| Long-Term Plan for Manageability Announced |
| Apr. 7, 2003 |
|
The Dynamic Systems Initiative (DSI) is a far-reaching plan to improve the manageability of Windows systems and applications. Through this initiative Microsoft hopes to provide the products, development tools, and industry leadership needed to make data centers more dynamic—that is, able to continually monitor applications and their underlying dependencies and automatically add or reallocate computing resources as needed. Coming on the heels of similar initiative announcements by IBM and Sun Microsystems, the DSI may help Microsoft strengthen its enterprise image and counteract the open-source Linux-based platform, which lacks built-in management capabilities. However, the risks are high, the details are sketchy, and the time frames are long. Most notably, the DSI will require a new generation of application software containing self-describing resource, installation, and management requirements. The DSI will also require new hardware and could get bogged down in competing and non-interoperable standards. New Leadership Rethinking Management With the release of Windows Server 2003, the advent of 64-bit Windows and SQL Server, and the availability of powerful and more sophisticated Intel-based servers from vendors such as Unisys and Fujitsu, Microsoft has finally reached parity with Unix systems in terms of raw performance and scalability, and will pass them if progress continues along current trends. However, to become the favored vendor for mission-critical enterprise systems, raw performance will not be enough. Microsoft must also be competitive in terms of security, reliability, and manageability. In 2002, the company stepped up efforts on security and reliability with its Trustworthy Computing initiative, but manageability still depends on a blend of overlapping and nonintegrated products and technologies, such as Systems Management Server, Application Center, Microsoft Operations Manager, Group Policy, and Software Update Services. The one element tying Microsoft's management products together is Windows Management Instrumentation (WMI), a standard interface to management data, but it alone does not suffice for end-to-end application management. That is, WMI cannot answer questions such as "Are all components of the CRM system, including all underlying components it depends on, installed and configured as desired and functioning properly?" To do this, a much more comprehensive approach is needed. DSI: Built-In Manageability At the 2003 Microsoft Management Summit in March, Microsoft indicated that it has rethought and reprioritized manageability and has begun formulating a long-range management strategy, known as the DSI, which stems from research work begun in the late 1990s. A precise definition of the DSI is still evolving and Microsoft has released only a few high-level details, but the company envisions it as a radical change in the way applications are developed, deployed, and operated over their life cycles. Most important: rather than support manageability through add-on, generic management products that incorporate little or no knowledge of specific applications or hardware, manageability will be built into applications, system software, and hardware from the outset. The DSI will accomplish this with a common management infrastructure that spans all elements of the system and a System Definition Model (SDM) that provides a Microsoft-standard method of building management knowledge into each hardware or software component during its design. The SDM is an XML document that defines all the information required to manage that component, which could include dependencies on OS components or applications, definitions of settings that can be configured through OS policy, definitions of variables and counters required to monitor the health of the component, and so on. Among other things, this will enable a new breed of "knowledge-aware" business applications that use the SDM information to interpret management and configuration policies as inputs and use them to tailor the information services they provide to users. Microsoft believes the DSI will benefit customers in the following ways: The "dynamic data center." The DSI allows OS services and hardware resources, such as storage, to become "virtual" from the perspective of applications. Instead of each application being bound to specific OS instances and hardware, these resources can be relegated to pools that can be dynamically allocated to various applications according to policies. This will make it easier to automatically deploy applications and dynamically add, grow, or move computing resources according to the information in each application’s SDM. For example, a database that will soon grow beyond its current storage volume could automatically request the volume to be expanded from a pool of reserve storage. End-to-end operations automation. The DSI will provide a comprehensive infrastructure for application-centric management and monitoring. This will provide the foundation for future Microsoft or third-party management applications that automate tasks such as log collection, configuration management, change management, and software distribution and patching. Better problem diagnosis and resource planning. By providing a means for each component and management function to be aware of the others, the DSI will provide a basis for systemwide event correlation and root-cause analysis without requiring extensive, manual modeling of dependencies. For example, if a Web server fails to connect to a back-end database server, the failure could be for dozens of reasons. A DSI-based management application could automatically comb through failure events generated by various involved components and determine that the root cause was a duplicate network address. Furthermore, once a model of all system components and their interdependencies is available, it will be possible to leverage this information using tools that can analyze performance bottlenecks and project the impact of adding, upgrading, or reallocating resources. Long Timetable The DSI is a long-term plan that may take until the end of the decade before its benefits are fully realized. It will require system support provided by the Blackcomb release of Windows Server (estimated to ship no earlier than late 2005), new hardware, and DSI-enabled management products, such as Microsoft’s Systems Center, which will ship after Blackcomb. Furthermore, it will take a new generation of DSI-instrumented applications and the tools needed to build them, the latter of which Microsoft plans to build into a version of Visual Studio to ship in 2004. However, some components and benefits of the DSI will be made available to customers earlier. Windows 2003 already has components that will be instrumental to the DSI, such as the Network Load Balancing Service, the Virtual Disk Service, and the Windows System Resource Manager (WSRM). WSRM is a Windows Server 2003 add-on that can allocate server resources—primarily memory and CPUs—to particular applications. This version allows system administrators to manage workloads when more than one application is hosted on the same server, and allows them to restrict applications to use only the number of CPUs for which they are licensed. Furthermore, Windows 2003 will get two additional capabilities in the second half of 2003: Automated Deployment Service (ADS). Intended for blade servers and Web server farms, ADS will provide a fast, highly automated way of building fully configured servers from the "bare metal," allowing administrators to quickly add capacity when needed. Virtual Server (VS). Based on technology recently acquired from Connectix, VS will allow multiple instances of Windows to run simultaneously on the same server. Although VS has some immediate uses, such as allowing the consolidation of older Windows NT–based applications onto fewer servers, it will undoubtedly be the starting point for future Windows OS virtualization technology. Although administrators statically configure these Windows 2003 technologies and none of them use the SDM, Microsoft claims they are precursors to future DSI technologies and will provide much of the foundation for the dynamic data center. Big Idea, Big Risks The DSI is a big idea that represents a huge jump from all previous Microsoft efforts to improve manageability. However, Microsoft is not alone in its thinking—recently other vendors have put forward initiatives with similar goals, most notably IBM’s "Autonomous Computing" and Sun’s "N1" initiatives. Vendor support will be critical, because a fully realized DSI-based system will require replacement of today’s hardware, OSs, and applications with SDM-instrumented successors. Although Microsoft states it has support from major vendors, such as Computer Associates, Dell, and Hewlett-Packard, as well as a few small players, much more will be needed. Microsoft will also need to encourage hardware vendors to define and build standardized DSI-enabled servers, storage, and networking building blocks and will need to evangelize the DSI to ISVs and corporate developers, who may be unwilling to invest resources in it until the infrastructure is in place. This could create a classic "chicken or egg" problem that stalls progress. To overcome these issues, Microsoft must coordinate the DSI with standards bodies and other industry efforts, and it must also find ways to roll out the DSI and related management offerings in stages so that customers gain significant benefits along the way. That task may not be easy since ownership of management is split among various groups even within Microsoft itself. The Windows Server team owns the DSI, the new Enterprise Management Division owns all add-on management products, and developer tools are in yet another group, but all must contribute crucial parts of the overall management solution. Although all of these groups eventually come together under Platforms Group Vice President Jim Allchin, it may be difficult for these disparate teams to coordinate activities and timetables. Furthermore, because Microsoft’s own server product groups are now individually responsible for making sure their products exploit Microsoft’s management platform, it remains to be seen how quickly they will build SDM-instrumented versions of their products once the DSI infrastructure is ready. However, it appears that support from the top is finally there: according to several members of the Enterprise Management Division, the security incidents of the past several years have persuaded Bill Gates to make Windows management a much higher priority. Continued engagement by Gates and Allchin undoubtedly will determine the success of the initiative. Resources Microsoft will more fully describe the architectural requirements for hardware that supports the DSI at the Windows Hardware Engineering Conference (WinHEC) 2003, to be held in New Orleans May 6–8. For more on VS, see "Connectix VM Technology Acquired" on page 14 of the Apr. 2003 Update and "Server Virtualization in the Enterprise: An Overview of the Technology and Its Business Implications" at www.microsoft.com/windowsserver2003/docs/Virtualization.doc. |