There are only two goals when configuring server disks: risk avoidance and best performance. Regardless of usage or application, some best practices are universal.

Risk Avoidance

• Separate system and data. Discrete containers for the operating system/applications and data simplify maintenance, security, and disaster recovery.
• Moderate cost. With this strategy, you spend moderate cost for hardware while avoiding expensive reconfiguration or disaster recovery using other approaches.
• Hardware RAID preferred. Implementing redundant disks with hardware is significantly faster than software RAID with easier maintenance and fewer failures.
• RAID 1 system container. Mirrored system drives allow a server to run with no downtime upon the failure of one drive.
• Entire disk for system. You cannot extend a system partition, so it is recommended to format the full disk for the operating system.
• OS and applications only. The system partition should not contain data. Installing applications on special role servers like domain controllers and firewalls should be avoided.

 Best Performance

• Buy the best disks you can afford. SSD is the most expensive. SAS is middle of the road for cost, but much better performance than SATA.
• RAID 5 most common. Striping with parity requires a minimum of 3 drives and continues operating with the loss of one drive. This configuration generally provides more useable disk space than other approaches, except RAID 0.
• RAID 50 best speed/space/redundancy combination. Requiring at least 6 disks, this approach uses RAID 0 striping across 2 or more RAID 5 elements. The collection may lose a drive in each RAID 5 element and still function. For a large number of high-capacity drives, a Storage Area Network with RAID 50 is recommended.
• RAID 10 for high I/O. A stripe of RAID 1 mirrors requires at least 4 drives. RAID 10 can function with the loss of a drive loss in each RAID 1 element, but has approximately half the useable disk space of RAID 50. This configuration provides the highest throughput of any other configuration except RAID 0 for e-mail and database applications.
• RAID 6 not recommended. Briefly popular as the assumed next version above RAID 5, this approach comprises striping with double parity and fault tolerance of 2 drive loss. Unfortunately, RAID 6 has marginal comparative useable disk space and comes with a significant performance penalty.
• RAID 0 fastest and highest risk. With simple striping across 2 or more disks, you achieve the highest throughput and the most useable disk space. However, the loss of a single drives results in complete data loss.

This guidance is provided as part of proven managed services standard operating procedures since 1996. For example, solutions like virtualization consist of RAID 1 system container for HyperV and RAID 5 data container for virtual servers.

RAID – Redundant Array of Inexpensive Disks

Hybrid RAID – Nested RAID levels

Published by Kevin Fream

Many people don't understand their current situation and become frustrated in business and life, so I created the Delta Method and wrote the #1 Best-Seller "Change Your Game" - as featured at Harvard and NASDAQ as well as national TV news. Streamline your technology for competitive advantage using vetted IT support and take a divergent approach to personal growth. View all posts by Kevin Fream