1. Will the vendor design an optimal file structure to take advantage of predictive caching or other features of their system?
2. What protocols do the clustered systems run? Can they take advantage of newer, lower cost protocols such as iSCSI?
3. Does the product require agents, adapters, initiators, drivers or other software to run on the application servers? These can add complexity or possibly hurt performance.
4. Is the product designed to deliver good sequential performance for streaming video or random I/O for transaction processing?
5. How many nodes will the cluster support without sacrificing ease of management and performance?
6. If additional hardware nodes are added, does the vendor charge more for clustering software?
7. If a node fails, how does the system handle data migration or replication to ensure uninterrupted application and data availability?
8. How easy is it to add new resources to the cluster?
BIO: Robert L. Scheier is a technology writer based in Boylston, Mass.
Material used here orginally appeared in Storage magazine.
This was first published in April 2009