We think of network and systems management as a group of disciplines that designs, configures, deploys, diagnoses, audits and projects network and application usage in a distributed environment. Making a network run takes technical expertise in inventorying desktops and servers, simulating networks and application loads, configuring network and infrastructure devices, drawing network topologies, and processing events and performance metrics.

Delivering the Basics

The elementary question, "Is it the network or is it the application?" has turned out to have an elusive answer. Nothing much has changed in terms of making network management work. Most managers are trying to a get a handle on what their networks are doing. The promised new services and future-predicting Java GUIs of prognosticators and product pundits have yet to be delivered.

Basic information about what devices are deployed on the network and what they're doing is still network managers' most pressing need. Products such as Ipswitch's WhatsUp Gold, Castle Rock Computing's SNMPc and even Microsoft Corp.'s Visio Enterprise offer the biggest bang for the network-management buck; they're simple and straightforward, and deliver what they promise.

These products provide autodiscovery inventory at Layer 3--the most cited reason for purchasing a network-management package. However, trying to get a logical Layer 3 topology is bound to start unraveling a network-management ball of string. Soon the vague Layer 3 discovery needs to be replaced by the much more difficult port-level Layer 2 discovery. It's at the port level of detail where physical meets logical, and the simple, relatively correct Layer 3 topology becomes the incredibly difficult, often-wrong port-level mapping.

Many vendors, including Aprisma Management Technologies, Entuity, Intel Corp., Microsoft, Peregrine Systems and RiverSoft Technologies, claim to offer Layer 2 discovery. But beware: Layer 2 discovery is easier to brag about than to accomplish. It's not necessarily the fault of the vendor, which queries the bridge table and proprietary MIBs of Cisco Systems, Nortel Networks and Extreme Networks. If those tables aren't correctly maintained, are part of VLANs (virtual LANs) or have unmanaged devices intervening in the managed path, it's hard (if not impossible) to be 100 percent accurate.

So the question becomes, how accurate is accurate enough? It's certainly not enough to diagnose a problem in which a 10 percent chance of error is possible. What about 95 percent accuracy? Maybe, but 98 percent is more comfortable. And so far, no product achieves better than 90 percent accuracy.

Companies To Watch

Services -- So Out, There Isn't Even a Box

Since you won't get out-of-the-box network and systems management anytime soon, it's time to do some outside-the-box thinking. Wouldn't it be nice, for example, if you could rent network management with no up-front costs and no search for experts?

If the network manager didn't give you what you needed--or even if what you needed were a moving target--you wouldn't have to fire, retrain or convert. You'd just sign a new contract, and like switching long-distance service--bim bam boom!--you'd have new network management.

MSPs have services that would make the above (bim bam boom!) possible. The establishment of the MSP Association in June, now comprising more than 60 members, is an indication of vendors' rush to offer these kinds of services to enterprise networks.

Two basic types of MSPs are emerging -- the really big and the medium sized. Small shops will continue using WhatsUp Gold and other ping monitors, at least for the foreseeable future.

The big MSPs, such as Nuclio Corp., are contracted by ISPs, and play an ever-increasing, behind-the-scenes role. Their services will come to many enterprises as SLAs offered by ASPs. Enterprise leverage in this case will be limited to that which you have with your ASP. Face it: Your ASP won't switch its MSP based on your complaint, unless you're a Fortune 50 company. Even then, these services will more closely resemble the flavor-of-the-month club than a long-term commitment.

So, then, what's the attraction of these "Venti"-sized MSPs?

Expertise and a blissful lack of baggage. They land the biggest contracts, in customer-rich environments, which they use to leverage the best management applications and talent. Meta Group says 30 percent of large enterprises are attempting to integrate disparate events under a single toolset. By 2003, according to Meta Group, 60 percent to 70 percent will have such projects under way (often as a side effect of pursuing an operational "command center of excellence"). In contrast, 90 percent of Internet-related service providers (such as ASPs and ISPs) possess consolidated event consoles--a result of having a standardized infrastructure--and have little legacy technology to support. Enterprises also have the added disadvantage of not practicing management as a core business.

Although the large-scale MSPs garner the headlines, the small MSPs have their own appeal. Providers such as SilverBack Technologies place a management appliance on-site, at an enterprise network, and act as the NOC (network operations center). The polling is local, and the outsourced NOC lets the enterprise helpdesk know what's going on, and vice versa. There's a monthly fee, but no other strings are attached. It's a model that has promise for small and midsize businesses, as such providers can offer 24x7 monitoring, as well as expertise, without up-front capital.