Manage your bandwidth

ATM is the way to go if you're looking for screaming bandwidth, but you'll need a set of tools to manage change, capacity, and complexity.

November 13, 1998
Manage your bandwidth
Main page | Toolbox | Specialized network tools

By Karen D. Schwartz

J.B. Hunt Transport Services just purchased a hot rod. The Lowell, Ark., transportation company has set up a new ATM network to connect its IBM mainframes and client/server network to run mission-critical administrative and planning systems over shared token ring hubs. The network promises more of everything. More power. More bandwidth. More trouble. But it's worth every bit of it, according to Ken Mangold, the company's network manager.

"This network gives us power we never had before. Those tweakable parameters you get with ATM are highly desirable in this environment. With ATM, we can really isolate some of the traffic from other areas of the company without impacting those areas," he says. "Some of our lower speed, high-volume traffic can be moved off the network where our business applications reside so we can get high availability and response time."

J.B. Hunt Transport Services

The company: J.B. Hunt is a $1.4 billion Arkansas-based transportation company with 8,000 trucks and locations scattered around the nation.

The problem: The sudden and major increase of its bandwidth thanks to ATM has created many new opportunities but also a serious management problem, especially centering on capacity planning and other management issues.

The solution: J.B. Hunt has standardized on IBM tools--NetView for AIX for the whole enterprise and the Nways set of ATM management tools for everyday network management tasks. IBM-compatible point solutions will follow.

The IT infrastructure: J.B. Hunt's IT configuration consists of an ATM network connected through shared token ring hubs to its IBM mainframes and client/server networks.

The company, which has a fleet of nearly 8,000 trucks and had revenues of $1.4 billion last year, is a prime candidate for ATM. For years, J.B. Hunt has increased sales and grown its business using a traditional 16Mbps token ring collapsed backbone network to run its administrative empire. Last year, as the trucking company started looking to a future that included more bandwidth-intensive applications like imaging, data warehousing, Internet applications, push technology, and faster processors on the desktop, Mangold decided the time was ripe to move to ATM.

The network as bottleneck

"As people get faster and faster processors on the desktop, your network becomes more of a bottleneck than it may have been with a regular Pentium 133 processor," Mangold says. "We wanted to eliminate that latency in the network, and the only way to do that was to get to a much higher speed than what we were getting with a standard source route and bridge-collapsed backbone and token ring design."

With the new communications tool, J.B. Hunt gets an embarrassment of riches. At its fastest rate of 622Mbps, ATM gives companies six times as much bandwidth as Fast Ethernet, and 600 times as much bandwidth as regular Ethernet.

But the bandwidth isn't free, and Mangold decided the best way to protect his investment was with a new set of tools. His company standardized on IBM tools, choosing IBM NetView for AIX as its enterprise manager, and the Nways set of ATM management tools for everyday network management tasks. But managing the network well may take more planning than Mangold had anticipated. "We have to manage all that extra bandwidth. We're starting to research additional tools to manage bandwidth capacity, performance tuning, and long-range capacity planning," he says.

Managing peaks and valleys

Capacity planning is a particularly important issue, since network usage varies throughout the day. "During prime time, we have to make sure that when the network peaks out, we can handle that," Mangold says. "It's the same thing with failover. If I've got a redundant network and I lose half of it, can the other half support the whole network during those peak periods?"

Mangold is researching specialized ATM tools to help him manage his network more closely, but such tools are just beginning to emerge (see "Niche ATM tools"). NetScout's ATM management tool isn't even on the market yet, but it's being investigated by several companies because its Ethernet management tool has been so popular. Other tools made just for ATM include Concord's Network Health, a combination capacity planning, traffic management, and network monitoring tool; ATM Sniffer Network Analyzer by Network General, which troubleshoots the network; and NetMaker XA from Make Systems, which performs network monitoring and design.

Many companies need the combination of an enterprise network management tool and a general-purpose ATM management tool. Enterprise management tools, which run closest to the operating system, can coalesce alarms and provide a variety of views of the network, all the way down to the hardware. Tools in this category include Hewlett-Packard's OpenView, Cabletron's Spectrum, Sun Microsystems' Solstice Enterprise Manager and Domain Manager, and IBM's NetView.

Move toward tools that are as self-managing as possible.

Sitting on top of the enterprise management layer are the general-purpose ATM tools--the class of tools on which ATM networks depend most heavily to provide everything from setting global policies, to detecting and reporting faults, to watching the flow of packets across the network. Tools in this class include Fore Systems' ForeView; IBM's Nways group of products; Bay Networks' Optivity; Cisco's CiscoWorks, CiscoView, and StrataSphere; and 3Com's Transcend.

Many ATM networks manage fairly well with nothing more than these two classes of tools. In cases where network performance suffers in one or two specific areas, network managers often can develop a homegrown solution to fix the problem.

For example, network engineers at DreamWorks SKG, the entertainment production company owned by Steven Spielberg, wrote their own scripts to deal with lost connections.

"We were having a problem for a while there with our [Silicon Graphics] Indigo 2 workstations, where we were getting the wrong memory addresses. That would lead us to suddenly lose connection to the server with no warning," says Suri Denduluri, an engineer responsible for systems and networking in DreamWorks' feature animation division. In an environment where animation artists work cooperatively over the network, dropped packets of data can mean dropped frames in a motion video sequence. To prevent glitches, Denduluri's staff wrote scripts to monitor the server. If it's down, "we mark that ATM interface as down, and the traffic is rerouted through an Ethernet interface until we can fix it."

This also maintains the highest level of availability at blinding speeds. "Timelines on movie production are tight," Denduluri says. "Even one hour of downtime translates into tens of thousands of dollars of production loss because our artists are very highly skilled and highly paid."

Know your network

To achieve a well-managed ATM network, assess your organization's present and future needs, advises Lap Vu, network administrator for Aetna's large ATM network. "Once you understand what is normal for your network, you can know what to expect," Vu says. "Only then can you find the right tools to tailor the network to your needs."

"A system with 10 CAD/CAM machines and a few servers might function well by connecting a single ATM switch to a bunch of adapter cards," notes Byron Young, director of product marketing for ATM at Cabletron Systems of Rochester, N.H. "You can probably manage that system using a text-based console pretty effectively, because odds are that you're not going to be doing a lot of moves and changes. If you want to know when you have an alarm, you can just look over and see if you have a red blinking light."

Systems such as the one Young describes lend themselves to homegrown solutions. For corporations with large, complex networks consisting of hundreds or thousands of ATM switches, the best way to achieve optimum performance may be to turn to one of the specialized ATM management solutions making their way to market. These tools tackle isolated network performance problems, such as performance monitoring, traffic management, and capacity planning, with software written specifically for those problems.

Tools like Concord's Network Health seem to be tailor-made for very large ATM networks, such as Aetna's 7,000 node network. The network uses Cisco routers in conjunction with IBM's NetView and OmniVision (now called X-Vision), a network management tool from Xylan of Calabasas, Calif., to handle both internal and external users.

The network is so large and complex that it would be virtually impossible to keep tabs on its usage without such a tool, Vu adds. "We need to know if something breaks right away, and we need to know if we are running under the gun or are running comfortably with the bandwidth we have," he says.

Network Health allows Vu to receive reports on trend analysis and traffic and capacity issues. "Before Network Health, we had to upgrade bandwidth on an emergency basis," he says. "We didn't know that we had a bandwidth problem until we got a call from a user saying he was having performance problems."

But for optimum performance, very large ATM networks may need tools even more specialized than all-in-one multipurpose tools.

Health First, a health care provider in Brevard County, Fla., has embarked on an ambitious plan to connect all of its hospitals, cost centers, and doctors' offices with a far-reaching ATM network based on Fore Systems' ForeView. Network manager Mark Amey says a network of the size planned by Health First simply couldn't be run efficiently and effectively without turning to specialized tools.

"The need to manage our ATM network rather strategically and well in a real-time scenario is key to us, because we are gearing up for an increase in bandwidth-intensive, real-time applications like medical imaging, videoconferencing, and telemedicine," says Amey. Along with basic requirements like reliability, scalability, and redundancy, engineers wanted to ensure that troubleshooting would be done in an automated and complete way.

Amey turned to two outside packages: Action Request View from Remedy Corp., which allows the network's SNMP station to open, process, and prioritize help desk tickets; and ATM Sniffer Network Analyzer from Network General, which allows engineers to analyze packets between the network's ATM switches and edge devices. With the sniffer, "you can pull cells off of your network between ATM devices and then analyze them to see what is going on," explains Steven Shim, a Health First network engineer.

Even Vu, who has been pleased with the help Network Health has provided, is considering augmenting his ATM network with more specialized tools. "Network Health only provides me with a link layer of statistics," he points out. "I need tools on top of this. I want to be able to get all the tools up to the protocol layer and the application layer."

Advice from the trenches

With the growing array of ATM management tools now available, it can be quite difficult to determine the combination of products to best manage your network. While organizations with a good depth and breadth of internal networking talent may very well be able to handle the job of setting up the ATM network and choosing the best add-on tools effectively, other organizations may be better off leaving that job to an outside consultant, says Geff Ambrose, senior network engineer for Wang Laboratories of Billerica, Mass. Wang, which offers consulting and integration services to corporations embarking on ATM network management, will assess the company's ATM network and determine how best to optimize it.

Turning to an integrator can help companies avoid the tendency to standardize on products from one vendor. By doing that, "we can take the best of breed from every category," Ambrose says. "There may be a software package that integrates very well into HP OpenView that might do the job better than the manufacturer's." Diversifying also helps companies avoid risking obsolescence when one vendor's technology falls behind.

No matter which route you take, it pays to heed advice from those who have been there. Be sure you have the network designed correctly from the beginning, Health First's Amey says. If that means bringing in a consultant who can best advise you on what you need, do it right away.

"For the most part, we are very pleased with how we set it up," Amey says. "But there were a couple of things we've had to reengineer, and reengineering a production network is a lot more difficult than getting it right at the beginning."

Don't forget about the simple things, Cabletron's Young advises. "You might have the best network management tools in the world, but if you only have one T3 line between two locations and it goes down, you're out of luck."

It also makes sense to move toward tools that are as self-managing as possible, notes Greg Marrow, brand manager for Nways at IBM. "Look for tools that are proactive and don't require you to spend a lot of time determining what the problem is," he says.

Wang's Ambrose advises you to "Look at the depth of management you really need. Some companies may say a certain link can go down and that's okay because it's not vital, but other companies might consider that link vital."

When you purchase ATM, your corporate highway suddenly grows by thousands of lanes. But nature abhors a vacuum, and that is very apparent on a newly expanded network. Wang's Ambrose advises watching your traffic closely. "Monitor everything that goes on with your switches to make sure you are making the most effective use of the bandwidth that is available," he says. In the end, you will get extra years out of your current setup and will keep what could become a congested nightmare down to a reasonably managed internal superhighway.

Main page | Toolbox | Specialized network tools

About the author: Karen D. Schwartz is a freelance technology writer based in Washington, D.C.

0 Comments (click to add your comment)
Comment and Contribute


(Maximum characters: 1200). You have characters left.