Windows shops: Start preparing for .Net

January 31, 2001 —

In case you haven't noticed, Microsoft Corp. has been flooding the market with previews of its new offerings. Product breakouts in the past few months include Office 10, Windows Whistler, Exchange Server 2000, Datacenter Server, Host Integration Server, Internet Security and Acceleration Server, Application Center 2000, BizTalk Server 2000, Commerce Server 2000, the .Net SDK (software development kit), Visual Studio.Net, and the C# (pronounced "C Sharp") programming language.

Everybody loves a party, but there's one big problem with this massive rollout: How does corporate IT make sense of all these overlapping announcements, previews, and beta tests? It's obvious that Microsoft is encouraging a change of direction. IT needs to figure out where Microsoft is going and decide whether it wants to come along.

Developers rule

The last energizing jolt that Microsoft handed to the Windows developer community was the pairing of Windows NT 4 and Visual Studio 6. That was five years ago, an eternity in technology years.

Developers are the chief beneficiaries of new Microsoft technologies. All the fuss over .Net and Visual Studio.Net confuses some nontechnical people, who often make the mistake of seeing these as vague statements of mission, mere demonstrations of technology that might play a role in Windows' future. In fact, .Net and Visual Studio.Net are real, working products, intentionally targeted at Microsoft's neglected audience: developers.

Microsoft hit that target with a sniper's accuracy. The beta release of the Visual Studio.Net IDE (integrated development environment) crystallized elements of the .Net platform. It solidified in particular the C# programming language, a standards-based C++ dialect that delivers most of Java's benefits without Java's learning curve or license restrictions. Developers are snapping up free 120-day evaluation copies of Windows 2000 and SQL Server 2000, just so they can explore C# and .Net. Microsoft is keeping the heat on by having high-level Microsoft staffers, instead of the usual non-Microsoft moderators, involved in .Net online discussions. Those discussions are active and often contentious, a very good sign where technology is concerned.

Scaling up

The centralized computing movement took hold faster than Microsoft expected. While large companies are moving to monolithic Unix servers with dozens or hundreds of processors, Windows has been unable to compete. You scale Windows by adding more servers, something that modern-day IT managers want to avoid. Microsoft's clustering and load-balancing technologies are fairly mature, enough so that Web farms of Windows servers are commonplace. Still, Windows needs in-the-box scalability.

Windows' key limitation is not software, but Intel's rather tired 32-bit architecture. Some PC server vendors have managed to squeeze eight Intel CPUs into one cabinet, but dual-and quad-CPU servers remain the norm. To go beyond that, Microsoft is counting on Intel's 64-bit architecture, dubbed IA-64. It's worth noting that AMD, which ignited the performance PC market with its Athlon 32-bit CPUs, is working on a competing 64-bit architecture for Windows. Users will get an immediate boost from the move to 64-bit processors, but faster, wider Intel and Intel-compatible CPUs will be severely hampered by antiquated I/O bus technologies.

Massive multiprocessor Unix systems move data among processors and peripherals as fast as 1GBps. That's nearly 10 times the speed of the 32-bit PCI bus, the reigning standard for PC server peripherals. Gigabit Ethernet, IEEE-1394 ("FireWire"), Ultra 160 SCSI, and USB 2.0 boost the speed of communication between the CPU and external devices. It will take similar innovation on the internal bus -- not double or triple PCI speed but at least an order of magnitude -- to push PCs into the arena of massively parallel processing systems dominated by Unix servers.

On schedule

Microsoft has surprised developers and analysts with the high quality of its beta releases. With few exceptions, the technologies Microsoft has previewed in recent months have been stable and feature-rich. This has created an unusual problem: Companies are putting Microsoft's prerelease technologies to commercial use. They do so at their peril, because Microsoft doesn't provide support for beta products. It is telling that some companies find next-generation Windows technologies so compelling that they cannot wait for them to ship.

Migrating to .Net step-by-step

1. Enterprise developers and contract coders should grab the latest .Net beta and start learning. C# isn't necessary, but it is the best choice if you're currently using JScript, C++, or Java. Visual Basic and Visual C++ have been overhauled for .Net as well.

2. IT managers should sit down with key suppliers and business partners to devise a Windows 2000 migration plan. The final .Net should run on Windows NT servers, but Windows 2000 is a far better enterprise operating system.

3. Companies running massive Unix servers should stick with them for now. Keep an eye on the horizon for 64-bit PC servers, and look for massive multiprocessor PC servers one or two years later.

4. Companies running clustered or load-balanced farms of 32-bit Windows servers are well-positioned for .Net. Lay in for speed upgrades -- faster CPUs, more RAM, faster disks -- rather than new servers. Plan your next major server purchase around 64-bit machines.

5. If you're holding out for a cross-platform .Net, Microsoft is only hinting so far. The buzz is that Linux will be the first target, although Solaris would make better business sense. Nothing will happen until .Net is firmly established on Windows.

For developers, the time to jump on new Microsoft technology is now. Microsoft is giving developers some extra time by shipping second-round betas and playing coy about shipping dates. Existing developers can ramp up by grabbing the latest .Net beta, which now runs on Windows 98 and Windows Me systems. The server components still require Windows 2000. If C# proves too daunting, developers should stay with their chosen language. C++ and Visual Basic are still well-supported in .Net, but they have undergone significant changes.

Web developers working with JScript and Active Server Pages (ASP) also need to read about ways in which the new compiled JScript and ASP+ change the way they work. Microsoft will continue to support old facilities -- ASP+ runs in parallel with ASP, not in place of it -- but you should prepare for a mandated switch to .Net facilities when these products ship in their retail iterations.

IT managers have the luxury of waiting a while longer. If you haven't started your Windows 2000 migration, sit down with your key suppliers and partners to work up a plan. If you're looking at Windows for the first time, but you're not interested in the cluster/farm approach to scalability, stay your current course and keep watching.

Starting in 2002, the combination of 64-bit CPUs, the 64-bit Windows OS, and new internal and external bus architectures should begin to close the scalability gap with Unix servers. You'll hear some plaintive calls from hardware vendors, claiming that 64 bits means you need half as many CPUs. Stand your ground. Once 16-and 32-CPU PC servers ship from major vendors, the massive multiprocessor systems many companies need won't be far behind. If you can't wait, stack lots of smaller (one-CPU to four-CPU) PC servers in a rack instead of buying freestanding systems. Thanks to AMD's welcome interference in the PC market, prices for small rack-mount servers have fallen to very affordable levels.

Timing is everything

Whether by accident or design, Microsoft picked a great time for its coming out party. Its new software technologies just barely precede the unveiling of 64-bit PCs. The hard-core developer community had nearly completed its migration from Windows to Linux; .Net and C# are bringing them back. That translates to more Windows enterprise applications, more skilled developers and administrators, and more picky geeks unearthing OS bugs. The launch of retail .Net will make a bigger splash than Windows 2000 did.

In IT, enterprise Windows is largely a waiting game. Facilities that like Windows server farms have no reason to do things differently; .Net won't change the way they work. Windows' real opportunity to turn the corner on enterprise computing will come with massively parallel PC servers. If those don't come soon, Microsoft and IT will have to wait until the pendulum swings away from centralized computing and toward the distributed model.

Whatever the future brings, it's good to see Microsoft return to its classic, brassy, driven style. The buzz Microsoft has created among developers should give it plenty to brag about as it moves toward enterprise Windows.

InfoWorld