Telescope 2007

Networking: Optimising the WAN

The data that flows across corporate networks is getting richer by the day and networks are becoming more complex as time goes by. However, connecting to remote locations is still a challenge. Be it through a VSAT link or a terrestrial one, every entity in an organisation has to be connected to the head office and to other locations—at least the important ones. And, when users multiply, it isn’t always possible to raise the bandwidth of WAN links to accommodate this surge. Aishwarya Ramani finds out how companies can and are tackling WAN bottlenecks.

Indian companies are making hay while the sun of high investment and outsourcing shines warmly on India. Businesses are flourishing, finances are looking healthy and clients are mushrooming. An obvious result of this rise in business is the vast amount of data that CIOs are required to deal with.

As data centres proliferate and companies shift to a centralised set-up, the ecosystem consists not just of servers that process and dispatch data over the network, but also of intermediate routers, file servers, application servers and satellite communications.

Hence, network bottlenecks are a common occurence. Bottlenecks occur for a variety of reasons but a common one is improper design of the network and bloating applications.

According to Akash Saraf, Joint Managing Director, Zenith Infotech, applications should be designed properly. Designers do not pay sufficient attention to designing and testing. This does not automatically declare the network architecture as squeaky-clean.

K V S S S Gunneswara Rao

K V S S S Gunneswara Rao, Director, VoIP, D-Link India believes that the network architecture and infrastructure need to be tweaked and tuned to get the desired performance.

T G Dhandapani

CIOs don’t necessarily agree with this contention. “The problem does not lie with the application nor is it the fault of the network architecture. It is the level of knowledge that is the hurdle,” feels T G Dhandapani, Corporate CIO, Sundaram Clayton.

Hilal Issar Khan, Head, IT, Honda Siel Cars India, attributes bottlenecks to improper network architecture.

Remote locations linked to the data centre by WANs are under pressure to support the sheer volume of packets that need to be transported. When a packet is transported over the network, latency comes into play contributing to bottlenecks. Latency worsens when there are multiple technologies communicating within a network. For instance, the latency issues of a satellite link are the highest.

“We use leased lines from different vendors, Tata Net provides us with our VSAT services, Tulip provides us the RF connection,” discloses Girish Rao, Head, IT Solutions, Marico Industries.

“Disparity adds complexity. I wouldn’t attribute bottlenecks to this,” comments Khan of Honda Siel.

Rahul Mahajan, Head, IT, Glenmark Pharmaceuticals says, “Bottleneck issues increase when different vendors are involved. Every vendor has his own way of designing his product. In such cases compatibility is always an issue, and it is highly likely the products may not work together. To overcome such situations we can deploy a mediating interface at the cost of performance.”

Sumit Mukhija, Business Development Manager, Cisco Systems India & SAARC opines, “Applications are generally designed to work in a LAN environment. But there is a huge disparity between LAN and WAN speeds. WANs cover larger distances and hence latency also becomes a critical factor when dealing with them.”

Prasad Babu

As organisations adopt VoIP and other convergence technologies, packets have to be treated differently depending on their tolerance of latency. Policies dictate the priorities that are set for different types of data. “You probably would set a low priority for your e-mail data, a slightly higher priority for video and the highest priority for voice,” adds Prasad Babu, Director SE and Operations, Juniper India.

The converged network has to be fine-tuned so that each category of data is given its fair share of bandwidth.

According to Rao of D-Link, VoIP has attained the status of a mature technology, and positioned to play a dominant role in the converged next-generation network architecture. Weaknesses in a data network will be magnified when IP telephony is introduced.

Are IT satraps truly concerned about WAN bottlenecks? While IT departments do seem content with the bandwidth of the WAN links, they feel that things may not remain hunky-dory in the near term. “We are looking to deploy solutions like Juniper’s WAN Acceleration technology,” comments Pradeep Kalra, Head, IT Infrastructure, Yes Bank.

CIOs are unanimous that bottlenecks are aggravated in the last mile. The terrestrial links that connect them to the service provider are susceptible to bottlenecks. Deepinder Bedi, Director, Sales, Tulip India agrees with this view. “Last mile connectivity does contribute to bottlenecks. Sometimes, unavailability of the last mile also causes bottlenecks,” he points out.

Another factor that must be taken into account is that of a sudden increase in the number of users accessing the network. For instance, this happens at the end of every month, when the payroll is being processed. Most organisations go for on-demand bandwidth when they are faced with such situations.

Rao of Marico, however, explains the catch. “There exists a maximum cap up to which the service provider can increase the bandwidth,” he says.

Average WAN links at most organisations are in the range of 256 Kbps to 2 Mbps. While most CIOs aren’t bothered over much about WAN bottlenecks they are not averse to the idea of deploying WAN optimisation technologies in the future.

In the meantime, IT heads are making arrangements to deal with the current causes of bottlenecks. Some common policies followed are to allocate higher bandwidth to those areas where user concentration is higher. Marico, for example, uses two 4 Mbps links and one 2 Mbps link for communication between the data centre and the head office. Marico’s Bandra office houses about 250 employees and hence the high bandwidth allocation.

Service providers are doing their bit to help their customers deal with the bottlenecks in the last mile. Tulip, for example, offers their clients a wireless last mile. “Setting up a wireless link is simple. Besides, increasing the cap in case the customers are in need of extra bandwidth is easier in the case of wireless,” says Bedi of Tulip. He also observes that organisations are going in for MPLS.

Best Practices

Rao of DLink says, “Test, Test, Test; Voice is a mission-critical and time-sensitive application and you have to look at issues in each of the OSI layers.” Mukhija of Cisco gives his checklist of dos and don’ts saying that one needs to go for solutions that integrate transparently into the network and preserve all port numbers, IP addresses and other network characteristics. He also adds that users need to go for solutions that can provide WAN optimisation and application-specific adapters such as CIFS (Common Internet File System) and print services in one box.

Downward compatibility is essential. “Any new infrastructure introduced in the network for application acceleration and WAN optimisation should be compatible with the existing infrastructure and be able to take advantage of the investments in the routing and switching products that you have already made,” says Mukhija. He also suggests that WAN optimisation and application acceleration solutions need to be embedded as service modules into your existing routing and switching hardware to enable a quicker RoI.