Storage Pain: The Consumer View (Part 1 of 2)
Food to soothe the savage beast…..and maybe solve some problems.
I’ve been traveling in recent weeks to several cities in the Southern U.S. as a guest of Computer Associates. From Charlotte and Nashville to Atlanta and Birmingham, I have met with about 100 IT decision-makers and practitioners who have shared with me their storage pain, and listened, sometimes through gritted teeth, to what I had to say about what is really required to solve some of their knottier problems.
It is worth noting that these events have seen no marketing or sales pitches by CA. PowerPoints are verboten. Instead, the CA organizers have arranged for seating at a good restaurant in each city, put out the invitation to IT folks, and kept their promise of a free meal in exchange for a frank discussion of consumer needs.
From where I’m sitting, it’s a win-win. CA’s salespeople get to learn what storage is really all about from the consumer’s point of view. They learn about backups that don’t complete on time, SANs that aren’t delivering the return promised in the vendor brochure, and staffing and resource limitations, unprecedented growth, and cost. If all of this registers, they will become better at their jobs, more responsive to their customers with products they can really use, and develop better strategies for deploying products to achieve objectives.
For their part, the attendees are also getting something of extraordinary value. They get to exchange tips and share troubles with each other. On many occasions, an attendee has discovered that another attendee, whose shop is just down the street (but until now may as well have been on Mars), is having exactly the same problem with exactly the same product. Each has been told that the problem they are confronting is “unique to their shop,” but now they find out differently. Some vendors can expect a few irritated e-mails this week or next, I suspect, as consumers compare notes.
Everyone (except yours truly) gets a free meal at a five-star restaurant that might otherwise cost a week’s pay. Instead of eating, however, my job is to visit each table, listen, and contribute to the discussion. Basically, I am supposed to keep the chit-chat flowing until the last coffee cup is cleared away and the last goodbye handshake is proffered.
My message is simple: you need to manage storage if you are going to derive any real value from it. Buy a management solution now (whether CA’s BrightStor or any of the other solid products in the market), then make support for your chosen management framework a key criterion in the evaluation and selection of all storage technology going forward. Just doing this will drive close to 40 percent of the labor cost out of storage, I argue. I summon a lot of facts and figures, case studies, and consulting experience to make my point. And, of course, there are some jokes to keep things in a pleasant, dinner conversation mode.
The experience has renewed my confidence about a lot of things. I made this trip last year to many of the same venues and noted spotty turnout at the events—only four or five people showed up in some cities. This time we are seeing nearly ten times that number. Why?
Food is still food. CA is still CA. I’m still just that nutty storage critic and consumer advocate with his hair on fire. So, what has changed?
Here's some of what I am hearing over and over again from the participants:
First, backups are a bigger pain than ever. They aren’t getting done in the time allotted and many people are afraid that even a minor interruption may become a major disaster because so much data is going unprotected. Interestingly, not many hold out much faith that throwing tiered storage at the problem is going to make any difference. They see all of the second-tier-of-high-capacity-SATA-disk-configured-as-a-tape-surrogate as less of a fix to the problem than as an effort by vendors to sell them more disk. With no more human resources to allocate to storage, they are wondering how they will manage more capacity together with its additional complexity.
That’s when I suggest that their backups would probably be a lot less problematic if they did some analysis of the backup job and then broke it up into more bite-sized chunks.
Smart guys in the tape arena have spent countless hours explaining to me that backups are essentially multiple threads of data from multiple disks that get braided together into a “superstream.” In a perfect world, all of these threads would be the same length and would form a perfect stream to the tape drive that executes in perfect synchronicity. Backup software gives you the impression that such a perfect stream exists when it provides a simplistic measurement of how long a job will take to complete: “X number of GB will take X number of hours to back up.”
But it doesn’t work that way. Since threads are of different lengths, they tend to complete (that is, their data gets written to tape) at different times during the course of the overall backup job. As these threads complete and drop out of the superstream, the “braid” begins to unravel. You aren’t running those tape drives at their rated speed anymore, and they start “back hitching” or “shoe shining”—colorful terms for a knotty problem. When the drive doesn’t receive data at its rated speed, it stops, waits for data to accumulate, rewinds a little, fast forwards a little, and writes the data. Now, the data isn’t streaming to your drive anymore—it is in a stop-and-go traffic jam. Things slow down considerably and backups don’t complete in the allotted time.
One way to attack the problem is by creating backups as a series of jobs (rather than as one big job) in which each forms a superstream comprised of more or less equal-length threads. Doing this requires a lot of analysis (this is when the attendee’s teeth first begin to grind) and you need to do the analysis almost every time you do a backup (more grinding) if your data is growing fast and frequently.
One fellow asked aloud, “Who has the time for that?” Everyone at his table grumbled their agreement. That’s when I suggested that they 1) get their vendor to help them set it up and teach them how to do it, and 2) pressure their backup software vendor to provide automated tools to do the analysis and make recommendations or create scripts on the fly.
They seemed to like the idea of beating up on their vendors—a good sign.
That gave me the “in” for making my second recommendation. How about culling all the stale data and other garbage from your backup set before you perform the backup?
I rattled off the statistics about stale data. “Up to 80 percent of databases contain non-changing information that doesn’t need to be included in every backup.” “E-mail is often being replicated over and over with all attachments intact, which would be unnecessary if you separated attachments into an archive and migrated e-mail headers themselves into the archive after a set period of time.” “Access to end-user files tends to drop to zero after 90 days; so, why back them up over and over?”
Maybe your backups wouldn’t be so painful, I posit delicately, if you just managed your data better. This leads into a discussion of tools for archiving different data types (e.g., database, e-mail, user files, etc.) and ultimately for Hierarchical Storage Management (HSM), a precursor to true ILM.
It is also the point of the discussion when the techs tend to start giving me incredulous looks: “We barely have the resources we need to cope with the problems we have today. Now you are suggesting that we undertake a big data classification effort, deploy, and learn to use new software tools, and the like.” It is also the point in the conversation when IT managers and CIOs start to lean forward in rapt attention, saying, “We need to start doing that data management thing for regulatory compliance anyway. Tell us more.”
Suddenly, the conversation has shifted to an entirely different set of issues—issues concerning data management rather than storage management. Storage management is about managing capacity and performance—back office stuff. Data management is about managing utilization and cost-efficiency, regulatory compliance, service level conformance—front office stuff.
The good news is that only one or two people suggested that it would be easier to just throw more storage at the problem of data management. Both were shouted down by everyone else at the table where they were eating. Pleasantly, of course, but firmly.
You had to love it. The protest provided a great springboard for a kind of Socratic question-and-answer session that guided everyone to greater clarity with respect to data management and how to get to a disciplined data environment.
We’ll look at that discussion in the next column. In the meantime, the success of these dinner roundtables appears to have astonished the folks in CA’s Islandia, New York headquarters. Organizers such as Mark Newberry and Chuck McConnell are walking around with big grins.
My understanding is that the events are perceived as so successful that CA will continue to host them in cities across America (and possibly abroad) over the next several months. If I’m not speaking out of turn, I would recommend that consumers everywhere come out to these events. Registration details are at http://www3.ca.com/events/eventdetail.aspx?CevID=72282.
Maybe a few of CA’s competitors ought to sneak in as well, to see what real consumer-facing marketing is really all about, or hold their own dinner get-togethers. In the meantime, I’m off to Miami, then Puerto Rico for the next two dinners. You can still reach me by e-mail at firstname.lastname@example.org.