[Image] [] [Image] By Dennis O'Flaherty Coming (Pretty) Soon - NT 5.0 Don't Call it Cairo! Some Microsoft beta codenames are around for what seems like no time [Image] at all before they undergo the magical transformation into a real-life product -- Denali, for instance, turned into Active Server Pages with admirable briskness. Cairo, on the other hand, has been around for what seems like forever. Just as an experiment, we riffled through some year-old trade magazines and came up with an article in the February, 1996 PCWorld discussing the beta release of the Windows-95-type GUI for NT 3.51, where Microsoft's NT Workstation Product Manager Megan Bliss spoke of Cairo as "the next major version of NT, slated for release in late 1996." The release she was talking about was version 4.0, but when it appeared Microsoft lost no time in informing the public that 4.0 wasn't Cairo. O.K., so maybe 5.0? But by January, 1997, PCWorld (and anybody we talked to at Microsoft) was already telling us "Don't mistake NT 5.0 for Cairo!" So What is Cairo? What gives? We put the question to Enzo Schiano, Group Product Manager for Windows NT Server, who gave us the current Microsoft position on Cairo: it was never really planned as a specific release or group of releases, but instead was seen as a broad set of technologies for distributed computing. These were developing at a steady but unremarkable pace, until the big Internet transformation hit Microsoft in the Winter of 1995-96; that was the catalyst that struck the Cairo technology family like a lightning bolt, and individual products within the group started getting pushed out into the world as fast as they could be readied for release (e.g., the rapid transformation of "Network OLE" into DCOM). Consider The Road to Cairo Closed In any event, says Schiano, we can stop looking for "Cairo." We have already seen many of the pieces of the original concept released in NT 4.0 and more of them will appear in 5.0, but Microsoft would like us to think in terms of evolution rather than revolution with the next release of NT. When, for instance, we asked Schiano if we could expect major performance gains in 5.0, he said it would be better to expect startling gains in functionality with no loss in performance -- if 4.0 made a secure, stable Windows platform a reality for small and medium-sized networks, then NT 5.0 is going to extend these benefits to the really big guys, while "enriching" the offerings of 4.0 substantially. The goal is clearly to position Windows NT as the premier infrastructure for distributed computing, so it needs to be presented as a logical, even conservative choice for all the needs of the enterprise. It's Impressive, but Can it Scale? This idea has been vigorously attacked in the past and probably will be for some time to come on the basis of Windows NT's supposed inability to scale far enough to meet enterprise needs in the big leagues. Microsoft's Schiano is equally vigorous in refuting this claim -- in his view, the perception of NT's limited scalability is largely the creation of UNIX vendors who have felt threatened by the fact that NT creates a "totally new pricing model." NT is, after all, non-proprietary as far as hardware is concerned, while its scalability critics have included IBM (with AIX), HP (with UX) and Sun (with SunOS and Solaris), who clearly have a vested interest in selling not only hardware but support services. In fact, according to Schiano, the one area where we should expect major performance gains in NT 5.0 is in scalability, both through its clustering capabilities and through the fact that Microsoft is already working with OEM's like Dell and Compaq to get the hardware to scale better with the new OS. But Schiano was emphatic in saying that "the reality is that NT already scales very well up to 8 processors." At Schiano's suggestion, we checked out tests run by TPC, a non-profit corporation that was founded to define transaction processing and database benchmarks (www.tpc.org). According to the results posted on the TPC site, NT 4.0 running on a Compaq Proliant 5000 with four processors was holding its own against a Sun running with twelve and an AIX system with 8, with unquestionably more favorable price-per-transaction figures. Schiano also emphasized the importance of optimized software in any scalability discussion, pointing to the difference between Oracle's "routine" early port to NT in version 7.2 and its optimized port in version 7.3. Hard-Core 5.0 Scalability aside, the core technologies that will be incorporated into 5.0 are impressive enough to attract plenty of enterprise attention. With the proviso that none of this is cast in concrete yet, some elements of the eventual full release can already be identified, and the most important of these will include Active Directory, Distributed File System (Dfs), Active Server, and Microsoft Management Server. As you read about these elements, both now and in the future, keep in mind the one big leitmotif for NT 5.0: Internet/intranet. If It's Not in Here, You Don't Need It The fact is that 5.0 represents a kind of "Big Bang" in the Microsoft development cosmos -- by the time it's ready for shrinkwrap, all of the applications that have been coming out of different corners of the Redmond campus in the drive to enable distributed computing across the Internet will have been unified into one giant, TCP/IP-enabled organism. If you want an emblematic example, just look at the User Interface that's promised for NT 5.0 -- instead of the old menus-and-dialogue-boxes look which we're all so familiar with, you're going to be seeing a "page-and-link" metaphor -- i.e., the GUI is going to look like a Web page throughout, and when you see a list of anything on a page, the elements in it are going to work like hyperlinks. 5.0's Leading Players Once you can get wrapped around that, the rest of this stuff begins to fall into place. Active Directory -- based on CCITT X.500 and LDAP (Lightweight Directory Access Protocol) is a massive hierarchical directory service that is, in Enzo Schiano's words, "essentially the repository for all the data and the components that you need in a distributed environment." With the ability to host objects as well as standard data formats, and integrating private key Kerberos and X. 509 public key security systems, the Active Directory will let Internet/intranet users access it using a wide variety of open standards so that it can simultaneously serve as the secure, dynamic DNS server within an enterprise, the single LDAP repository for any application within an enterprise or the sole X.500 directory system. (Besides which, it can integrate with Windows NT 3.x and 4.x directory services, NetWare Directory Services and NetWare 3.x binderies). Distributed File System (Dfs) is the obvious complement to Active Directory, aiming at making it easier to find and manage data on the network by uniting files on different machines into a single name space. With Dfs, you can easily build a single, hierarchical view of multiple file servers and file server shares on your network, so that instead of seeing a physical network of dozens of file servers, each with a separate directory structure, you'll now be able to see a few logical directories that include all of the important file servers and file server shares. File maintenance will be a lot easier with Dfs, simplifying virus scanning across a network, backup tasks, and content indexing (this will be tightly integrated with Microsoft Index Server, formerly known as Tripoli). The Active Server element of NT 5.0 will incorporate a lot of things we're already familiar with, but they will be working together within this larger structure for new purposes. So, although DCOM, Active Server Pages and the Microsoft Transaction server (for instance) are already out there and on the job in many networks, they will be pulled together under NT 5.0 to maximize the advantage that their language-independence gives then for creating Internet and intranet component software. A prime example of Active Server employment within NT 5.0 will be the Microsoft Management Console (MMC, codenamed Slate), which will provide an ISV-extensible common console framework for management applications. MMC doesn't provide management functions itself, but it provides a common environment for "snap-ins" which will be written both by Microsoft and ISV's as ActiveX modules. And since MMS heavily leverages Internet technologies like Active Server Pages, it will help Administrators to manage their systems remotely using IIS and a Web browser. (In fact, IIS will be a central element of NT 5.0, since its role in producing Active Server Pages will be crucial to the functioning of other services in the OS). Rumored Players Many other elements -- old, new and improved -- are being discussed as projected elements of NT 5.0, including plug-and-play, power management, full DirectX support, hierarchical storage management (HSM, from Wang), volume management (from Veritas), a dynamic DNS, support for 64-bit virtual memory (VLM, which can handle memory in the 4GB to 32GB address range) and others that are definitely far from finalized. A couple of possibilities, though, look fairly strong: the K2 version of IIS and Steelhead, Microsoft's new routing software. And Keep an Eye out for K2 and Steelhead K2 builds on IIS 3.0 in a number of ways, including a management console that will take the lid off the Registry to allow access to features like the pooling of ODBC connections and their timeouts, the caching of ISAPI applications, and filter mapping, as well as allowing the administrator to use a Web browser to view the virtual name space. K2 also permits Internet-based X.500 certificates to access NT directories, opening them up to Internet and intranet users. Finally, K2 will integrate PPTP (Point-to-Point Tunneling Protocol) into NT 5.0, which leads us to Steelhead. Available now in beta, Steelhead for Windows NT provides a step up from NT's present LAN-only support to multiprotocol routing over wide-area links. According to a review in the Feb. 24 PCWeek (www.pcweek.com), Steelhead has thoroughly revamped NT's routing architecture, fully integrating it with RAS for demand-dial routing and PPTP, adding RIP version 2 and OSPF routing protocols and supporting LAN-to-LAN VPN's (Virtual Private Networks) by extending the functionality of PPTP beyond client/server to server/server. All this, apparently, with a GUI so easy to use that configuring is a snap -- in fact, about the only thing the PCWeek reviewer could find to complain about was the thought that concentrating all your routing on the server instead of using a standalone router made your operation that much more vulnerable. We asked Lloyd Spencer, Microsoft's Group Product Manager for NT Communications, about this point and he seemed a little bemused. He suggested that anyone who was worried should check out www.edgeserver.com to see what U.S. Robotics has done with the Edgeserver system that lets ISP's consolidate RAS, Intranet, Internet and Web site processing in one remote access server box that has NT server imbedded right in it. It has to run all the time and the MTBF is outstanding, so why should Steelhead worry anybody on the vulnerability front? Anyway, Spencer emphasizes that Microsoft isn't out to do in standalone routers with Steelhead, basically they look at it as a platform on which they want vendors to build products for routing and remote access. And Steelhead does have one big advantage over hardware routers: it offers extensibility through APIs, so if you want to go to a 3rd party and work on new protocols to extend the platform, no problem ... which makes it a great deal for developers. As Spencer says, it basically boils down to how much do you want to invest in a router by comparison with routing that comes with NT and is dead easy to use? Maybe that last thought gives us the clue as to whether Steelhead will be in the shrinkwrap when NT 5.0 finally comes out or available as an add-on. Frankly, we'd guess it'll be right in there with K2 and all the other powerful goodies that are clustered under the outer reaches of the 5.0 umbrella right now.