Last week's article about a customer who thought his network-attached storage appliance was protected from outages because of the Redundant Array of Independent/Inexpensive Disk configuration, sparked quite a few responses from readers. I thoroughly enjoy receiving your e-mails as it keeps me on my toes and I would like to address some of those comments here.
One reader wrote:
"First of all I agree this person must have had some high budgetary constraints. If the box had been configured RAID 5 with a global spare and a parity disk this disaster could have been avoided. It doesn't matter that is was a NAS, even if it was direct attached configured this way, they were setting themselves up for a disaster."
Excellent points! However, the box did not support RAID 5, with the ability to stripe the data across multiple disks with parity. In a RAID 5 configuration, space is reserved on all drives for parity information. This parity information allows data to be reconstructed automatically if one of the drives fails. RAID 5 is a common configuration for protecting data but requires at least three disk drives and commonly uses five drives to implement, of course adding to the overall cost.
This system only had two drives.
I would also like to support the comment that any RAID array, whether it is attached directly to an application server or on a storage area network (SAN) with two disk drives, configured in a mirrored environment is at risk of experiencing the same problem of not failing-over. So, for buyers, be sure you understand and have documentation for the fail-over procedure on your RAID array.
I will note, though, that when using disk arrays in any consolidated environment, such as NAS or SAN, customers are at greater risk of wide-spread effects, in the event of a failure, due to the shared nature of the environment. This is why it is so important to understand and balance both the capital costs and the downtime costs when making storage infrastructure choices. For more information on RAID levels, there are many sites on-line that describe the different levels as well as the benefits and limitations. Go to http://www.google.com/ and search for "RAID definition."
Several of you reminded me of another age-old adage, "You get what you pay for." At least, you hope you do. More to the point, "pay for what you need!" One reader, who has obviously spent a good deal of time thinking about this, wrote:
"In my job, I find myself trying to push the themes of 'financial impact', 'operational impact' and 'business impact', in an attempt to ensure that my customers fully understand the opportunities and risks - both of doing something, and doing nothing - involved in storage consolidation.
"Once the opportunities and risks are qualified, and financial impact fully considered (both operational and capital), my customers are better able to make a business decision regarding storage. Most will also find that they can justify a great deal more money, once potential downtime costs, reduced maintenance, economies of scale, improved backup windows, etc. are taken into account - even in small shops.
There are reasons that cheap storage solutions are cheap. It is for the same reason that cheap cars are cheap. Usually safety and comfort are sacrificed. Although I have a modest salary, my wife and children do not travel in a cheap automobile - simply because I cannot begin to put a price on their lives, and nor could I imagine living without them. It's an unfair parallel, but for similar reasons, my company's valuable data does not reside on cheap storage."
Well, said! Both Mike Karp, my colleague, and I couldn't agree more and have stated as much in our columns. The return on your investment in time and energy doing the up-front business analysis of your storage requirements will be significant in maintaining a high quality of service to your applications and end-users.
Anne Skamarock is an analyst with Enterprise Management Associates (http://www.enterprisemanagement.com). She has worked with networked storage for the last 15 years and is currently focused on the storage practice within EMA. She can be reached at firstname.lastname@example.org.