Butler Group believes that it is time for IT management to take stock of their increasingly heterogeneous IT environments, and to look at the true costs that are in part driven by adherence to their incumbent Server Operating Systems. There is significant convergence of hardware technologies, and serious opportunities to reduce the TCO through consolidation of servers, and adoption of Open Source Software.
Very few IT managers ever go out to purchase an operating system. Whilst they might set internal standards for compliance and interoperability, at the end of the day, selection of an application for the business purpose is the main driver, with the operating system, and sometimes the hardware, being subservient to that choice. There will be less of a requirement for this compromise on standards, and potentially greater control, and management of the corporate IT environment, as more of the major Independent Software Vendors (ISVs) are certifying their applications for the Open Source Linux operating system, which will run on a variety of hardware, from low-end Intel server to proprietary mainframes.
In this context we expect to see the decline in dominance at the high-end by proprietary processor-based UNIX systems from HP, IBM, and Sun, with the migration to Linux and .NET. There will also be significant consolidation in the data centre, with multiple servers being aggregated onto single servers to simplify management and improve security.
In the selection of a Server Operating System the availability of specific features and functionality are important, but other aspects are of equal, if not more, relevance such as a clear vendor road map, use of non-proprietary processors, and adherence to standards.
Business Issues
TCO is an important selection criterion, with many of the costs hidden, including not just the purchase or acquisition costs, but also the resultant operational costs for deployment, maintenance, and support. Furthermore, there are a number of other considerations that effect overall running costs, such as system robustness, frequency of releases, need to upgrade hardware, vendor lock-in, and management time.
A number of benefits can accrue from server consolidation, such as a reduction in TCO, a more secure and efficient infrastructure, and easier maintenance and deployment of applications. When contemplating consolidation it is important to take into account more than just the hardware implications. There are numerous other considerations, such as Quality of Service (QoS) requirements, software needs, administration overheads, and even human factors. There is no right or wrong answer, the aim must be to find the correct balance between a centralised and fully distributed environment to suit the unique circumstances of each organisation.
The deployment of thin client architecture is a persuasive alternative to the usual `fat' client/server methodology with the benefits that can be achieved, most notably reduced administration and running costs, offsetting the relatively few issues. Nevertheless, thin client computing should not be seen as the total solution and must be used where appropriate. There will be situations where local PC processing is still the better approach. Again, the aim must be to reach a cost-effective balanced environment between the two extremes by using all the available technologies and system features to meet the requirements of both users and applications.
Technical Issues
Architectures
The three main architectures found in Server Operating Systems are UNIX (including Linux), Windows Server, and one on its own, IBM OS/400. The design criteria of UNIX systems make them inherently stable and secure, being originally developed by programmers for application development teams. The use of peer review, often employed during the construction of UNIX and Linux products, has produced more elegant and dependable software.
Windows Server's fundamental design concepts include the use of objects, modularity, and the use of a Hardware-Abstraction Layer (HAL) making the operating system, if Microsoft chose, portable across any number of hardware platforms. The central themes for the evolution of the Windows Server include interoperability, reliability, extensibility, and portability. Interestingly, POSIX-based interfaces are available within Windows 2000 Server, opening up the possibility of the Windows kernel supporting UNIX utilities and applications, if and when Microsoft decides to embrace this feature seriously.
IBM's proprietary OS/400 also adopts an object-oriented approach giving in-built security and integrity. A technology-independent interface with the hardware provides OS/400 with hardware portability. The tight integration between the hardware, currently the iSeries, and software allows the entire package to be optimised for enterprise processing. A total business solution is included within the operating environment from database to server applications.
Server Operating System Functionality
The range of technical and operational features form only a part of selection criteria for a Server Operating System. Reliability, flexibility, security, and adherence to standards are also key. The issue of support, both from vendors, ISVs, and Systems Integrators (SIs) will become more apparent, as the differences in functionality of respective operating systems decreases. Against this background, Butler Group believes that the clear winner will be Linux.
Market Analysis
Short-Term - 2002-2004
Butler Group expects that over the next three years we will see rapid penetration of Linux for file and print servers, either replacing NT operating systems on the same box, or as low cost replacements. Not having support from a vendor is a small issue, as file and print services are relatively simple operations, and seldom mission critical. Indeed, in Butler Group's opinion, there is no reason for even the smallest organisations not to start implementing and experimenting with Linux in order to gain the skills base and knowledge that will undoubtedly be needed in the next few years. The three proprietary processor-based UNIX offerings, HP-UX, AIX, and Solaris, will still be the systems of choice at the high-end.
Medium-Term - 2005-2008
Undoubtedly there will be consolidation in the Server Operating System market. This will be partly driven by the vendors themselves, as shown with the merger of Compaq and HP, which will see the planned eventual demise of the Tru64 UNIX operating system. Low-end UNIX will wither and be replaced by Linux in this period. Linux is also likely to have taken significant share from Windows. The high-end UNIX offerings will be facing significant challenge from .NET and a maturing Linux.
Long-Term - 2009 onwards
The winners in the long term will be Linux and .NET, with the Open Source offering penetrating the market from file and print server, through to the mainframe. The high-end proprietary UNIX offerings will have lost to both Linux and .NET, but predominantly the former as there is the opportunity for considerable skills transfer from the UNIX environment.
To cater for longer-term requirements, Butler Group believes it is imperative that the industry as a whole starts collaboration on a next generation Server Operating System. The current offerings are tired and no amount of papering over the cracks with new features hides the fact that the fundamental architecture designs are over twenty years old, dating back to a time when computing hardware and business needs were very different. In addition, the computing landscape is due to change dramatically over the next ten years with the arrival of Grid and Web services technology. The explosion of wireless intelligent devices and appliances using voice and streaming multimedia interfaces will, for example, make device driver software obsolete.
The present penchant for large, highly integrated operating environments is not the long-term answer. It just increases the complexity and reduces the operational flexibility of IT managers. It also locks the organisation into one particular vendors offering. The industry must consider repositioning to provide Server Operating Systems that are small, simple, and meet the two main objectives of virtual machine and resource management, with more distinct demarcation between the operating system, utilities, and application, allowing users to implement the rudimentary components and then choose which other options they wish to deploy.
Der vollständige Bericht kann bei der Butler Group bestellt werden.