We're in the middle of an evaluation pitting IBM Shark vs. EMC Symmetrix vs. the Hitachi 9900 platform. Our expectation is to share the disk storage across OS390, AIX (and non-IBM unix), VMS, NT and Novell. Major mission critical enterprise-wide applications reside on UNIX, MVS and VMS platforms (all across a WAN). NT and Novell are used across the WAN as well.
When evaluating these three vendors, what should I consider and where are the "gotchas"? They all are making claims about state-of-the-art performance and customer service (naturally, we are doing our reference checks, but they only tell half the story).
All I can say here is do your due diligence and test your production applications on each vendors solution BEFORE you make a purchase decision. Each vendor should be able to supply you with evaluation boxes to do your testing. I know Hitachi always does. Things to consider here are storage density, host connectivity, application performance degradation when using mixed environments, availability, and of course, cost. Everyone can make claims but I would use the nuclear arms protocol of "Trust But Verify."
In all honesty though, I have been through a number of similar benchmarks at quite a few customer sites and all being equal in everything else, Hitachi usually outperforms the others hands down, especially when connecting MANY hosts into a single subsystem. This is due to Hitachi having the only crossbar switch based internal architecture. But, don't trust me, do the testing yourself.
Other things to consider:
- Ability to include NAS into the SAN for the remote WAN nodes using a single subsystem (does NAS and SAN at the same time).
- Point in time copy robustness for backup and instant restores (ability to instantly create new BCVs on the fly).
- Time Stamping and Sequence Ids for writes if doing remote copy for disaster tolerance. (This is important for database transactional integrity when using hardware based remote copy solutions.)
- Ability to do ASYNC remote copy for extended distances with the least cost and best performance (multi-hop not required).
- Requirements for special drivers for HBAs due to timing conditions in an array (openness of solution).
- Ability to quickly make configuration or LUN changes ONLINE without cost or scheduled downtime.
- Ability to review performance and Cache statistics on a LUN by LUN basis in a graphical or report based format.
- Support for many switch vendors to enable "best of class" decisions on Infrastructure.
- Ability to scale the platform to full capacity and max out host connections with no impact to performance.
- Ability to configure subsystem to maximum connectivity without having to remove other components.
This was first published in June 2001