Dave Merrill has an interesting new post on his HDS blog. He’s discussing reclamation options for recovering wastage and avoiding spending.
He divides storage up into four categories:
- Old or Worthless Data
- Duplicate Data
- Usable but unallocated to hosts
- Allocated by unused
Fortunately for Dave, the answers to these storage wastage issues nicely point to HDS products/services:
- Old or Worthless Data – Archive – Hitachi Content Archive Platform
- Duplicate Data – Dedupe – VTL with ProtecTIER
- Usable but unallocated to hosts – virtualisation – Universal Volume Manager
- Allocated but unused – Thin Provisioning – Hitachi Dynamic Provisioning
I guess there’s no surprise that Dave is highlighting this stuff and to be fair, his blog entry doesn’t actually point to or reference HDS products. From my perspective, this attempt at saving resources is commendable, but typically, I have my own Point of View on the subject.
First, the process by which this data is retrieved needs to be understood. That needs to be tied to a cost/benefit analysis based on current storage purchasing costs. So, if it takes 1 day of effort to retrieve 50GB of storage, it may not be worth the effort.
Second, review the provisioning process. Why did the wastage occur in the first place? Were processes defective and could they be corrected to ensure future allocations aren’t over-provisioned. In addition, migrations occur continuously, as hosts are migrated to new technology, then the migration process should right-size them for the target array, implicitly retrieving this over-allocation.
Third, have a joined up storage strategy. There’s no benefit implementing point solutions that don’t follow an overall storage strategy – so for instance implementing a specific archiving solution that can only be deployed on one platform or technology may cause problems in the future.
What’s my point? Well, I agree with Dave – but ensure any cost savings you implement form part of a joined-up process or you may end up saving less than you think.
4 Responses to Enterprise Computing: Storage Reclamation
Leave a Reply
You must be logged in to post a comment.
The first thought that comes to my mind when I read Chris’s blog is ‘accidental architectures’. Not long ago did I see a presentation on this very topic by one of the major analyst firms. The questions what Chris raises are valid. Why did these wastes occur in the first place? The answer to a large extent lies in the way things were architected and / or procured. The term ‘accidental’ does have a negative connotation but what it implies is different groups within an organization were growing organically and doing piece meal acquisitions. Blame it on the server proliferation fueled by Microsoft. Also, to be fair, some of these technologies that allow for aggregation and pooling did not exist when the storage sprawl began or the server sprawl for that matter. This is just hind sight judgment. To make things worse many of the file system went and claimed every square inch of disk space provisioned to them.
So what does it take to quantify the effort and to justify the investments for reclamation. Some of the capex investments are quite straight forward. That is the cost of entry for getting into the game. However as Chris rightly points out that the effort involved needs to be better understood. Storage utilization is at an all time low. However if you are talking of only a few tens of terabytes virtualization may not be for you. You could possibly do a better job managing it.
From a thin provisioning perspective the reclamation happens when you move the data from a thick to thin volume. My understanding is this can be done non disruptively. Also the zero page reclaim is also done non disruptively.
The amount of space reclaimed depends upon the application and the file system. Please note your mileage could vary. I am studying this more now a days and hope to provide a deeper insight with times to come.
Anyways, a well written perspective by Chris Evans.
-Vijay Ramaswamy’ HDS
You’re spot on. Technology is only ever a part of the solution to problems it often shared the blame for causing in the first place. The answer is always a combination of process and technology. With process, as you rightly point out, leading the way.
The problem for a lot of organisations is that sorting out the process can seem a lot more difficult that implementing another technology led solution. But the best results come when processes manage growth effectively without being an impediment to the business and which are able to quickly accommodate new technologies (think thin-provisioning) without needing to be radically re-engineered.
For example, there are great products out there to take the load off high-end storage for applications generating scads of data. Outside of email, filing, ERP and other ‘well-known’ applications though, archiving (or subsetting) is still ‘hard’. The solution is to ensure that the question of data lifecycle is addressed during application development as part of the governance process for new applications.
Governance processes which ask the right questions at the right point in application development and deployment can help to make information management part of the culture of an IT organisation (not just the storage team). Do that and IT can be way more successful in preventing waste in the first place.
Ultimately though, that wasted space is out there today in most if not all organisations. For many, while a long-term strategy of sorting out process and governance can bring real benefits, there is still a great deal to be gained through implementation of short-term technology ‘fixes’. The solution in the short term is, as you say, to weigh the benefits in terms of storage reclamation against the cost of new technology and the effort required to implement. Just remember that the alternative will usually be buying yet more storage and perpetuating the problem.
[...] This is not a new topic on my blog – there is an old blog entry on stop buying storage; and an entry on physical reduction, which is different from reclamation. Then a stand-alone entry on basic disk reclamation; and one from the Storage Architect. [...]
Thanks Vijay. There’s no doubt that resources can be saved – just takes some thought!