devices or connection-oriented lines (e.g. ISDN), respectively. An increase in the jitter buffer size decreases the likelihood of data being missed but also has the drawback that it increases latency of a connection. Latency The delay or time span between the voice being digitalized at the senders Location and then output at the receivers end is the latency of a connection. Latency is influenced by the distance the data has to travel, the packet size, the number and delay time of network

7 elements between the terminals and of course the latency generated by the terminals themselves when sending, receiving, encoding, decoding and compensating jitter. LPCP LPCP (Lightweight Phone Control Protocol) is a standard that is used to control telephones in a pragmatic and simple way. Thus, the memory and resources needed on VoIP telephones and terminals can be reduced to the minimum needed which in turn will be more cost effective. The call signalling (SIP / H.323) is done for the phone on a server that has enough memory resources. QoS Quality of Service pertains to the quality of a connection and this is especially important for connections relaying voice since the user feels the impact immediately. A retransmission cannot make up for the lost data. The internet protocol was devised as a best effort data network and thus it does consider jitter, latency or even data loss a problem. Ergo, it does not handle voice well per se. To make the transmission of voice possible it must be given the necessary priority and bandwidth. There are mechanisms for reserving bandwidth (see RSVP) but they add network equipment with an additional burden of handling this functionality and slow down establishing connections. The other pragmatic approach to this problem is to acknowledge that normally the access point (interconnection between LAN and WAN) is the most critical section. By prioritising the packets (see ToS) (of course the network equipment has to Support this) and ensuring that the access point is not overloaded good QoS can be achieved. The data traffic load in the backbone is about 10 times that of voice (thanks to WWW) of carriers so this should not be the problem. RTP The RTP (realtime transport protocol) labels all information transferred by a sender with a timestamp. By examining the timestamps the receiver is able to sort the packets in the original order and synchronize real time streams and/or compensate jitter in audio data. RTCP The RTCP (realtime transport control protocol) was devised to give Applications a status on the quality of a network. With this information parameters affecting the transmission of data, e.g. the jitter buffer size, can be optimized. RSVP The RSVP (resource reservation protocol) makes it possible to reserve bandwidth in nonterminal network elements such as routers. This and prioritisation (see TOS) is done to

8 practically eliminate latency, jitter and loss of packet problems for realtime application such as VoIP. SIP SIP (Session Initiation Protocol) is a highly pragmatic, ASCII-based protocol and competing standard to H.323. Its main advantages are that it is easy to implement, debug and to integrate applications. It is newer than H.323 and is pushing H.323 out of the scene. TOS In order to specify the priority of a packet the internet protocol has a ToS (type of service) field. VoIP VoIP (Voice over Internet Protocol) is a term used for voice being transported via the internet, intranet or data links to the internet regardless whether H.323, SIP or a proprietary standard is used. STUN Solutions for NAT traversal in SIP Environment Basics - NAT in SIP Network Address Translation (NAT) is a big problem for SIP, because both the SIP signalling and the media uses UDP to transport their information. The reason for introducing NAT was the shortage of public addresses available on the Internet. Therefore, users started to share one IP address through a NAT gateway in a private network. Typically, addresses in the private network have the form x.x or 10.x.x.x. When a phone sends a packet from a private network to a public network, the NAT gateway allocates a port for this new "connection" and patches the IP packet according to this port. Looking at the IP transport layer of the packet, the recipient of the message thinks it came from that port on the NAT gateway. However, because SIP uses explicit addressing in the SIP contents (and that address has priority in SIP), the response to that packet will not find the right way. The same problem occurs when the phone wants to invite a person for a call. In the invitation message, it needs to put the IP address and the port where it expects the media to go to. If it puts in its private address here, the media will not find its way from the public Internet to the phone.

9 Firewalls use this mechanism to filter traffic entering and leaving a zone. Often, this is combined with NAT. There are different approaches to solving this problem: Use a SIP-aware NAT router Set up a static route in the NAT gateway Use STUN to measure out ports SIP-Aware NAT Router With the increasing acceptance of SIP as a standard protocol and its needs, more and more network equipment becomes "SIP aware". That means, if a SIP packet passes a NAT gateway, the gateway will inspect it and make the necessary patches to the packet. It allocates ports for the media and puts the right addresses into the SIP address. This solution is recommended in environments where there is sufficient number of subscribers in the private network. There are "Application Layer Gateways" available to solve this problem (like the snom 4S SIP NAT gateway), which are installed on the firewall or on the NAT gateway computer. Some gateway vendors offer special add-ons to their standard firmware that make their equipment SIP-aware. Some advanced DSL routers include this ALG in their standard configuration. When you use a SIP-aware router, you should make sure that every SIP message goes to the NAT gateway directly. There are several ways of doing this, depending on the equipment you are using in the private network. For this mode, NAT detection should be set to "Off" or "Automatic" with the STUN server field empty (this is the default). The fields "dynamic RTP port start", "dynamic RTP port end", "Network identity" and "Network port" should be left empty and "Local SIP port" should be set to default. If all messages are to go over the NAT gateway, you can set the outbound proxy to the address and port of the private NAT gateway. This address takes the format of a SIP URL. If there is no initial "sip:", the phone will complete the URL according to the rules for the SIP URL. Valid examples for this field include " ", " :5062", "nat.company.com" and "sip:nat@company.com:5065". In this case you should set the "Treat as initial route" field to "Address", so that no additional headers are inserted into the SIP messages. Especially where there are several phones located within the same NAT, you might want to avoid all traffic going through the NAT gateway. In such cases, we recommend setting up a SIP proxy (like the snom 4S SIP proxy) in the private network and pointing all traffic to this proxy. It will then take care of the traffic leaving the NAT and redirect the packets to the NAT gateway itself.

10 STUN Solution Setting up the NAT router is impossible in many cases, and new equipment may be too expensive. For these environments, "Simple Traversal of UDP Through NATs" (STUN) has been defined in the SIP environment (see RFC3489). STUN uses a server located in the public Internet. The phone sends a test message to the server and receives in the response which IP address and port the server received. The client may ask the STUN server to send the packet back from a different location. In this way the client can determine the type of NAT present. You can use the snom 4S proxy server for this. To set up a phone for STUN, set the NAT detection to "Automatic" and enter the address of the STUN server in the field "STUN server". The field must have the format hostname [:port], a valid STUN SERVER is " :5060". If you set up DNS, please include a SRV entry for _stun._udp, which points to the right address. The phone should then be able to find the STUN server on its own. SNOM STUN SERVER -For eval use only :5060 When you use STUN, the fields "dynamic RTP port start", "dynamic RTP port end", "Network identity" and "Network port" should all be left empty and the "Local SIP port" field should be default. Static Route Firewall Configuration Setting up a static route on the NAT gateway is the most powerful, but also most complicated way of setting up the phone in a NAT environment. For this mode, NAT detection should be set to "Static". There are a couple of settings available for the static route: Dynamic RTP port start, end: The range of ports that are used by the phone for media including start and excluding the end port. Network identity (hostname, port): The phone will insert the inserted name as hostname and port into the SIP messages. The values must match the router set-up. Local SIP port: With this flag you can decide whether the phone uses the standard port (5060) or the port provided in the network identity as local port. If the router is not able to translate ports, you must use network port.

11 On the router, you need to set up one UDP port for the SIP traffic and several ports for the RTP (media) traffic. Try to set up at least ten ports for RTP, so that the probability of a port conflict is not too high. Make sure that the settings on the phone are exactly the same as those on the router. This is very important because otherwise the service will be unreliable and frustrating. Changing these settings requires a reboot of the phone. Decision-making matrix Scenario Best Solution Home user with old DSL router STUN Experienced home user Firewall Static Route Use software ALG Small office Use advanced SIP aware router (IX66) Phone on public Internet Automatic NAT or NAT detection off Phone only used in private network NAT detection off *********************************************************************