Hu Yoshida

Avoid Buying More Storage Than You Need Today

Blog Post created by Hu Yoshida Employee on Nov 7, 2014

Mar 3, 2014


Many data centers purchase storage infrastructure on a 3 to 5 year depreciation cycle. The problem with buying storage this way is that you have to buy more than you need for the first few years. These data centers have to over provision their storage requirements with capacity that is declining in price by about 24% per year. In 3 to 5 years when they are ready to use this capacity on the storage system technology that they purchased back then, new technology is already available to make storage systems even more efficient than before.  The decisions that they made 3 -5 years ago are still carried on the books as depreciation.


Some of the reasons for doing this are that it is too disruptive to add incremental capacity from a procurement and operational perspective. Another concern is that the capacity that you add 3 to 4 years into the cycle will need to be replaced with new technology before it can be depreciated.

Prior to the end of the depreciation cycle, data centers must plan for the migration to the next technology cycle. As storage systems became larger and larger, with an increasing number of applications and servers, the migration time where you have both the old and new storage systems sitting on the floor gets longer and longer. This is another form of over buying that is required for migration or technology refresh.

Data centers are well aware of the cost of over buying and are seeking ways to eliminate it. The ideal way to eliminate over purchasing is to go to an on demand, pay as you go model for provisioning.Here are some of the ways that data centers are addressing over purchasing of storage:

  • Buying limited storage capacity in each frame. While this gives up the economy of scale and will increase management overhead, it makes the depreciation and migration tasks more manageable by breaking them up into smaller increments.
  • Cloud provisioning. Contracting with a 3rd party to provide infrastructure over the network as required. The cost is carried as an expense. The terms and conditions vary by cloud provider.
  • Leasing where you upgrade and increase resources when demand increases and you don’t have to worry about removal of the hardware. Operating lease charges are treated as operational costs. The leasing company caries the inventory risk and factors that into his lease charges.
  • Managed services where a third party takes over the IT operations on site or offsite and charges the customer for usage. These charges are reported as operational expense.
  • Capacity on demand is a common model where additional capacity is pre-provisioned by the vendor or the reseller and turned on and charged when it is actually used.
  • Pay per use is a model, which is used with servers where a fee is charged based on how much compute power is used on a periodic basis. This does not apply well to storage, which is stateful, meaning once you put data on storage it remains on storage.  Pay per use may apply to tiered storage where fees are determined by the length of time that data occupies a certain tier of storage.

Whether or not these different approaches for avoiding over purchasing will reduce costs will depend on the utilization of technologies like virtualization and the terms and conditions of these different operational models.

Virtualization can help to grow a storage system, on demand, non-disruptively, and add new functionality like dynamic tiering without ripping and replacing the whole infrastructure. Virtualization essentially separates the storage functions from the storage capacity, so that you can add new capacity and continue to depreciate it while new storage functions are upgraded in the virtualization layer. Non-disruptive migrations – meaning no downtime for the application servers connected to these devices – are now available with virtualization. The enterprise storage controllers from HDS, starting from USP, which was introduced in 2004, can be non disruptively migrated to the current VSP through the virtualization of mount points greatly reducing the overprovisioning time required for tech refresh.

VSP can scale to 2048 x2.5 inch HDDs, with a system bandwidth of 192 GB/s and over 1 million IOPS, with the highest reliability and availability. With the ability to non-disruptively migrate to the next generation VSP there is no need to sacrifice economy of scale. You can consolidate multiple storage frames into one pool of managed resources and extend the life of storage assets with the assurance of non-disruptive migration to the next generation of Hitachi Technology.

When evaluating different operational models, you need to carefully evaluate the terms and conditions in relation to your business requirements. When trusting your data to a third party you need to do your due diligence. How viable is the third party and is he exposed to his suppliers? When Nirvanix went bankrupt they affected several other big name cloud providers. And, of course, security and compliance is also a concern when sending corporate data to a third party provider.

What is involved in the charges of a third party provider? There is usually a capacity charge based on type of service. Some charge on data transferred and the number of times you use PUT, COPY, POST, LIST, and GET. It may be cheap to store data but expensive to do any thing else with it. Length of contract will be a factor as well as the statement of work or service agreement. If your business changes and you require more services than agreed to up front, the additional charges may be outside your budgeted expenses. The end of the lease or service agreement may also pose problems if unplanned extensions are required. Keep in mind that the providers of these OPEX models are carrying the risk of the depreciation and they need to make money to remain in business. If their business model is not sound your data can end up locked in foreclosure.

Over buying capacity can be addressed in many ways. Storage virtualization is a key enabling technology to add resources when you need it, extend the life of existing assets, and non disruptively migrate to the next generation whether you do it in house or use a third party. There are many financial solutions and provisioning options available. A careful economic assessment should be done to decide the best options for your business.