When it comes to boosting application performance, it's about a lot more than just the bandwidth.
"Our network is scattered across the country with 82 field offices and 10 regional locations. Our field offices don't have local servers, and they had pretty slow lines," says Chris Finucane, CTO for the US Department of Health and Human Services, Office of Inspector General in Washington, DC.
"We optimised the network and upgraded all our links to T-1 lines, and the field offices didn't notice a huge difference. That's when we realised bandwidth can't be everything and we needed to investigate other means to improve application performance."
Finucane's challenge is not unique, according to industry watchers who say today's increasingly complex applications simply aren't designed to run smoothly across large distributed networks that support branch, remote and mobile workers.
That has network executives scrambling to invest in new technologies - such as application acceleration and WAN optimisation tools - and others working with system administrators and application developers to tweak internal servers and fine-tune application code.
We spoke with a handful of network professionals and industry analysts to get their advice on how to boost application performance across a network.
1. Emulate WAN conditions
Sometimes the best defence is a good offence. That's the premise behind products that simulate network - specifically WAN - conditions to help network managers (and application developers) get a handle on how an application will behave on a given network under certain conditions.
Robert Wieters, enterprise network architect at cosmetics company Mary Kay in Dallas, says his staff uses Shunra's Virtual Enterprise network simulation tool to "create a mock-up of the network, configure typical behaviours and stress the application in that environment to discover where challenges exist."
Products such as Virtual Enterprise, Itheon Network Emulator, Network Nightmare and Candela Technologies' LANforge enable network managers to work with application developers and educate them about network latency, application round-trip times and other network-specific metrics that application stress testing on a LAN-based server wouldn't include.
"A little thought on the application development side can have a huge impact on performance," says Joe Skorupa, research director at Gartner. He adds that such devices reduce application adjustments post-deployment, which not only improves performance but reduces labour and costs.
"Overseas bandwidth is expensive and development/deployment becomes more costly with each iteration to make the application successful," Wieters explains. "My team would really like to deploy with as much performance efficiency as possible."
2. Study application traffic
You can't manage what you can't see. Network traffic analysis products from Fluke Networks, Network General, Network Instruments, Network Physics, NetQoS and NetScout can deliver a picture of protocols, packets and application traffic crossing network lines.
This information can be invaluable when determining which applications consume the most bandwidth and which deserve higher priority for business needs.
"The No. 1 thing is that whatever tools are in place, they should tell you what traffic is actually on your network," says George Hamilton, director of enterprise computing and networking at the Yankee Group. "Having that visibility is the first step toward larger [QoS] initiatives and ultimately filtering out the less critical application traffic."
Michael Nix, assistant director of IT services and communications technologies at the Kansas University Hospital Authority, says understanding the traffic flowing across network links can help IT staff decide whether certain types of traffic can be eliminated or if they should consider different application delivery approaches.
"If we look at methodologies that eliminate traffic from the network, for simple problems like streaming media, particularly to large numbers of people, we can explore multicast, which is a send once, distribute to many technology," Nix says.
3. Put a terminal server in place
Don't send what you don't need. Often between data centres and branch offices, a lot of unnecessary data travelling from headquarters to remote offices clogs network lines.
Chris Majauckas, computer technology manager for Metrocorp Publications in Boston, which distributes Boston Magazine and Philadelphia Magazine, says he put Microsoft's Windows 2000 Terminal Server in place to "get rid of excess data going back and forth between Boston and Philadelphia."
The product, similar to those available from Citrix, lets IT systems do work and compile results in Philadelphia, and only send the results to be viewed by staff in Boston.
"I am trying whenever possible, with as many applications, to decrease the data that needs to travel between offices," Majauckas says. "Terminal Server works on that premise. End users don't need to know the data is not in Boston, when all they want is to be able to view the results."
4. Invest in WAN optimisation or application acceleration technology
If the pain is acute enough, technology designed to compress, cache and speed application traffic could be the fastest fix and help avoid an investment in bandwidth.
For John Shaffer, CIO at Greenhill & Co, a global investment banking firm in New York, DS-3 lines cost about US$6,000 per month and he still wasn't getting the type of data transfer rates he wanted on Messaging Applications Programming Interface and Common Internet File System traffic.
The company had just consolidated servers to a central data centre and he was dealing with users accustomed to having their applications delivered from local servers.
"I was looking primarily at ways to reduce bandwidth costs but knew it was also an opportunity to clean up the network and optimise performance," Shaffer says.
"Exchange was a problem. If your system is on a LAN and you move it to a remote location, there is going to be some contention. There is no way to mask that from end users."
Shaffer invested in Riverbed Steelhead appliances, five of which cost less than his yearly DS-3 bill, to accelerate traffic to the company's distributed locations in Dallas, London and Frankfurt.
"We achieved upwards of 80 percent to 90 percent optimisation in our data transfer rates," he says. "It's not exactly like being on the LAN, but it does speed the traffic for less cost."
5. Investigate the apps on your servers
Not all application performance issues stem from long-distance trips over the WAN. Sometimes applications act up right on the local network.
Bret Moeller, CIO at Bunker Hill Community College in Massachussetts, says server monitoring software from Heroix enables him to monitor application performance, react to problems in real-time and plan for the long term.
He says the software monitors multiple operating systems and can help IT staff determine when to add memory, increase disk space or upgrade CPUs to ensure application availability.
"We know when a process may be using too much memory, for instance, and can take short-term measures such as a reboot or a configuration change followed by a more strategic measure, such as a hardware component upgrade," Moeller says.
Yankee Group's Hamilton adds that using monitoring software to find and remove unused applications on servers can improve performance of mission-critical applications.
"People are great at deploying applications and terrible at decommissioning them," he says. "These apps can be running on servers, using resources and not serving the business in any capacity."
6. Upgrade to MPLS
For those planning a network upgrade, consider MPLS, which has gained its greatest prominence among carriers looking to consolidate voice, data and video networks.
Robert Whiteley, senior analyst at Forrester Research, says network managers could couple two initiatives - convergence and application optimisation - in one fell swoop by upgrading to MPLS.
"An MPLS WAN helps people in a couple of ways. First, most people can actually get twice as much bandwidth for about a 5 percent increase in costs by getting rid of their frame, ATM or private links," Whiteley says. "More importantly, MPLS also lays the foundation for traffic engineering and quality of service that allows for performance optimisation."
John Hines, IT manager at Crafton, Tull & Associates, is thinking about MPLS, mostly because Verizon has planted the seed in his brain. Hines six months ago invested in a Packeteer PacketShaper appliance to speed traffic among offices on the company's star network that supports remote locations throughout Arkansas, Alabama and Oklahoma, but now is considering MPLS.
"MPLS can help me shape and prioritise bandwidths," he says. "I am talking to Verizon about switching, but I am waiting to see if MPLS becomes a natural next step."
7. Become the network "Renaissance person"
Don't wait for the application developers to come to you. Seek them out and share your network insight. Kernels of knowledge about network hops, round trips, latency, bandwidth consumption and response times can help developers build better applications and reduce your headaches when applications go live.
Industry watchers say that not only have the job duties for network managers evolved to include oversight of application performance, but a new role is emerging.
"We are calling the new position the Network Renaissance person," Whiteley says.