Today, a simple suggestion for delaying the purchase of expensive disk drives.
During the American Revolution the colonists were fond of saying "millions for defense, but not a dollar for tribute."
Today, many IT shops seem to have paraphrased this to "millions for hardware, but not a dollar for managing it."
The common thread running through my last several articles is that many shops continue to buy disks even though they really may not need them. They do this because, for whatever reason, management of storage resources has never been properly implemented at their sites. So, to be on the safe side, they add disks each quarter to prepare for the anticipated data explosion. The result is a feedback loop with more data on more disks, and none of it under any sort of management.
Most of us are aware that properly implemented storage resource management (SRM) software offers significant total cost of ownership and return on investment benefits over the long run.
So, I am suggesting that companies hold off on buying storage hardware for a quarter or two, and direct the savings to purchasing SRM tools.
At issue is how to avoid buying that quarterly shipment of disk drives and arrays. Our workarounds for saving money will only work for a quarter or two at most - don't plan on doing this stuff by hand over the long run. Hopefully when you save one quarter's worth of hardware expenditure you will then invest the savings in management tools that offer real long-term efficiency.
Last week, we talked about regaining space you already own.
Our two suggestions: defragmenting your disks (a small gain in space, but a big payback in system performance and stability), and flushing inappropriate files from your servers (a potentially huge increase in available space, but at the cost of losing lots of fun stuff on your systems).
Today's point acknowledges that optimizing disk usage is a lot more than just making sure disks are being used to their full capacity. We will try to dodge "infrastructure overspend" by suggesting some inexpensive ways to migrate files to cheaper storage environments as a way to avoid buying expensive hardware.
First, a reality check. If you have been purchasing disks for the last several years, you know technologies have changed. By now you have a mix-and-match collection in your shop that consists of high-speed devices (10K to 15K rpm spindle speeds), mid range devices (say, 7200 rpm), and lots of slower devices (5400 rpm or less) that you have been unable to kill.
Large capacity 15K hard disk drives (HDD) cost $35 per gigabyte or more (MUCH more if you want to install them in a RAID subsystem requiring vendor-approved devices). And an oft-quoted rule of thumb indicates that real annual costs (including support) may be six to 10 times that. Figure out what a best-case purchase costs you per-100G-byte of disk space.
How will moving files help? Here is the historical overview.
The usefulness of migrating files has been apparent for two decades now. Complicated and expensive Hierarchical Storage Management (HSM) systems that automatically moved unused files to tape or optical disk were available for mainframes in the early 1980s, and began showing up for open systems by the end of that decade. But managing such systems was typically unwieldy, and by the 1990s HSM fell on hard times.
Today the technology is making a comeback of sorts. A kind of HSM (but never identified as being such) appears as a part of some storage virtualization software, where data can be allocated to specific devices in the storage pool.
In the near future we expect to see an exciting new incarnation. Data will be assigned to disks based not just on how recently data has been used, but also on how important the data and programs are to the business. This will all be implemented by a sophisticated set of policies driven by rules engines such as those that can be seen today in other contexts from companies like Tivoli and BMC.
For you however, rather than spending the money up front, why not do some "pump priming?" With only a few steps you can move some less valuable data to less valuable online media, freeing up space on higher performance (and more expensive) disks. And every gigabyte you reclaim on a high-speed disk is an expensive gigabyte you don't have to buy.
At the simplest level, here is what to do.
Step 1: List your online storage devices, by performance level.
Identify the solid-state disks, the fast HDDs, and the progressively slower drives.
Step 2: Come up with a useful name for the project, to help win the hearts and minds of the people using your services. Make sure everyone understands the goal is to provide faster access to key data.
Step 3: Poll your stakeholders regarding the relative value of the data their department uses. Data accessed by batch processes (or by programs that haven't been used since 1995) are likely candidates for migration to slower devices.
Database tables are not.
Step 4: Sort files by their time-date stamp, to account for whatever is missed in the previous step. And remember that just because a file was hit 400 times in the last 24 hours doesn't mean it has any value to your company. It could be the football pool.
Step 5: In a methodical and well-documented way, move the least valuable/least used files to cheaper disks.
This sort of effort is not for the faint of heart, and it certainly isn't an approach you would want to use over a long period of time. Remember however that your objective is to save by doing this only once, and then to put your savings toward a strategic purchase of management software.
This won't be nearly as easy as what we discussed last week.
But it's a near certainty that saving one quarter's worth of disk expenditure and investing in SRM software will save you big bucks over the long haul.