Bandwidth control answers SOS: a close examination of poor performance results in a cure - Testing and Diagnostics

Communications News, Sept, 2002

From Warri, Nigeria, to Vung Tau, Vietnam, to Kuala Lumpur, Malaysia, when the phone rings at International SOS, the call could be a matter of life and death. The large, comprehensive medical and security assistance company for travelers experiencing medical crises organizes emergency evacuations, surgeons and specialists. Critical administrative details like immigration clearances and passports are also handled.

With coverage in 41 countries across five continents, International SOS' work can be complicated and extremely time sensitive. Its systems must be up and running optimally to allow timely delivery of critical information when a major medical emergency strikes.

The company's global frame relay network plays a pivotal role. Company representatives from around the world use it to access a homegrown CRM application delivered via a Citrix server at the company's base in Singapore. When Jimmy Lim, International SOS' network manager, found that the mission-critical application was performing poorly across the WAN, he knew he needed to do something fast.

"We must be able to answer questions on the spot, especially when there is an emergency," Lim says, "but sometimes the links were so congested that the remote sites simply couldn't get access to needed information." Users often gave up on the network, and used the phone to enlist staff in Singapore to assist. "The inability to prioritize traffic created additional work for everyone," says Lim.

To find a solution that would give critical traffic top priority on International SOS' links, Lim enlisted reseller Vanda Solutions, whose answer to the dilemma was NetReality's WiseWan network application priority switch, designed to monitor, report on and control network traffic. What the switch uncovered about International SOS' use of its bandwidth resources surprised the company's IT department.

"We found that the majority of International SOS' bandwidth--about 70%--was being consumed by FTP, while mission-critical applications were grabbing only 10% to 20%," says Gee Meng, Vanda product manager. "While some users heeded the IT department's request and sent large FTPs during off-hours, others simply had to FTP on an ad hoc basis when the need arose. WiseWan revealed that files like a 150-meg marketing database were so big that even if sent during off-hours, they would still be clogging the links the next morning when the office opened and people needed to get their e-mail."

The solution also highlighted the massive network impact of NetBIOS over IP. Lira thought that traveling employees used NetBIOS occasionally to browse their home NT servers and pull files instead of initiating a proper FTP. "WiseWan made it clear that NetBIOS was frequently consuming bandwidth," he observes.

To solve the problem, Lim set up four policies:

First, because International SOS uses Citrix to deliver three mission-critical applications, Lim assigned high priority to all Citrix traffic. In the future, if the company decides to deliver less-critical applications via Citrix, Lim will utilize WiseWan's ability to see within the Citrix packet and assign different subapplications different priority levels.

Second, off-hours, FTP can take up as much of International SOS' pipes as it needs, but limiting it during office, hours assures needed bandwidth for mission-critical traffic.

Third, NetBIOS--which is not WAN friendly--is now limited to 12 kbps. Those who use it are not rewarded with quick connections. Says Lim, "We reiterated to users that we do not condone NetBIOS usage, and we have enforced a policy to control it."

And fourth, by limiting e-mail to 16 kbps, it can be delivered in a timely fashion without impeding more time-sensitive applications.

The solution's historical reports showed that the Sydney, Prague and Geneva sites were overprovisioned. "We'll be able to efficiently run all our mission-critical applications with only 50% of the current bandwidth," says Lira. Downsizing these links will save an estimated 15% to 20% of the WAN bill.

"We use WiseWan to check our network service level. We can verify that we are getting the promised CIR (committed information rate) and burst traffic for our PVCs (permanent virtual circuits)," offers Lim. "Our service provider really is giving us what we pay for.

"Before, we didn't have a view into our WAN traffic. When an issue arose, we questioned users as to whether they were doing anything that could impact performance. We were just guessing. The situation was out of control," adds Lim. "Now, we know what's going on, and we have a means to enforce policies and use our links to their fullest."

For more information from NetReality: www.mleads.com/209cn-259