It's a scenario common to organizations of all sizes--a key application is going to be upgraded, and the platform...
upon which it is to run must be selected. With hardware costs for Windows, Solaris and Linux running neck and neck these days, there are major points beyond initial acquisition costs to consider. Here's a rundown of the top 10 things beyond initial costs that should enter into your decision-making process:
- Establish a partitioning strategy based on utilization. Co-locate multiple system instances (monitoring and reporting, for example) on a single system with adequate spare capacity. Isolate system instances with high utilization requirements on servers with adequate capacity for future scaling (adding additional CPUs or memory).
- Skill level of IT staff. Do you plan to run clusters or multiple servers in a load-balanced configuration to increase the availability of the environment? Practical experience shows that organizations with lower administration and engineering skill sets should stick with running multiple servers at lower utilization rates. Lack of skilled staff combined with running clustered servers in any OS can actually reduce availability, rather than increase it.
- Security concerns. Every OS is susceptible to security breaches if not patched properly and in a timely manner. The increased popularity of Linux in the last few years has led to ever-increasing attempts to exploit vulnerabilities in that OS. Check security sites like www.Secunia.com for a broad perspective on the risks inherent in each OS. Compare the risks and analyze the total costs associated with patching the operating systems you are considering.
- Staffing. When considering the TCO of a platform don't forget to include staffing costs. Studies show that the single biggest cost in IT is staffing. Hardware and software costs rarely account for more than 28% of TCO over 3 years. These percentages fall even further over a 4-5 year timeline. Linux and UNIX administrative support costs 10%-30% more than equivalent Windows staff, and fewer qualified Linux personnel in the job market increase direct hire support costs.
- Remember that open source isn't free. Every major Linux vendor charges for technical support, warranties and licensing indemnification. These costs must be factored into the TCO and ROI.
- Hardware support and repair options. Consider all available options before settling on one. Buying and deploying redundant hardware that the staff isn't trained on can result in a longer downtime scenario because it takes the staff longer to diagnose and fix the problem. It may be more cost effective to purchase a premium 4-hour on-site support contract with a vendor to repair/replace/troubleshoot an issue.
- Power requirements. Windows- and Linux-based servers have approximately half the power consumption of Sun servers. In a data center with UPS and generators nearing capacity, this can make a difference in the timing of an expensive facility upgrade.
- Read the roadmap. Get in touch with the product marketing team for the hardware and operating systems you are considering. Find out the 3-5 year plans for specific features and components of interest to your organization. Try to match up corporate rollout plans with the products that will have the longest support window.
- Synch up with development teams. If you're making a platform selection that will become an enterprise standard, ensure all infrastructure in development and test environments is included in efforts to roll out the corporate platform, at least in terms of OS and complexity. Include these costs in the TCO analysis.
- Scalability and performance. Whether you scale up or out, start by picking the right hardware for the job. The basic requirements to gather and analyze include:
- I/O and CPU intensity of the application
- Maximum CPU scalability of the application
- Acceptable rate of utilization
- Required level of availability
- Application quality (homegrown applications are typically not as stable as commercial apps)
- Peak vs. off-peak server demand
- Number of front-end and back-end application connections
- Number of infrastructure applications (monitoring, backups, etc.)
About the author Kackie Cohen is a Silicon Valley-based consultant providing data center planning and operations management to government and private sector clients. Kackie is the author of Windows 2000 Routing and Remote Access Service and co-author of Windows XP Networking.