WO 2001052460 A2
A direct internet access system providing high band width internet access to users, in addition to voice services, without any changes to the existing cabling infrastructure and an always-on internet access permanently available at the user premises, comprising a DSU which combines voice and internet data packets on a single twisted-pair wire at the user premises; an IAN, at the service provider's premises, for separating the voice and data traffic from a number of users and routing them independently to the PSTN and the internet respectively the IAN being connected to the internet through E1 data ports or an Ethernet port and to the PSTN through (1) and E1 voice port or (2) POTS lines with addition of an optional SMUX module that converts a single E1 line to multiple POTS lines.
A DIRECT INTERNET ACCESS SYSTEM
This invention relates to a direct internet access system.
According to this invention, the direct internet access system providing high band width internet access to users, in addition to voice services, without any changes to the existing cabling infrastructure and an always-on internet access permanently available at the user premises, comprising a DSU which combines voice and internet data packets on a single twisted-pair wire at the user premises; an IAN, at the service provider's premises, for separating the voice and data traffic from a number of users and routing them independently to the PSTN and the internet respectively the IAN being connected to the internet through E1 data ports or an Ethernet port and to the PSTN through (1) an E1 voice port or (2) POTS lines with addition of an optional SMUX module that converts a single E1 line to multiple POTS lines, wherein the said direct internet access system further comprises (a) a DSP based router accelerator module for implementing time-critical functions at high speed and reducing total routing time (b) a DSP based fire wall implemented by a configuration of access rules, applied to each packet (c) a subscriber lineframer for multiplexing and transmitting the signaling information for voice and internet data on a single channel as packets from the DSU to the IAN; (d) a dynamic bandwidth controller provided in the DSU and the IAN, where by the DSL bandwidth is dynamically split between voice and data packets (e) a subscriber key generator for assigning each DSU with a unique confidential and user- unchangeable ID (f) a packet discriminator incorporated in the DSU, for eliminating or resolving specified data packets (g) a hot swap controller provided for the IAN whereby when one of its line cards is removed and replaced, the operation of the said system is not affected.
This invention will now be described in further detail by referring to the accompanying drawings, wherein
Fig.1 illustrates internet connectivity today Fig.2 illustrates, by way of example, and not by way of limitation, one of various possible embodiments of the architecture of the direct internet access system, according to this invention
Fig.3. illustrates, by way of example, and not by way of limitation, another of various possible embodiments of the architecture of the direct internet access system, according to this invention
Fig.4. illustrates Class A service with 4 IP addresses
Fig.5. illustrates Class B Subscriber connection
Fig.6. illustrates multiple links to the internet
Fig.7. illustrates the stacking layout for data and voice
Fig.8. illustrates the normalized packet delay diagram as a function of data rate.
Fig.9. Illustrates a diagram pertaining to the estimation of back haul capacity requirements (P=250)
The Internet started as a network to connect Universities together. Today it has evolved into a network that provides seamless communication for millions of people worldwide. The user base of the Internet is experiencing exponential growth and is expected to surpass 100 million by the year 2000. Services such as e-mail and World Wide Web are transforming the way people interact and are fast becoming the norm for personal and business communications.
The Internet is fast becoming as essential as the Telephone, and therefore, it is conceivable that soon every home and every office will need Internet access along with their phone service.
Today an Internet subscriber connects to the Internet using a Modem connected to his PC to place a call over the telephone line to connect to the Internet Service Provider (ISP) through a bank of modems connected to phone lines as shown in figure 1. The ISP, in turn, maintains backbone connectivity of their Router to other ISPs, via. leased lines.
Today's Internet connectivity solution is entirely dependent on the telephone network. For an ISP, the key is to acquire a sufficient number of telephone lines to provide on- demand access to subscribers. For example, if an ISP has 10 telephone lines, no more than 10 subscribers can connect to the Internet at any given time. Of course, the ISP can have more than 10 customers, as not all of them will be using the Internet at the same time. If only 20% of the total number of subscribers are expected to use the network simultaneously, the ISP could have 50 Internet subscribers.
The user experience on the Internet is primarily dependent on Bandwidth, the data rate at which communication takes place. This depends on the data rate supported between the subscriber's modem and the ISP's modem and also on the quality of the telephone lines connecting the two modems. Whiie the fully digitized telephone network may support connection speeds of up to 56Kbps, it is typical for connections in small towns, having only analog trunks, to support data rates of only 4.8Kbps or less.
Outlined above is the Internet access mechanism widely adopted today. It works rather well. The percentage of telephone users who use the Internet is growing slowly to 1-2% in most developing countries. However, in some areas, a peculiar problem is being faced. With the number of Internet users in these areas growing to 5%-10% of all telephone users, the telephone network appears to get congested during certain busy hours. The reason is not difficult to find.
While the average telephone call lasts for a few minutes, the average Internet call lasts for 30 to 45 minutes. Present day telephone switches were not designed to function like this. They were built under the assumption that no subscriber will keep a line busy for more than few minutes at a time. This is an important assumption because when a subscriber makes a call, very often he holds a trunk line because the call goes from one switch to another most of the time. The number of trunk circuits were chosen with the assumption that trunks will be held only for a few minutes and statistically, a particular number of trunk lines are enough to cater for a particular number of customers.
When the calls are to the Internet, this reasoning breaks down totally. Now the calls are held for as long as 45 minutes. This means the trunks, which would have normally been free to serve other customers, are now held up for the Internet calls. The subscribers start getting more and more busy tones. The reason is simple - there are just not enough trunk lines to take the calls! *
A normal telephone user generates O. Erlang traffic, meaning he will use the lines for about 10% of the time. As long as the number of Internet users is around 1% of the telephone users, even if each of them were to generate lErlang traffic, the total traffic (voice plus Internet) on the Public Switched Telephone Network (PSTN) would not exceed 0.11E. But when the Internet usage reaches 10%, even if the traffic generated per subscriber is only 0.3E, the total load on the network would get close to 0.13E.
The telephone network is usually designed to handle 0.1E traffic per subscriber with a blocking (or Quality of Service - QoS) of 0.5%. This implies that, with the average traffic generated by a subscriber being 0.1E, only 1 out of 200 calls will be blocked (not completed) due to non-availability of resources at the PSTN. However, if the PSTN is loaded with 0.13E traffic, the blocking probability exceeds 5%, and with loading of 0.15E traffic, the blocking exceeds 15%. The network would be congested with many calls not maturing due to blocking. As the traffic increases further, the network may collapse and fail to complete most calls.
Such a debacle is not widely observed today because the percentage of Internet users is small now. As the usage grows, however, most telephone networks would get congested and collapse.
The PSTN is the most widespread wired network today. Other mediums connecting homes and offices are the power and cable TV networks. Presumably these could be used to provide Internet access.
Conventional Power lines have not been designed for communications. The lines linking homes and offices are in fact, a tap from the main power lines on the street. It is difficult to provide widespread, high-bit-rate Internet access on these lines, though attempts are now being made to explore this option.
Cable TV networks, on the other hand, are actually a communication medium. The only problem is that they were designed for one way communication only, from the service provider to the subscriber. Advances in technology are making two-way communication and Internet connectivity possible on the Cable network. The main problem still associated with cable networks is the lack of structured wiring. Particularly in developing countries, where cable networks are often laid in an unplanned manner, and where taps are taken without consideration to the possible transport of data, providing Internet service at even PSTN quality is questionable.
While other options may become very attractive tomorrow, there is little doubt that today the telephone network is the most promising Internet Access Medium, provided one could use it without causing concomitant congestion in the PSTN. The solution to this dilemma is to find a way to occupy the PSTN more efficiently than do present day Internet Access methods.
The congestion of the telephone network when extensively used for Internet access is due to the fact that Internet access sessions last much longer than a telephone call, and throughout the duration of connection a circuit is occupied on the circuit switched PSTN. But if one closely studies the Internet traffic, one finds that the total bandwidth required for medium speed Internet access rarely exceeds that required for a voice connection. Telephones typically generate 0.1E traffic implying that a telephone, on an average, is used only 10% of line. But they generate voice traffic continuously during the period of connection. On the other hand, Internet sessions are usually of a longer duration, but do not generate continuous data traffic. The data traffic is in fact bursty and for most of the connection time, no data is transmitted or received. Even a busy user rarely transmits or receives data more than 10-15% of the time implying that even if the user is continuously on, the data bursts rarely use more than 10-15% of the available bandwidth. The problem is, irrespective of the usage of available bandwidth, the circuit-switched PSTN connection is on, utilising the resources continuously.
The obvious solution is to find a way to occupy the PSTN resources only to the extent that is required by the subscriber data.
One way would be to set-up and drop connections whenever data bursts need to be transmitted. The only problem is that these data bursts normally are too small, barely a few hundred milliseconds and could be very frequent. Dropping and setting up connections would, by themselves, consume much more time as compared to that of data transmissions. Also, most PSTN networks just cannot handle such frequent connections and disconnections.
The approach is being attempted in a new standard known as AO-DI (for Always On Dynamic ISDN). This relies on two important characteristics of PSTN - that ISDN services are available throughout the PSTN, and secondly, in an ISDN network, the subscriber is permanently connected with the ISP via a low-bit-rate packet data channel called D channel. When larger data bursts occur, a circuit-switched B channel is temporarily set-up. The ISDN signaling allows fast set-up and disconnection of B channels. This property of the ISDN network, in addition to the fact that most small bursts could be carried on the D channel itself, is effectively used to provide Internet communication using AO-DI without causing network congestion.
The procedure is currently being incorporated in some of the networks. Though it looks attractive, it is not clear as to how well this would work under conditions of heavy load. Also, except in few countries, ISDN has not really penetrated the PSTN so much to be an effective transport for Internet.
An alternative to this and probably a better solution to this is similar to what is being increasingly used in recent years for voice networks. The best way to take care of statistical variation of usage is to combine the traffic from several subscribers.
A telephone exchange uses this and reduces the number of trunks. For example, a 40,000 subscriber telephone exchange typically has only 4,000 trunk lines, and assigns them to users on demand. The individual, random behavior of each subscriber gets significantly less varied when traffic from multiple subscribers is combined and resources allocated on demand.
This has been recognized since a long time and it has also been recognized that concentration closer to a group of subscribers helps tremendously. For example, instead of carrying out the concentration for 40,000 subscribers at the exchange (which would require 40,000 lines from subscribers' premises to that of the exchange), it helps tremendously if the concentration is carried out at the curb (street corner) in groups of 1000 subscribers. The lines for the 1000 subscribers are now terminated at the curb and only about 120 to 150 ports are taken to the exchange after concentration. The total number of ports at the exchange for 40,000 subscribers therefore reduces to 5000 - 6000.
A similar approach could be adopted for bursty data. And the key to this is two fold -
Separate Internet data traffic from voice traffic at the earliest point possible and
Carry out the concentration of bursty data.
Since data traffic is bursty, and the Internet sessions are on for a long time, it is essential that the data is handled differently. It is therefore necessary that the local loop (which could carry both voice and data) terminates at the earliest possible point, possibly at the curb. The Internet data now needs to be separated from voice. The voice calls are to be connected to circuit-switched trunk lines on demand. The bursty Internet data from multiple subscribers are now combined and concentrated. The burstiness is largely removed and the line-utilization becomes nearly constant. The concentrated data is now carried to the Internet Service Provider (ISP) using leased (preferable) or switched circuits. Even if switched circuits are used, it does not create congestion as the data on these circuits is the concentrated data from multiple subscribers. Even if we assume a channel utilization of 10% by each subscriber, due to the bursty nature of the transmission, on concentration, one channel can serve about 10 subscribers. The net load on the PSTN, therefore, does not exceed 0.1E per subscriber.
The Traffic Aggregation technique described above is generally referred to as Direct Access Technique. Internet access thus provided overcomes the limitations of the telephone network and offers direct access. The Direct Internet Access System (DIAS) proposed herein incorporates this technique to provide high bit-rate Internet access in addition to telephone access for each subscriber.
The Direct Internet Access System (DIAS), proposed herein allows telecom service providers to provide high bandwidth Internet access to residential and corporate subscribers, in addition to voice services, without any changes to the existing cabling infrastructure. In contrast to current residential PSTN (Public Switched Telephone Network) and ISDN (Integrated Switched Digital Network) dial-up access, the DIAS provides an Always On internet Access that is permanently available at the customer's premises. Using Digital Subscriber Loop (DSL) techniques, seamless voice and data connectivity is provided to the customer over the same pair of copper wires.
Implementation of this system is often done using the existing cables. All that is required is the installation of the IAN (Integrated Access Node) at the Telephone exchange and a DSU (Digital Subscriber Unit) at the customer premises.
The DIAS has a Digital Subscriber Unit (DSU) that combines voice and Internet data packets on a single twisted-pair wire at the subscriber's premises. At the service provider's premises, the Integrated Access Node (IAN) separates voice and data traffic from a number of subscribers and routes them independently to the PSTN and the Internet respectively.
The IAN is connected to the PSTN via an E1 voice port, and to the Internet either through E1 data ports or through an Ethernet port. Alternatively the PSTN connectivity can be achieved through POTS lines with the addition of an optional Subscriber MUX (SMUX) module, that converts a single E1 line to multiple POTS lines.
The DIAS system as shown in figure 2 provides 2 types of voice and data services -
♦ The BDSU (Basic Digital Subscriber Unit) is designed for the SOHO (Small Office Home Office) and residential Internet user. It provides a permanent Internet connection at a maximum data rate of 144 kbps, which drops to 80 kbps when the telephone is in use and transparently goes back to 144 kbps when the telephone goes on-hook.
♦ The HDSU (High bitrate Digital Subscriber Unit), which is designed for corporate subscribers, can provide voice connectivity for upto 8 telephones and permanent data connectivity of upto 2 Mbps, which drops by 512Kbps when all the 8 telephones are in use.
The Basic Rate DSU (BDSU) located at the subscriber's premises has a telephone interface (RJ11) and an Ethernet port (RJ45) to provide Internet Access. The High Bit Rate DSU (HDSU) has an Ethernet port and 4/8 Telephone Interfaces (RJ11), thus having the ability to connect to 4 or 8 independent telephones at a corporate office. The BDSU (BSX 200LP) has local powering off the AC Mains (110V/230V) and a backup battery to support the telephone on power failure. The Ethernet port providing Internet access is off during power failure. In addition, a low cost BDSU unit (BSX 200) comes without a battery back-up and is powered over the network from the IAN to provide telephone service on power failure. The IAN can provide power to this unit only if the length of the copper is around 800m to 1 Km depending on the wire gauge.
The BDSU and the HDSU are connected to the Integrated Access Node, located either at a street corner (curb) or at the central office, using a twisted pair copper wire. For the BDSU, the maximum length of the copper can be 5.5 km when 0.4 mm tp copper is used. 144 kbps Internet access can be provided on the BDSU Ethernet port in such a configuration when the telephone is not being used. The Internet access rate seamlessly drops to 80kbps when the BDSU telephone is being used providing the telephone a 64kbps circuit-switched access. The HDSU is also connected using copper to the IAN. The maximum rate at which Internet Access is provided to the HDSU Ethernet port is 2 Mbps, and this is possible when the length of the copper is less than 2 Km (0.4mm tp copper). The bit-rate on the HDSU-IAN link drops for higher lengths of copper, thus reducing Internet access rate on the HDSU ethernet port. Each telephone on the HDSU uses 64kbps when off-hook and reduces the bit-rate of the HDSU Internet access by the same amount. In on-hook mode, the Internet access rate seamlessly reverts to the original value.
Both the HDSU and the BDSU have a minimum routing function built-in and routes the packets on the Ethernet meant for ISP on to the IAN.
The IAN separates the voice traffic and the Internet traffic. The voice traffic from each BDSU/HDSU is circuit switched on demand to one of the 64kbps slots on E1 lines connected to an exchange. The protocol used between the IAN and the switch is the ITU standard V5.2. Optionally the IAN can have a SMUX unit which would connect the DIAS subscriber to the exchange on 2wire interface as shown in figure 2.2. Thus the IAN acts as an access unit of the exchange and provides all the features and services of the exchange to the subscriber. The Internet traffic from each BDSU/HDSU is concentrated at the IAN (IAN essentially acts as a RAS at this point) and passed on to the ISP on an 10Mbps Ethernet or an E1 leased line. Upto 2 such E1 ports are provided on the IAN for connection to the ISP.
Besides the E1 port to connect to a telephone exchange, (also referred to as E1 IN port) IAN has another E1 voice cascade port (referred to as E1 OUT port), which enables cascading multiple lANs as discussed in the next section.
Each IAN is designed to support a combination of HDSU and BDSU subscribers. The options are,
60 BDSU subscribers
48 BDSU subscribers, 4 HDSU subscribers,
36 BDSU subscribers, 8 HDSU subscribers,
24 BDSU subscribers, 12 HDSU subscribers,
12 BDSU subscribers, 16 HDSU subscribers,
20 HDSU subscribers.
The DIAS system is designed so that multiple lANs could be cascaded at the service provider's premises. Multiple lANs are connected to each other on a 10 Base T Ethernet (switched/shared). The Internet data from all the lANs can thus be combined and taken to the ISP on one or more leased E1 lines.
Optionally, it is possible for lANs to set up a m x 64 kbps switched call to an ISP router. The concentrated Internet data from various subscribers can thus be taken on m slots of the E1 port to the exchange and then on to the ISP. The combination of leased lines and switched Internet access lines provides Always-ON service to a subscriber and still allows extra bandwidth to be made available for Internet on demand.
As mentioned earlier, an IAN has an E1 (IN) port, to interconnect it to an exchange. Interconnecting multiple lANs on E1 (OUT) cascade port allows the switched service of multiple lANs to be on a single E1 line to the exchange. This is especially important as voice traffic on a single IAN from its subscribers is unlikely to use a full E1 port to an exchange. Stacking for data and voice is explained in detail hereafter.
Summary of features of the proposed invention
♦ Upto 60 Basic Rate (BDSU) Subscribers and upto 20 high bit rate (HDSU) subscribers.
♦ Up to 4 Mbps leased WAN connectivity
♦ E1 trunk interface to PSTN with V5.2 signaling.
♦ Optional 2-wire connectivity to exchange using SMUX.
♦ SNMP manageable
♦ CLI for configuration
♦ RADIUS for user database
♦ Power Feed to subscriber units
♦ Cascadable for more capacity
♦ 10BaseT port for management as well as DIAS stacking
♦ Remote Management of subscriber units At the Central Office
The Integrated Access Node
The IAN is housed in a 6U/19" sub rack. All the I/O connections are at the front with the power supply connections at the rear. It has the following cards:
Each BDSL line card has a connector at the front that has 24 connections for 12 subscribers. Each HDSL Line card has 8 connections for 4 subscribers. All these connections are through Euro Connectors. In addition to these, the Switch Card has an Ethernet Port and a RS-232 port for Management and cascading. All E1 connections come from the E1 Uplink Card. There are four E1 ports on the E1 Uplink Card. Two are for voice trunks and are marked E1 IN and E1 OUT. The other two are for data connection over the WAN. A second port on E1 uplink card is for a remote alarm unit.
The IAN consumes 100W without any power feed to the subscribers. Each subscriber power feed adds about 5W to the power. For supporting 60 subscribers, the DIAS will consume 400W in total. The supply is powered use 48V +12V.
The power feed circuits are protected against short circuit to a maximum current of 100mA.
The PSTN Interface in the DIAS is an E1 (IN) port conforming to ITU-T G.703, G.704 and G.823. Besides there is a E1 (OUT) cascade port. The E1 IN Port and the E1 cascade or E1 OUT Port can be used to cascade a IAN to two other lANs. When a single IAN is used, the IN E1 port will be used for connecting to PSTN. The PSTN is the clock master on the E1 Port and the IAN loops back the clock. WAN Interface
The WAN Interface in the IAN is made up of two E1 ports for Internet connectivity to the WAN. Both ports are channelized E1 ports and conform to ITU-T G.703 electrical specifications.
Management and Metering
The entire system is capable of being centrally managed. The management can be from a terminal using CLI or from a remote location, using SNMP. The system supports standard SNMP management protocols.
In order to be maintainable, the DIAS provides a set of operational testing and maintenance options.
All faults like Loss of Frame Sync, Loss of PSTN connection, Loss of WAN Interface connection etc are indicated by LEDs (for internal inspection) as well as over SNMP and CLI for the Manager.
The following is a list of remote operations possible:
Remote Restart of subscriber Units
Remote Software Update
Remote Subscriber Unit diagnostics
Line quality test (BER test)
There are two basic types of subscriber units - the Basic Rate Digital subscriber Unit and the High Bit Rate Digital subscriber Unit.
Basic Rate Digital Subscriber Unit (BSX-200XX)
The Basic Rate Digital Subscriber Unit has one Ethernet port and one telephone port. It connects to the IAN over a pair of copper wires. The BSX-200 provides 144 kbps data connection while the telephone is not under use. When the telephone is used, the data rate goes down to 80kbps. This change is automatic and needs no user intervention.
There are two versions of the Basic Rate Digital Subscriber Unit. The BSX-200 is powered off the 230V mains as well as remote powered from the IAN. In case of the 230V failing, the BSX-200 enters a low power mode in which it keeps the telephone operational with the power from the IAN.
The BSX-200LP does not have the power feed from the IAN. It however has a battery that provides backup power in case the 230V input fails. The battery is a 12V, 1.3Ah battery. Standby time of 12 hours is expected from this battery. The BSX-200LP also has a in-built battery charging circuits to keep the solid lead acid battery charged properly while the mains supply is on.
High Bit Rate Digital Subscriber Unit (HSX-200X)
The High Bit Rate Digital subscriber units provide higher bit rates and more telephone ports. The basic HDSU is the HSX-200X that has just one 10BaseT port. This can provide upto 2 Mbps data rate to the IAN. By adding optional plug in modules, the number of telephone ports can be increased from zero to eight. Each telephone card plugged into the baseboard will provide 4 telephone ports. The HSX2000, HSX-2004 and HSX-2008 have 0, 4 or 8 telephone ports respectively.
All HDSU units are locally powered. In case of failure of the 230V Mains supply, these units have a 12V battery for backup power. In case the 230V power fails, the HDSU units enter a low power state wherein only one telephone port works. All the other telephone ports and the Ethernet port are shut down. All HSX-200X units need a 12V, 1.3Ah sealed lead acid battery for backup.
The HSX-200X units are all housed in a tabletop/wall mountable box with provisions for fitting the optional telephone cards. The Basic Unit is a Data only Module that has provisions to accept one or two Telephone Cards that have four Telephone lines each. By adding these cards as plug in modules, the number of telephones the HDSU can support can be four, eight or twelve. .
The reach of the DIAS is dependent on three factors - the wire gauge, the bit rate required at the subscriber end and the need for power feed. For various bit rates and gauges, the reach distances are shown here. There is also a column that lists the reach without power feed.
Reach on 0.8mm wire Network Powering
Basic Rate Subscriber Unit BSX-200.
The BDSU comes in two versions. BSX-200 is a line-powered unit. Under loss of power from the local Mains, the unit is powered from the line. In this reduced mode, the power consumed by the BSX-200 is 3W. The IAN feed power is 4.8W. The distance from the IAN to the BSX-200 now depends on the wire gauge that can deliver this power at the subscriber end. For various gauges, the expected distance is as shown below
Reach with Power Feed
In this section we discuss the technical features of DIAS, especially those that pertain to deployment and operation. We discuss routing and addressing issues, security, management, billing, configuration and stacking.
An IAN acts as a router considering each subscriber as a routing port. Hence each IAN requires one or more IP addresses (IPS). With the explosive growth of the Internet, an operator may face difficulty in obtaining a large number of IP addresses.
Hence, we categorize three types of services to make efficient use of the available IP addresses:
Class-A Service: Leased-line connection with more than one subscriber node
Class-B Service: Leased-line connection with only one subscriber node
Class-C Service: Dial-up connection Class-A Service
This class of service is similar to the leased-line connection available through an Internet Service Provider such as VSNL A set of IP addresses \Pι ... IP„), where n = 2 is assigned such that the addresses fall under the same subnet. One address from the set is assigned to each xDSU and the others are assigned to the subscriber nodes.
Users of this class of service can have their own domain name, (e.g abc.res.in). Individual machines on the subscriber's network can be assigned names like mad. abc.res.in, mac2. abc.res.in, etc.
Functionally, this class of service is the same as Class-A. But in the situation where the subscriber requires only one node to be connected to the Internet, the Class-A requires two IP addresses, one for the DSU and the other for the node. To overcome this problem, we slightly vary the scheme by allocating the lAN's IP addresses (IPs) to the DSU box and allocating only one IP address (say IP^ to the subscriber node.
Users of this class of service can have their own domain name, (e.g abc.res.in). This domain name is attached to only one subscriber node (node with address IPι), which we call the subscriber-Internet-node.
The number of effective nodes on the Internet can be increased by running a proxy server on the machine IP1, and connecting more nodes to the subscriber LAN, with locally significant addresses. These local nodes can then access the Internet indirectly through the subscriber-Internet-node.
The above two classes of services provide leased-line connectivity, and it is possible to associate a domain name with subscriber machines. However, for most domestic purposes, it is not necessary to have the above features. To reduce the number of IP addresses used by a DIAS system, Class-C is provided. Class-C service allows only one subscriber node, and the IP address for that node is allocated from the IAN from a pool of local IP addresses. These are not well known addresses and cannot directly go on to the Internet. In this case, a NAT server is required at the IAN side. The NAT server will translate all traffic from the subscribers (who will have local IP addresses) to a well known IP address. The subscriber will not have a domain name on his own. He will still be connected permanently to the Internet. However, the subscriber loses the domain name facility. This service will be useful for residential subscribers who would normally not have to host a web site or an email server. See the section Servers and IP addresses for more details on this.
In this class of service, no domain name is attached to the subscriber. This class of subscriber normally uses the Internet for e-mail and web-browsing. E-mails are accessed by logging on to the server of the ISP. The number of effective nodes can be increased by using a proxy server at the subscriber premises, assigned a Class-C address as described above. Now other nodes on Ethernet (at subscriber premises) can access Internet through this proxy server. This service will be available as a future upgrade.
Each IAN acts as a router, forwarding IP packets between the Internet and subscribers connected to its BDSUs and HDSUs. Each subscriber machine on the Ethernet port of a DSU is considered as one routing port on par with the Ethernet port and the E1 port on the IAN. Each interface is assigned a global IP address and each subscriber is assigned a global or a local IP address depending on the class of service, as discussed in the previous section.
In most cases, a subscriber has only a single connection to the Internet, through one specific IAN. For such cases, static routing is sufficient. The IAN maintains its routing tables independently using only information on IP addresses assigned to its subscribers.
There are two situations in which an IAN needs to exchange routing table information with other routers. The first is when a stack of lANs are interconnected via Ethernet to form a large DIAS. While the stack as a whole does static routing, each IAN needs to know the IP addresses of subscribers connected to other lANs in the stack. This exchange is done using the Routing Information Protocol (RIP).
The second situation arises when a subscriber has two or more connections to the Internet, to the same or to different ISPs (figure 4.3). This is likely with large organizations. In this example, there are 3 possible routes for each packet: via IANι, IAN2 or the dial-up line. Which one is used depends on the destination address and the prevailing congestion on each of the routes. Hence, the lANs need to dynamically update their routing tables. Again, this is accomplished using RIP.
Apart from RIP, most routers also use the OSPF (Open Shortest Path First) routing protocol. In order that the DIAS inter-operates with them, an OSPF protocol software upgrade is planned.
DIAS management can be split into two parts : management of the data-portion and of the voice-portion
As mentioned earlier, each subscriber is considered as a routing port, i.e., the Ethernet port on the DSU box is considered as an extended Ethernet port of the IAN. The data portion is completely SNMP manageable. All the standard MIBs associated with the entities are supported: MIB-II, PPP, RIP and routing-table. In addition, a proprietary MIB is defined for handling enterprise, call management, RADIUS, ML-PPP and the xDSL link.
This basically consists of managing V5.2. V5.2 is currently managed by another proprietary management interface. When two-wire interface is used with a SMUX, the management of the voice-portion is carried out using SNMP. Command Line Interface (CL|)
The IAN has an RS232 port, through which the DIAS could be completely managed. DIAS also supports the telnet protocol, using which one could login to the IAN and the complete CLI command set could be accessed remotely.
DIAS provides secure Internet access to subscribers. Rigourous checking is done with the subscriber, before allowing him/her to access the Internet. Since each port is associated with one DSU box, there is little chance of an intruder gaining access through that port.
The reasons are:
A proprietary protocol is observed between the DSU and DIAS server.
Each subscriber is associated with two IDs. One is the public-id, which is well known, and the other is the secret-id. Each DSU box is assigned this secret-id (assignment is done though proprietary methods) and the DSU box is checked for access only with this secret-id, which can be kept secret even from the subscriber. Accounting and access information are managed by the ISP using this secret-id of the subscriber. Configuring this secret-id cannot be done across the Internet. It can be done only through the CLI, using the privileged password.
Accounting and access information are accessed through RADIUS. RADIUS transactions are encrypted. This provides a secure way of managing the accounting and access information.
The DIAS system and its components (DSUs), are designed to require minimal configuration in the field. Most of the parameters and subscriber information are accessed through a server located on the Internet network. This method enables quick and easy replacement of a failed IAN, since most of the information is stored in the off-the-site server. This design strategy allows the ISPs to keep their downtime minimal and prevent loss of information due to failed units.
The other main problem with any system is upgradation. In lANs, software upgradation can be carried out remotely using TFTP (Trivial File Transport Protocol). The same mechanism of software upgradation is also extended to the DSU. One of the important issues with the DSU is that it resides in the users' premises and any wrong upgradation of the software would result in total loss of connectivity. One copy is fused in the factory, and has a minimal set of features that provide connectivity to the IAN. Hence, should there be a problem using the new software, the DSU will revert to disconnection. To avoid this situation, each DSU has two copies of the software the factory-programmed software. Using this factory-programmed software, the new software can be downloaded to the DSU.
Servers and IP addresses
An ISP must have a Radius Server connected to the IAN on the Internet for authentication and billing. Failure of connection between the IAN and the Radius Server results in disconnection of service to all services.
Deployment with unique IP addresses
No server other than the Radius Server is required if each IAN and each DSU / user machine (using class A or class B) have a well known IP address (a well known IP address is a unique address on the Internet globally recognized and is sometimes referred to as a permanent address). These machines are said to be directly connected to the Internet.
Deployment using local IP addresses
Quite often, sufficient number of IP addresses are not available and one such address cannot be assigned to each user. It is then possible to assign local IP addresses (not recognized outside the DIAS network) to each user and even to the IAN. The IAN will then have to be connected (on the Ethernet Port) to a NAT (Network Address Translation) Server.
All communications to and from any user as well as the IAN need to go through this NAT server. The NAT server converts local IP addresses into some well known IP addresses before passing them on to the Internet.
It is possible to configure a Pentium machine with Linux to act as a NAT server. In future the NAT server function will be integrated into the IAN.
Mail Server and Web Server
A subscriber with a permanent (well known) IP address connected all the time to the Internet does not require a separate mail server or a web server. The subscriber's machine that is connected to the Internet will usually perform these functions as seemed necessary by the subscriber (he can use his machine as a web server if he wishes to). If the subscriber has only a local IP address, he will not be able to host a web site or have a unique domain name. An ISP who provides only locally managed IP addresses should have an email server as well as a web server for the subscribers. These servers should have well known IP addresses and must be permanently connected to the Internet.
Stacking of more than one IAN is required in the case where a higher number of subscribers are concentrated through the same voice-E1 and data-E1 lines. The stacking strategy differs for voice and data.
Stacking of Data
Stacking of data is carried at over the Ethernet backbone. Four lANs can be stacked to serve a maximum of 240 subscribers. A proprietary stack- management protocol distributes the routing table to the lANs connected through the backbone. Thus, over a period of time, all the lANs will get complete information about all the other lANs that are connected to the backbone. One of the lANs connected to the E1 link of the ISP is considered as the default gateway by the other lANs.
For additional bandwidth and redundancy, two lANs can be connected to the ISP. When one E1 connection fails, the other IAN becomes the default gateway.
Stacking of Voice
A single voice E1 line, connected to the exchange, is shared by all other lANs in a stack as shown in the figure above. Each IAN is allocated a few slots on the E1 link. Individual lANs establish and receive calls using the slots allocated to them. Basically, each IAN acts like a drop/insert multiplexer for those slots. A Distributed V5.2 stack is run across the lANs.
It is important that a service provider using the DIAS has an estimate of the voice and Internet traffic of his subscribers and provides an appropriate number of E1 ports to the exchange and to the router. For n telephone subscribers, each having an e Eriang traffic, the total voice traffic generated is ne Eriangs. It is necessary that sufficient numbers of lANs are cascaded so that ne Eriangs traffic be carried on the E1 lines connected to the exchange.
If these switched circuits on the E1 link to an exchange are also to carry the concentrated Internet traffic to a service provider, the estimated Internet traffic must be added to the ne. Eriang estimated earlier.
For example, if 108 subscribers are to each have 0.1 Eriang voice traffic and the switched Internet traffic requires 8 x 64kbps connection, the total Eriang traffic out of the IAN to the switch will be equal to
(108 x 0.1) + 8 i - - -j 18.8 Eriangs One E1 line to an exchange would then be just about sufficient and provide a Grade of service (or QoS) of 0.5%, which is normally acceptable. If the number of telephone subscribers increase, or the Internet traffic estimates increase, one E1 line may not be sufficient.
The Always-On or permanent Internet Access provides immense bandwidth to the customer. However, the subscriber is unlikely to use the connection all the time and the Internet access uses bursty traffic. Since IAN concentrates traffic, it is important to obtain an estimate of the average Internet traffic generated by all the subscribers together and its variance. This is a difficult task and the estimate will be different based on the nature of subscribers. The usage pattern may in fact change over time.
It is therefore best to have an initial estimate of this traffic and then actually monitor the traffic. Using the actual usage pattern, dimensioning for Internet access is to be adjusted. This adjustment needs to be carried out on an ongoing basis to provide the desired QoS to the subscriber.
One could define an activity factor for every subscriber having Internet access. It would be computed using the peak rate available to the user and the actual data bytes transmitted (or estimates of this) by the subscriber during some peak hour. For example, if each user with 144 kbps Internet Access (BDSU without voice) is to transmit or receive (or estimated to transmit or receive) 5 Mbytes of data during a peak hour, the activity factor for each user would be
5 x 10δ x 8b / 144 x 103 x 60 x 60 = 7.7%
If the number of users is large (more than 20) it is unlikely that average activity factor would be as high as this. A typical 5% activity factor for BDSU would normally be a good initial estimate.
The HDSU Internet access is for the corporate client and usage is expected to be more. Let us assume a 25Mbyte data transfer by an average HDSU user in an hour.
Assuming 2 Mb per sec connectivity, the activity factor for the HDSU subscriber therefore works out to be
25 x 106 x 8b / 2 x 106 x 60 x 60 = 2.7% Thus, estimating that each HDSU, on an average, would transmit and receive 25 Mb of data in an hour, a 2.5% activity factor would be reasonable.
One can now calculate the average Internet data access rate required by an IAN supporting 48 BDSU and 4 HDSU subscribers with the BDSU subscriber having an activity factor of 5% ad the HDSU subscriber having an activity factor of 2.5%. The average data rate is
48 x 144 x 103 x 0.05 + 4 x 2 x 106 x 0.025 bits/sec = 0.5456 Mbps
Even after accounting for a large variance, one third of an E1 link may suffice. One can use ten 64 kbps switched circuits and add and drop circuits on a dynamic basis, depending on the total amount of traffic encountered each way by an IAN. Since the IAN is provided with 2 E1 links to the ISP, care has been taken to enable handling of much higher Internet traffic for tomorrow. Also, the Ethernet port of the IAN is capable of carrying Internet traffic to a higher bit-rate transmission unit. Thus the subscribers could presume to be connected on a virtually leased Internet line from their premises.
For a DIAS to provide good service, it must have adequate bandwidth to the ISP and the PSTN. Outlined below are some sample dimensioning calculations.
Assuming 0.1E per subscriber and 8 telephones per HDSU, we can calculate the 'total traffic per IAN, and hence the number of voice channels to the switch for 1% blocking. This is shown in Table for various configurations. In all the • cases, it is observed that a single E1 from the IAN to the switch is more than sufficient. A large number of slots are free for data traffic.
Capacity of the Voice connections from a DIAS to the PSTN
To improve the utilization of the E1 line, we could have, say, 4 lANs colocated with both voice and data cascaded through one IAN. Serving about 200 voice subscribers at 0.1 E, the voice traffic is 20E. This requires a full E1 for 1% blocking. Assuming 0.37 Mb/s average Internet traffic per IAN, one would require a second full E1 to the ISP.
Data traffic is bursty. Even during an active Internet session, the average traffic on a link is typically much less than the link capacity. We can then define activity factor, the ratio of average traffic during the busy hour to link capacity. Activity factor is likely to range from 1-10%, for subscribers with a single PC to an organization with a large LAN.
Given the activity factors of subscribers, the aggregate average data rate on the backhaul from the IAN to the ISP can be calculated.
Since the IAN concentrates traffic, it is important to obtain an estimate of the average traffic generated by all the subscribers together and its variance. This is a difficult task as the traffic depends on the nature of subscriber. The usage pattern may in fact change over time.
Starting with an initial estimate of the traffic, the dimensioning should be adjusted based on monitoring of actual traffic. The activity factor can be computed using the actual data bytes transmitted (or estimates of this) by the subscriber during some peak hour. For example, if each user with 144 kbps Internet Access (BDSU with no voice) transmits or receives 5 MB of data during a peak hour, the activity factor is (5 x 106x 8b) / 144 x 1000bps x 3600s = 7.7%
The HDSU Internet access is for the corporate client and usage is expected to be more. Let us assume 25 MB of data transfer in an hour. The activity factor works out to be
(25 x 106x 8b) / 144 x 1000bps x 3600s = 2.7%
One can now calculate the average Internet data access rate required by an IAN supporting 24 BDSU and 4 HDSU subscribers with the BDSU subscriber having an activity factor of 5% and the HDSU subscriber having an activity factor of 2.5%. The average rate is
24 times 144 x 103 x 0.05 + 4 x 2 x 106 x 0.025 bps = 0.3728 Mbps
For packet transmission over a single link, the delay increases as the average data rate approaches the link capacity (figure 8).
For the delay experienced by the Internet user to be acceptably low, the total round-trip delay contributed by the DIAS system should be limited to a small fraction of a second, say 0.1 second. IP packets range in size from 40 bytes to 1500 bytes, with an average of about 200 - 300 bytes. The average one-way delay of a packet over the back haul link is D = (P/C)/(1-μ), where P is the average packet size in bits, C is the link capacity in bps and μ is the link utilization.
In figure 9, we plot the average one-way delay, D for several user activity profiles, as a function of backhaul capacity. The activity profile is BDSU activity factor/HDSU activity factor. If we wish to have an average one way delay of 50 ms, then for the 1%:1% profile, 3 channels need to be allocated on an average from the IAN to the ISP. This could be achieved by permanently having 2 slots and diaiing-up additional channels during times of peak traffic. For the heavy traffic 10%: 10% profile, an average of 19 slots gives an average delay of 28 ms. Taking into account the statistical variation in total traffic, we could lease, say 8- 10 channels and dial-up additional channels only when traffic increases.
The DIAS system is sufficiently flexible to serve several different applications.
Internet Access Network
A DIAS system can be used to provide Internet access to subscribers over an area of a few km. square, independent of the voice telephone network. This may be useful where voice plus Internet is not feasible, owing to regulatory constraints. For instance, currently in India, only two basic service operators are licensed in each state, but there is no limit on the number of Internet Service Providers.
Such a system may be used in an organization situated on a large campus, which already has a PBX for voice and possibly Ethernet LANs in some buildings. A DIAS could be used to provide a campus-wide LAN with connection to the Internet.
Voice + Internet Access Network
DIAS is most cost effective when used to provide both voice and Internet access. This can be operated in several ways:
A single public operator— the basic service operator (BSO) provides both voice and Internet services.
Franchise — each DIAS is operated by a franchisee who provides voice + Internet access in a locality. The DIAS is connected through a BSO to the PSTN and Internet.
Private network— an organization could use DIAS to provide both voice and Internet services in its campus. Deployment Scenarios
DIAS can be deployed in different configurations depending on the subscriber density, and the nature of the voice and ISP networks.
In a dense business or residential area, there may be hundreds of subscribers within a radius of a few km. Several of these may be medium to large organizations. These could be served by a cluster of co-located lANs.
Since the aggregate average data traffic is likely to be of the order of 100s kb/s to a few Mb/s, a dedicated backhaul to the ISP of several Mb/s is called for. This could be provided by a 8 or 34 Mb/s radio link, or a number of leased E1 circuits from the BSO.
Dense with Isolated Pockets
It is unlikely that all subscribers will be uniformly spread over a single "dense" area. It is more likely that there will be a number of isolated pockets of perhaps a few 10s of subscribers located within 4-8 km of the cluster of lANs. Each such pocket is too small to justify an independent backhaul to the ISP.
Initially, the subscriber base may consist only of isolated pockets, perhaps separated by 5 km or more. It is then not feasible to use the master-slave architecture. Each pocket needs to be served by an independent DIAS. The data traffic is likely to be at the most a few hundred kb/s. Hence, the backhaul to the ISP could be provided by one or two leased 64 kb/s lines (or an ISDN line), with additional capacity being added when needed, using dial-up lines. Glossary of Terms
BDSL - Basic rate DSL Subscriber Line
BDSU - Basic rate DSL Subscriber Unit
HDSL - High bit rate DSL Line
HDSU - High bit rate DSL Subscriber Unit
Basic Rate - Term used in the DIAS to refer to a 144Kbps link
DIAS - Direct Internet Access System. The Integrated Multi Access Solution that provides both Voice and Data over the same pair of copper wires
IAN - Integrated Access Node or the DIAS Central Office.
SU - Subscriber Unit
E1 - A 2.048 Mbps 32 channel serial link used by various countries to carry
64 kbps channels, voice or data.
WAN - Wide Area Network
LAN - Local Area Network
IP - Internet Protocol
SNMP - Simple Network Management Protocol
PPP - Point to Point Protocol
Unit No Unit
SX-1000 IAN with SMUX for 30 voice subscriber lines
SX-2000 IAN with V5.2 interface for upto 60 BDSU or upto 20 HDSU subscribers
BSX-200 Basic Rate Subscriber Unit with Power Feed
BSX-200LP Basic Rate Subscriber Unit with Battery backup
HSX-2000 High Bit-Rate Subscriber Unit with 0 telephone ports
HSX-2004 High Bit-Rate Subscriber Unit with 4 telephone ports
HSX-2008 High Bit-Rate Subscriber Unit with 8 telephone ports
SRA-004 4 Port Telephone Module for HDSU Technical Specifications
IAN - SMUX Version
Number of BDSL Ports Maximum 24 subscribers per IAN Number of HDSL Ports Maximum of 20 subscribers per IAN (only 24 telephones can be used)
Number of Lines per BDSL Line Card 12 Number of Lines per HDSL Line Card 4
Number of Lines 24 2-wire Analog ring-down loop start
12Khz/16Khz pulse detection on 2 out of every 15 Analog Lines
Two numbers of E1 ports
One 10 Base T interface conforming to IEEE 802.3 LAN standard. RJ45 Connector, MDI-X mode (connects to Ethernet Hub or Switch) Line Interface
Basic Rate Subscribers
Number of Subscribers Upto 24 Basic rate (BDSL) 2-Wire Interface
High Bit Rate Subscribers
Number of Subscribers Upto 20 HDSL 2-wire interface
IP, PPP, ARP, ICMP, TFTP, UDP, SNMP
SNMP version 1, Local CLI over RS232
IAN V5.2 Version
Number of BDSL Ports Maximum 60 subscribers per IAN
Number of HDSL Ports Maximum of 20 subscribers per IAN
Number of Lines per BDSL Line Card 12
Number of Lines per HDSL Line Card 4
E1 port with V5.2 signalling as conforming to ETSI 300-347-1, ITU-T G.965 and Indian GR TEC G/VAN-02/01 Sep 96 Internet Interface
Two numbers of E1 ports
One 10 Base T interface conforming to IEEE 802.3 LAN standard. RJ45 Connector, MDI-X mode (connects to Ethernet Hub or Switch)
Cascade E1 Port Interface
Cascades multiple lANs, for voice communication
Basic Rate Subscribers
Number of Subscribers Upto 60 Basic rate (BDSL) 2-Wire Interface
Cable Pairs required One pair cable with 1200E max resistance
High Bit Rate Subscribers
Number of Subscribers Upto 20 HDSL 2 wire interface
Line Transmission Rate 528Kbps/1168Kbps/2320Kbps Cable Pairs required One pair cable
IP, PPP, ARP, ICMP, TFTP, UDP, SNMP, V5.2
BDSU With Power-feed from Exchange
Connector RJ 11 Distance 1Km on 0.5mm, 800m on 0.4mm Data rate 128 kbps. Switches to 64kbps when telephone is used
One 10 Base T interface conforming to IEEE 802.3 LAN standard. RJ45 Connector, MDI mode (connects to PC)
One 2 Wire voice port with RJ-11 connector
Dimension 23cm(W) x 23cm(D) x 4cm(H)
Mounting Wall Mounting
Input Voltage 85VAC to 265VAC, 50Hz/60Hz
Power Consumption 7W
Power feed from exchange
Power Feed Voltage 30VDC minimum
Power Feed Current 100mA
Sync with IAN Green LED
Ethernet LINK Green LED
Ethernet TX/RX Yellow LED BDSU with battery back-up
Connector RJ 11
Maximum loop resistance 1200Ω
Distance 5Km on 0.4mm
Data rate 128 kbps. Switches to 64kbps when telephone is used
One 10 Base T interface conforming to IEEE 802.3 LAN standard. RJ45 Connector, MDI mode (connects to PC)
One 2W voice port with RJ-11 connector
Dimension 25cm(W) x 25cm(D) x 5cm(H)
Mounting Wail Mounting
Input Voltage 85VAC to 265VAC, 50Hz/60Hz
Power Consumption 10W
Power Backup 12V, 1.3Ah Sealed Lead Acid Battery
Backup Idle Time 10 Hours
Backup Talk Time 5 Hours LED indications
Sync with IAN Green LED
Ethernet LINK Green LED Ethernet TX/RX Yellow LED Battery Low Flashing Green LED
Connector RJ 11
Data rate 512Kbps/1Mbps/2Mbps Decrements by 256kbp for 4 Telephone units and by 512kbps for 8 Telephone units.
One 10 Base T Ethernet data port with RJ-45 Connector in MDI Mode (Connects to Computer)
IEEE 802.3 LAN standard
Upto Eight 2 Wire voice ports with RJ-11 connector.
Dimension 25cm(W) x 25cm(D) x 5cm(H)
Weight TBD Mounting Wail Mounting / Table Top Power Suply
Input Voltage 85VAC to 265VAC, 50Hz 60Hz
Power Consumption 20W Power Backup Needs external UPS
Sync with IAN Green LED
Red LED - 512Kbps Yellow LED - 1Mbps Green LED - 2Mbps
Fault Red LED
Ethernet LINK Green LED Ethernet TX/RX Yellow LED