Enterprise data networks are getting increasingly complex. Much work is still done locally, by a user at a desktop computer, but increasingly the work, and the dissemination of the work's results, requires connections to other desktop computers and servers within the organization, and through WANs and the Internet to distant locales and external organizations. The productivity gains of all this connectivity are well known, but networks don't stay up all the time, and when they go down, organizations can experience heavy financial losses.

There are two types of network downtime: degradation, when a service is slower than usual, perhaps to the point of being useless, and outright outages, when a service is unavailable. The second is usually more serious than the first (although both cause productivity losses), but not always. A customer accessing the enterprise Web site may forgive--once--finding the site unavailable, but may become so frustrated at a slow-responding site as to leave in a huff and never return.

In our recent study, "The Costs of Enterprise Downtime 2003," we present case studies of six large organizations in different industries and the effects downtime has on them, in terms of both direct revenue loss and lost productivity. What follows are some key findings from that study.

The table below gives an overall comparison of our six case studies, looking at revenue and downtime, and the association between them. Companies that we interviewed represent some of the most advanced IT organizations in the world, and most can be very proud of their performance. Losses ranged from less than one-tenth of 1% of total revenue to almost 1%. Possibly most interesting is the cost per hour of downtime; when we look at that data, it is our health care respondent who comes out the worst. They are the biggest company in our study and have, by far, the largest and most geographically distributed IT infrastructure.

There are three cases in which the costs of outages and degradations are about equal--suggesting that many companies tackle the outages and degradations--and there are 3 cases that are weighted much more heavily one way or the other--suggesting they have handled one problem but not the other.

Most people expect that, overall, outages would be far more costly than degradations because service is interrupted completely, and for some organizations, this is in fact the case. Outages are easier to spot than degradations, though, and in many cases are easier to fix as well. In the case of the high-tech and travel companies, outages are something they look at across their entire infrastructure and attempt to eliminate.

Service degradations can be a hideous problem because they are difficult to track and difficult to fix, and require investment in technologies that optimize performance, which many companies are not willing to make because the financial benefits are generally unclear.

Most IT managers suspect that downtime costs a ton of money. They have a sinking feeling that money is draining from the company coffers every time a server crashes, an Ethernet switch flashes yellow and then goes dark, or a service provider sends a courteous e-mail a week after a T-1 went down. They are right.

That is the bad news. Most organizations don't know where to start or how to go about fixing problems when they find them. Tracking downtime is the first step, and that is something that networking products, management vendors and service providers could make easier.

Next, companies should look at outages and service degradations separately, and attack each with a different set of solutions (prioritized by how much the downtime costs them). Outages require high-availability hardware and software, redundant configurations, fault monitoring and reporting, and a responsive IT staff. Service degradations require analysis of users and traffic patterns, tuning of network devices, switches and servers, monitoring/management of applications and performance, and constant vigilance. Users will slog along while the network underperforms, muttering about how slow the "system" is, but most are unlikely to make helpful reports of slowdowns that can be acted on by an IT department.

All areas of downtime are worth investigating, but based on the results of this research, it is likely that many companies could seriously improve their bottom line by investing in technologies that decrease the number and duration of service degradations, because they have a huge (and often hidden) impact on both revenue and productivity loss.

Jeff Wilson is Executive Director for Infonetics Research, and can be reached at jeff@infonetics.com.