W2K Launch is What ENT Has Been Watching For
ENT began in 1996 with a mission to cover Windows NT in the enterprise. Over the years that meant covering departmental deployments and applications, NT as a network operating system, a print and file server, and even desktop software, which has been an important driver of Microsoft Corp.’s growth in corporate servers.
That’s the way NT was used in the enterprise. But the real reason for our being was, and is, to cover Windows NT in enterprise-scale deployments as a corporatewide messaging server, a database server, an online transaction processing system, or other major back-end system. In other words, in a capacity that would compete in the market space traditionally reserved for Unix and proprietary minicomputers. For a long time we’ve been crying in the wilderness. Users suffered with reliability problems and ran into scalability limitations. Microsoft's competitors laughed about the Blue Screens of Death. Conventional wisdom was that Microsoft was not there yet and had a long way to go.
It’s a view we shared. We haven’t been cheerleaders for Microsoft as the only solution or the best solution for every installation. Yes, we’ve believed this operating system would crack the toughest enterprise applications one day, but we kept a level head and always told you that time had not yet come.
We were there when Service Pack 2 proved a cascading nightmare for many end users. We were there for Scalability Day -- which was one of Microsoft’s more memorable PR flops -- with its smoke-and-mirrors attempts to convince users it was ready for enterprise deployment when it wasn’t. We sat in, and kept you informed, as Microsoft failed to get demonstrations working during keynote presentations. We covered the problems with Terraserver, Microsoft’s SQL Server 7.0-based Web site that was supposed to demonstrate the scalability of the RDBMS/NT platform.
There were reasons we thought Windows NT would make it. One was Microsoft’s track record of introducing poor products that slowly but surely turned into very good products. Then there is Microsoft’s effective marketing machine, its dominance of the desktop market, and its willingness to leverage that strength with server-level lock ins. The pace of innovation at Intel Corp. and among the server hardware vendors cannot be forgotten. Most importantly, though, there is the low cost of the Wintel platform based on the extremely stiff price competition among the system OEMs.
Over the last year, there has been real reason to expect that Microsoft is finally approaching an enterprise-capable operating system. The Windows 2000 Beta 3 release was remarkably stable. Microsoft was saying -- and doing -- all the right things with its beta testing program: farming out the system to hundreds of thousands of testers, and running longer term tests on the operating system to turn up some of the kinds of memory leaks that plagued NT. The company also became stricter about hardware and software compatibility requirements.
At the same time, a dichotomy was emerging among Windows NT 4.0 users. While many users continued to encounter serious problems, those who deployed NT 4.0 with Service Pack 3 or greater and followed accepted data center practices were getting remarkable reliability and performance out of their Windows NT systems. Recently, the Aberdeen Group (www.aberdeen.com) concluded that NT 4.0, properly run, was a stable enterprise operating environment -- a reversal of the firm’s earlier position on NT. Additionally, the emergence of Intel of a reasonably linearly scalable eight-way chipset out of Intel, and the demonstration that Windows NT can keep up, has been encouraging.
Microsoft’s launch of Windows 2000 reinforces what we’ve been predicting all along: Microsoft’s real debut in the enterprise data center is here.
To be sure, Windows 2000 is only a first step, with limited overlap with lower-end Unix performance in the data center. But if Microsoft’s completely new SKU, Windows 2000 Datacenter Server, scales effectively to 32 processors and provides up to four-node failover clustering as promised, it will take Microsoft higher in that arena.
Launch doesn’t translate to deployment. Predictions for enterprise-level deployments as a network operating system call for Windows 2000 usage to really pick up next year. Expect glass-house deployments, -- outside of Microsoft’s bleeding edge partners -- later than that.
Nonetheless, this takes us into a new phase at ENT. We go from observing Microsoft’s approach to the glass house, to covering Microsoft’s actual performance in the datacenter. Hang on. This will be quite a ride.