Cover Story

The Network Management Trap

by Art Wittmann and Bruce Boardman

Do you have a network management console? Do you use it? Are you happy with it? We didn't think so. Network management is in a terrible state (we'll spare you the New Jersey jokes). The products we tested are big, complex and expensive, yet don't come close to delivering on their promises.

They require either a dedicated staff to perform tons of customization (something only public network providers and very large international companies are willing or capable of doing), or they're underpowered for enterprise management.

We're talking about products that require a Unix workstation, fat wallets and a high tolerance for mental anguish, but for the most part miss the mark in the corporate enterprise. In the end, they bait you with promises of omniscience, through Management Information Bases (MIBs), pretty interfaces and the Simple Network Management Protocol (SNMP). Yet what you find is that there's very little to manage, unless you buy into that vendor's hardware as well: the ultimate trap.

Still, it's hard to point a finger at the network management vendors alone. The products are improving greatly. In the next few pages, we'll offer some hands-on Sneak Previews of upcoming products from Hewlett-Packard, Cabletron and Sun, and a review of IBM's NetView. But, they still need to work out some basic architectural flaws.

Why So Bad? Fundamental network management essentially amounts t o finding routers, determining the networks attached to them and then finding all the nodes on those networks--usually just a matter of pinging all of the available addresses or reading a router's Address Resolution Protocol (ARP) cache. Over time, the network management station provides "important" information by occasionally pinging nodes to see if they're still alive.

It's useful, but not brain surgery, right? Is it really worth buying a $30,000 Unix workstation and spending thousands more on software and committing a small staff to the task? Certainly not. Such products should cost about $500, run on any PC or workstation and just send you e-mail or a page when necessary. It's simple network monitoring, and most of us are barely getting this out of our management stations.

We have no problem with complex software or with tools that need some work to configureŭif they deliver value. But if it takes a small team of operators and managers to configure and maintain software that doesn't offer much payback, something is seriously wrong. Moreover, vendors are starting to include systems management. Since most network managers are not even satisfied with what they have, adding systems management is hardly an attractive option. Nonetheless, many organizations have separate groups responsible for network and systems management. They need those groups in each other's way like they need more printer-door-open traps.

And What About SNMP? SNMP through its Management Information Base (MIB) construct provides a way for vendors to make available any important piece of information about their network-connected device. While SNMP's ability to deliver that information is somewhat problematic, the bigger problem is that every device vendor has been left, like a fox in a hen house, to define its own device information.

While all network management vendors support two standard MIBs--MIB-I and MIB-II--there are no hub MIBs, no router MIBs, no bridge MIBs and no server MIBs supported by both network ma nagement vendors and equipment providers. This is the crux of the problem. If you've paid $20,000 for network management software and another $30,000 for the hardware to run it, you should be able to find a menu that says "Manage Routers" or "Manage Hubs."

That's the good news. The bad news is that it won't get any better. Virtually every management vendor has a vested interest in its entire product line. If you buy HP management software and want to manage a Bay Networks router or worse yet, a Sun Microsystems server, you're pretty much out of luck. On the other hand, if you've got a bunch HP servers, HP has a much better story to tell. Cabletron is good at managing Cabletron hardware, IBM is good with IBM hardware, a good Merlot goes nicely with a T-bone and so on.

The vendors see network management as a way to strengthen their hardware lines, or at least their priorities are to manage their own hardware first. They don't see their network management system as a product that should stand on its own and simply offer the best possible features for managing a network--any network. We do. Because vendors approach management with such duplicity, and because of the lack of standardized MIBs, these products don't ever seem to become better generic network management tools. To be fair, even if that were the goal, finding and incorporating all the private MIBs would be prohibitively difficult anyway.

Is There Any Hype, er Hope? New management standards for new network paradigms may soon come to our rescue. The Internet Engineering Task Force (IETF) has proposed various MIBs to manage networking hardware from bridges to Ethernet segments. These MIBs, however, are typically either unsupported by hardware vendors or not integrated into management stations, or both. So as we move from our broadcast-oriented networks to connection-oriented networks, we can only hope that network management is higher on the list of critical features than it has been. In fact, you must demand that it is. (If you hav e, take the rest of the day off.)

Until a new generation of management standards makes its way, we need to pressure network management and network equipment vendors into at least embracing the general purpose MIBs that do exist. At a minimum, every network management system should natively support RMON. Vendors have done a good job of improving their user interfaces and making their products scalable. Now they must get back to the basics.

Enterprise Management Is Just Around The Corner

HP OpenView OperationsCenter and AdminCenter

Changes in SunNet Manager Version 2.2.2

ATM and VLAN Management Will Require Strategy Shift

October 15, 1995