WAN Choices

Before we start evaluating choices, let us first setup a working definition of a WAN. A Wide Area Network, for our purposes can be described as a multiple individual networks, preferably at different locations, connected together over public networks such as the Internet. Our focus here is on the links rather than on the component networks. These connects are normally available round the clock, but could also be for specified times or on requirement. The bandwidth available on the links could be fixed or could vary across time or with demand.

So, when we talk of WAN choices, we are talking of the method of interconnection and issues with them. The biggest concerns in setting up a WAN link is not the distance or the bandwidth, but the latency, or delay that will occur as data travels from one end to another and the throughput. Applications and business needs have their latency limits. Business needs with latency limits? Consider a stock broking firm with offices across the country. If they opt to go in for a WAN connecting all their branches with a central server from which, all orders are placed, then too much of latency could lead to wide disparities between the price at which the client placed an order and the price at which it was executed given the rapid changes in stock prices during trading time. Application latency requirements come into play when users access applications like the ERP or Production management system over the WAN. If the latency is beyond the tolerance of the application, then the user inputs will simply time out, much like what happens on a bad connect with web sites. Similarly, your applications will have a throughput requirement, as in the volume of data that needs to be sent across. The connect option as well as the vendor you choose will have to be one which guarantees the latency as well as the throughput. Typically, you would enter into a SLA (Service level agreement) with the vendor, defining the levels of service that needs to be provided, with prices and penalty clauses.

What about future needs? Unlike in the days before the Internet, when you had to build capacities upfront for potential future use, today, you can afford to buy bandwidth in installments, even at relatively short notice. Only, check with your vendor before hand on his network plans and availability. Again, you need not contract for your full peak demand all through. You could for example, contract for higher bandwidth for two hours during peak business hours and for lesser bandwidth for the rest of the day. You could even negotiate a setup where for most of the month the bandwidth is a constant, going up at the end of the month, when you reconcile accounts and inventory.

Your job is not over once you have contracted with a service provider. You need to constantly measure whether you are getting what you have contracted for, bandwidth, latency and throughput. There are many tools for doing this, and some of them have been covered elsewhere in this issue.

You would use these measurements to ensure that required service levels are met and would have to reconcile them with the vendor's own measurements to work out the SLA linked payments and deductions. Obviously the reconcilement has to happen as

and when appreciable variations happen and not at a later time. 

Typically, a WAN link would be made up of multiple components, with different underlying technologies. If you are on a land based network, the leased line from the service provider's POP (point of presence) to your location could be a leased line, while the service providers network would be partly leased and partly public networks. Unless your data traffic is very 



high, it does not matter what the underlying technologies are.

Other than the land based option, the other option is to go in for satellite links. While the

obvious advantages are of not having to lay out physical infrastructure, and in the ability to provide connectivity to remote locations, the key problem area 



is in latency. The latency is caused by the sheer distance that a data packet has to travel, to the satellite and back; a distance of seventy thousand plus kilometers. There are also throughput issues with
satellite links.

By Ankit Kawatra, Anoop Mangla, krishna Kumar, Sushil Oswal

Frame Relay

Frame Relay is a HDLC-based packet switching technology operating at the physical and data link layers of OSI. It uses variable packet length for transmission to multiple locations simultaneously and statistical multiplexing techniques to control network access. The packets are switched between the networks until the destination is reached. Frame Relay is a correction-less system that does not attempt to perform any corrections itself in the packets. If it encounters any framing errors, the frame is simply dropped and transmission continues. It is left to the underlying protocol (TCP/IP, etc) to perform the corrections. The Frame Relay packet structure matches the standard sequence used by SDLC, HDLC, X.25 and other systems, containing the routing address, control flags, the data itself and the CRC checksum.

Two devices are used for transmission, Data Terminal Equipment (DTE) and the Data Circuit Terminating Equipment (DCE). DTE includes personal computers, routers and bridges while DCE are the transmission infrastructure owned by the service provider. Frame relay uses either Switched Virtual Circuits (SVC) or Permanent Virtual Circuits (PVC) for data transfer. In either case, bandwidth remains under-utilized for long periods at a time since virtual circuits remain open regardless of activity, which would be sporadic at best. Frame relay networks can transfer data at speeds between 56 Kbps to 1.5 Mbps with transmission delays in the order of a few milliseconds (compared to a few seconds with X.25 networks), comparable to direct-modem networks. Frame Relay an ideal mechanism to easily implement in high-availability scenarios like polling systems, file transfers and voice transmissions. 

Digital Subscriber Line 



Digital Subscriber Line or DSL is a packet switching technology and fifty times faster than dial up. Voice and data gets transferred simultaneously over the existing copper telephone by

using a different frequency range for each on the same line. 



Data is transferred on higher frequency and voice on lower ones. The technologies that make this possible are the DSL modem and DSL Access Multiplexer
(DSLAM). 

A DSL modem on your end sends data over the telephone line to your service provider's central office. At the provider's end, a DSL Access Multiplexer (DSLAM) terminates and aggregates incoming ADSL lines. It then redirects the voice traffic through the normal telephone network (PSTN) and the data through the high-speed digital line that connects to the Internet. 



DSL is also known as xDSL, with the 'x' standing for various kinds of DSL technologies. Some of the popular ones are ADSL (Asymmetric DSL), ADSL Lite, HDSL (High Bit Rate DSL), SDSL (Single DSL) and so on. 

DSL offers speeds up to 1.5Mbps and is the popular choice for bandwidth-intensive applications like multimedia communications, video-on-demand, Internet and intranet access, and remote LAN access on your PC.