Ask any company if it would be interested in increasing its network efficiency and proactively managing its network, and the answer would be yes. Ask it how much it would be willing to pay for such benefits, and the answer is not as clear.
The cost of adding RMON capabilities to a large network can exceed $100,000, and the benefits are difficult to quantify.
Although vendors now offer companies more RMON capabilities than ever, many companies don’t use them. “Most companies still manage their networks in a reactive, rather than a proactive, manner,” said Elizabeth Rainge, an analyst at International Data Corp., in Framingham, Mass.
RMON probes sit on devices or network segments and collect information–such as available bandwidth, the number of retransmissions and which applications generate the most traffic–to help network administrators identify network areas that need a boost in performance.
The problem is that a wide variety of devices can slow network activity. A client workstation’s processor may be unable to keep pace with large file transfers, a LAN segment may have too many users sending information, a server may not have sufficient processing power to keep up with transaction loads or a frame relay line may drop packets when users transmit video files.
To pinpoint a problem, a company must deploy RMON probes at each potential trouble spot. LAN probes can cost as little as $1,500, but frame relay probes can cost $10,000 or more. Thus, a large organization with hundreds of LANs can quickly spend $100,000 to outfit its network.
RMON costs don’t end with probes, either. Running a probe consumes processing resources on each device, so a company may have to upgrade switches or computers.
Companies must also purchase a management tool, which costs $10,000 to $100,000. The tool transforms the information RMON probes collect into reports and graphs that illustrate network performance. Depending on how much data is to be collected and how often it is used, the reporting tool may require its own PC or workstation.
Last, the corporation has to train its staff to use the probes and management software. With competition for skilled networking personnel increasing, pay for such personnel is rising.
Many small and medium-size companies opt for the less expensive alternative of spending a few thousand dollars for a protocol analyzer. When response-time issues arise, they remove the analyzer, place it on the network and determine the source of the problem.
This option is not as appealing for large companies. Each time a problem arose, a company would have to send a technician to numerous locations, collect a mountain of performance data and then analyze it.
Vendors claim the savings from more efficient network design make RMON worthwhile. “To ensure adequate response time, many companies overbuild their networks,” said Ivan Shefrin, president of Response Networks Inc., an Alexandria, Va., supplier of service-level agreement software.
For instance, a company may identify 10 network segments that could be creating bottlenecks. Rather than use RMON tools to discover that six are overloaded, a company might simply upgrade all 10.
How many organizations have overconstructed their networks is difficult to determine, but one RMON customer has been able to avert major network upgrades. Arizona State University, in Tempe, relies on 20,000 Ethernet connections to support its 49,000 students and 12,000 faculty and staff.
The university decided that monitoring its network was important and purchased Concord Communications Inc.’s Network Health. Now, despite a large number of connections, the university has a modest network infrastructure: Shared 10M-bps connections to users’ desktops feed into an FDDI backbone.
Joseph Askins, the university’s director of data communications, said Network Health has enabled ASU to gradually upgrade select network segments, rather than completely overhaul the infrastructure whenever response-time problems arose. “By placing Ethernet switches in front of a few of our FDDI connections, we have been able to stay away from moving to a higher speed infrastructure like ATM,” Askins explained.
Avoiding application brownouts
Application availability is another major factor in justifying the cost of RMON. “Corporations want to avoid application brownouts–periods of time when they are not available to employees,” said Shefrin of Response Networks.
But determining how much bandwidth and computing power are needed to keep applications running is difficult. When mainframes were widely used, it was easy to gauge network usage patterns because users had limited options. Now, something as simple as workers attaching video clips to e-mail messages can create network slowdowns.
In addition, IT departments can no longer tightly control applications running on corporate networks. With the widespread emergence of department-developed and shrink-wrapped software, end users have more control and can add new applications without warning.
These problems were the driving force behind the development of RMON2. The first version of the standard could examine only bulk network bandwidth; the second iteration breaks down transmission load by application.
This allows companies to identify users who spend a lot of time downloading large files from the Internet and either give them more bandwidth or instructions not to download large files.
RMON2 probes have slowly been making their way to market. “RMON2 is now in the early adoption phase and will become more common during the next few years,” predicted IDC’s Rainge.
Suppliers are also trying to make RMON purchases less costly by developing probes that can be used on multiple network segments. Rather than spending $1,500 to monitor half a dozen LANs, a company can purchase one probe and use it on all six.
Users would like RMON capabilities incorporated in all of their network devices, simply turning them on when they wanted to use them. Although many vendors include some RMON functions in their switches, most have avoided including all capabilities.
“Embedding RMON in a switch increases a company’s product development time,” said Craig Easley, Optivity product line marketing manager at Bay Networks Inc., in Santa Clara, Calif. “Competition is so fierce and product life cycles so short that vendors prefer to get their products to market quicker than [waiting] to include RMON in them.”
Thus, companies must continue to balance the amount of monitoring they’d like to perform against the amount they can cost-justify.