I realize how the standard is defined for "overall high-availability" in a storage device. I wonder if there are accepted industry standards for performance degradation or unavailability in a high-end storage machine under declared failure. The following two examples illustrate my question:
1. A high-end storage system has yielded a certain performance benchmark under normal functioning. Assume there is now a failure in one of the disk devices and it has to be replaced. All redundancy mechanisms work properly and availability is thus unaffected. However, while the disk has not been replaced, overall performance is indeed affected. Is there a standard for the degradation performance allowed in such a situation?
2. A storage system has a dual write redundancy mechanism. At some point there is a cache memory failure and the dual write mechanism is properly activated and thus no data is loss. However, this mechanism requires that certain IO requests will be refused for a brief period of time, while recovery is being processed. Is there any standard for what is a "reasonably short period of time" in such a case?
I could add some other examples to this list but I think that these two should suffice at this point.
I find it very interesting that you say you understand "how the standard is defined for 'overall high-availability' in a storage device." I don't believe that there is a standard definition for overall HA in a storage device. I'd be most interested in your interpretation.I have found that every vendor, whether system, storage or software, each has its own definition of high availability and that its products coincidentally fit that definition precisely.
In your examples, you are looking for permitted degradation levels and outage periods that fit the model for high availability. To the best of my knowledge, no such standards exist either for the examples you describe or for storage.
I have been writing in this space for some time that high availability is not defined by the tools that increase system availability but rather is defined by the enterprise who employs those tools in an attempt to increase their availability. What would be considered highly available to an online vendor such as amazon.com or ebay.com is quite different from what would be considered highly available by a small accounting firm's billing department.
Bottom line: there are no standards that define high availability because everyone's needs are different. If your critical systems need higher levels of availability than mine, it's because your systems return greater value to the enterprise than mine do and so you are willing to spend the extra money that is required to reach those higher levels of availability.
Evan L. Marcus
Editor's note: Do you agree with this expert's response? If you have more to share, post it in one of our .bphAaR2qhqA^0@/searchstorage>discussion forums.
This was first published in February 2003