Q: How many servers can be attached to a storage device?
A common mistake is to size the number of attached servers based upon the usable capacity of a storage device or the number of available ports. However, many factors can influence how many servers can be attached to a storage device, including the following:
- Application characteristics and workload performance requirements;
- Performance capabilities of the target storage device;
- Speed and number of physical ports on the device;
- Number of LUNs or target volumes;
- How many initiators are supported;
- Number of concurrent I/Os supported;
- Interoperability criteria;
- How many switches a vendor supports between servers and storage;
- Storage device LUN volume mapping and masking capabilities; and
- License fees that are dependent upon the number of attached servers.
Exercise caution and understand what the application performance characteristics and requirements are, along with what the other existing or planned workloads will be for a target storage device.
Other factors to consider include how many servers you can afford to impact should the storage device be taken offline for maintenance or upgrades. Also consider the complexity of scheduling maintenance and storage downtime as more servers are attached to the storage. Some subsystems have features to support attachment and consolidation of multiple and diverse server types and operating systems. Features include volume and storage groups, storage partitions, and virtual host ports, among others.
Host-based dependencies to consider include path management and device driver capabilities and limitations. In terms of performance, consideration should be given to the types of applications and their I/O access profiles. Don’t forget to include other workloads such as backup, data migration, and replication.
Greg Schulz is a senior analyst with the Evaluator Group research and consulting firm, and the author of Resilient Storage Networks (Elsevier). He can be reached at email@example.com.