First, some articles on this topic, and the HDS offering from Wikibon.

Second, the added cost of thin provisioning has been defined as $1-per-GB.

Then go tell your CFO about thin provisioning.

Primarily, where is the economic sweet spot for thin provisioning.

First, thin provisioning has been called revolutionary, in that it can chance the behavior and usage patters, and buying patters for storage systems. This revolution comes comes by way of investments in:

• Processes defined and controlled for provisioning
• Capacity planning process
• CMDB
• Operational skills and perhaps some training
• Software tools, installation, and yearly maintenance

With an investment like this, there has to be a payback. My next blog will talk about the quantifiable benefits of thin provisioning, but I first want to help define the ranges for the best possible business case for thin provisioning.

With my research and reviewing this topic with several analysts, here are my findings for consideration:

1. Not all hosts and OS will be a candidate for thin provisioning
• Mainframe tends to be managed and tracked better
• Open systems are the more likely area to focus this technology
2. Some applications and file systems are better than others
• Unstructured data may not provide the best results
• Structured data should be the first area to plan for
• Some applications and databases pre-allocate storage blocks and volumes, so check with the TP vendor and the application vendors to ensure the right fit
3. Lower aggregate utilized systems are best
• If you are already running at 60-70% utilization (not allocation!) then thin provisioning may not have a strong case
• Thin provisioning can enable you to pick up 20 percentage points in the utilization category, so systems that are in the 20 percentile would be excellent candidates
4. The candidate storage needs to be pooled, (no DAS or SAS)
5. With my research, there appears to be a floor of around 40-50TB where the economics for thin provisioning and payback start to be strong.

So, if you are above 50TB, have chronically poor utilization of an open system SAN and have applications that play well with this technology, look to the next blog entry for some quantifiable calculations on determining the cost effectiveness and payback of thin provisioning.

In a time when IT budgets are under pressure, management is placing a closer eye on ROA. Problem is, with annual data growth rates at 30-100%, we cannot afford to go on a diet. Perhaps you have been part of, or overheard a conversation between an IT money man and an IT techie guy:

Tech: I need to buy some more storage for our growing business.

Finance: You have lots of spare capacity.

Tech: Oh common, disk is cheap.

Finance: But the CIO wants better ROA!

Tech: But the storage team is already maxed-out!

Finance: But we are wasting 50-60% of our capacity!

Tech: But it is too hard and time consuming to do a reclamation project !

Finance: But our capital expense cannot support more spending at the current levels of waste!

Tech: It is cheaper to buy disk and waste it then to reclaim or manage it.

The above dialog can go on and on, and it is too painful to write any more.

A recent development may be a source of help with these conflicting issues - Thin Provisioning. This technology has been around for a little while, but with new announcements this week this technology enters the high-end storage arena. These links can take to you announcement and other press sites.

HDS       Infoworld       ITWire

Thin provisioning can help reduce wasted space (white space) by 20%. That is a lot of thinning, providing better ROA!. Furthermore, this technology when mapped to storage virtualization is environmentally very attractive. According to ITCentrix:

“ITCentrix’s analysis of market leading high-end storage solutions shows that for I/O-intensive configurations of between 54 terabytes to 134 terabytes, competitive solutions consume 24- percent to 52-percent more budget for power, cooling and space over a five-year period than comparably configured Hitachi USP Vs using virtualization and Dynamic Provisioning,” said Dave Vellante, president and CEO of ITCentrix.”

 Another source for the economic impact of virtualization bundled with thin provisioning in an enterprise-class storage array, this article is from ESG.

Thinner and Greener, a powerful combination for TCO reductions in the storage space.  

There should be no surprise that the maturity progress also links to investment strategies and TCO strategies appropriate to the stage where you currently reside, or one or 2 steps ahead. Quite often we see IT organizations making a quantum leap 3 or 4 steps ahead with advanced concepts aimed at reducing cost, or providing a competitive edge, but what they often get is:

• Missed objectives
• project implementations that take forever to complete, if even completed at all
• Difficult organization and political barriers for radical change
• Processes that are not up to the task of some new technologies

A good example of this might be storage virtualization, thin provisioning, data de-duplication or advanced disk-level encryption. With some of these approaches a solid change control process or chargeback system would be absolutly essential for the implementation of a new technology. Sure the technology exists to deliver this capability, but if the infrastructure foundation of people and processes is not strong enough to support the new initiatives, the projects will fail. 

So as new solutions are evaluated, and technologies considered for advanced capability, take a little bit of time to ensure that you are at a level of storage maturity to handle the new function. Otherwise…… “you may not be able to handle the truth…..”

 Some take-away points that I am incorporating in my environmental econ pitch:

• We are nearing the point where the environmental costs for 3-4 years will be more than the purchase cost of storage infrastructure
• Planners need to consider 6kW and more per rack in the data center
• 2.5″ drives pull 5 volts each, and up to 12 volts for 3.5″ drives, but there is more power cost than the sum of the drives/volts (controllers, cache, ports)
• IT architects need to start using TB-per-square ft and TB-per-kW in plans
• Energy bills can be $60/sq. ft.

The current power and cooling growth may be unsustainable, but there are things to do now to flatten the cost curve:

1. Increase utilization of the array to increase the power/square ft. This can be done with a variety of activities
2. Implement tiered storage, since all data storage and access patterns are not the same
3. Rearrange the data center to get better cooling efficiencies. This may offset savings in floor space over time…..
4. Push more data off-line to tape, MAID or other data-at-rest technologies
5. Tech refresh may help with per-unit costs of cooling and power, since newer devices are more efficient

I also liked the comments about storage de-commissioning, and the costs to end-of-life the equipment and remove all data from disk systems. This may have to be added to the type of storage costs, it may be number 33……

Click here for other blog entries on environmental economics with storage.

After the workshop, in a conversation with a CSC colleague on this topic, some key points were discussed, because people often ask me about storage virtualization economics, if and how does this technology (like the USP or NSC) really save money. Here is what I told my colleague yesterday:

1. Although an important technology, virtualization is not a direct cost impact functionality. Rather, virtualization is an enabler of other cost reduction functions
2. Virtualization requires some advanced operational and architecture capability to realize the full benefits
3. Virtualization is not free
4. There is a cross-over point (as with most technologies) where the cost to virtualize provides long-term payback from the initial investment. With very small storage environments, the cross over point may never be realized.
5. The direct cost-lowing-functions that virtualization enables are (partial listing):
• Integrated tiered storage, with the cost benefit of the right data on the right value of storage infrastructure
• ILM and DLM with rapid data movement and remastering
• Single management point, with multiple storage types, technologies
• Better asset utilization, reducing long-term CAPEX
• Reduction of software licenses
• Reduction of HW and SW maintenance
• Although a better-trained SA is needed, the TB-per-FTE can be much higher
• With better aggregate utilization, the environmental costs per unit of storage is reduced

You can see from the list above that virtualization impacts several types of storage ownership cost, many of which can be reduced when virtualization is applied at the right place with the right investment and architecture.  See the link here for more on Virtualization from the HDS CTO Hu Yoshida.

Actually, I have blogged about the last 2 items, remastering and security in an earlier post.

First lets talk about data remastering, and why this is a cost factor for IT planners. For many, this term may invoke ideas about buying that Pink Floyd Dark Side of the Moon album for the 6th time, since it has now been digitally remastered (I bought it on 33 1/3 record, stereo cassette, 8-track, CD and then the remastered CD, and finally the iTune file). Movie makers and record producers can generate more revenue by offering and old media format and calling it ‘remastered’. The same is true with digital data, except that the remastering is needed due to media obsolescence or refresh. When you buy that new storage array, you have to move, migrate or remaster the data onto the new frame. Changes may occur with RAID level, LUN size, connection type etc. Just like the old vinyl, the music is the same recording, but we have to re-purchase to get it on the latest media format.

A customer once pointed out to me that the media migration cost and time was so long (2 months to get it on the new platform, and 2 months to de-stage at the end of the asset life) that the actual media and asset lifetime was 4-5 months shorter than the depreciation schedule. This can be factored into the ROA or depreciation/useful life cost of the asset. Dividing the useful life was changed from 36 months to 31 or 32 months.

There is also labor cost, tools cost and cost of loss or corruption every time data is moved or remastered. New laws and regulations requiring data to be around for 5, 10, 50 or 100 years means that during the lifetime of the data, remastering events may need to happen 3-20 times!

Next is the cost of securing or encrypting data, as part of the storage infrastructure. Some requirements require securing the data at-rest, other times it has to be protected or secured in-flight. The age, access patterns and movement/migration can also complicate the efforts, and increase the costs.

 Well the 32 areas have now been briefly summarized. No doubt that we will discover or define more types of storage ownership costs in the near future. Remember that price does not equal cost, and the cost of ownership, with its 32 options, can change how we plan for or architect storage solutions to be the lowest cost, even if it is not the lowest price.

First, the costs are just that, costs or the money involved with storage and data ownership. These costs are wide and various. I have been blogging about 32 types of storage ownership costs. Some of the money is hard, some is soft, that is up to you. Some of the money manifests itself in the short-term, some in the long term.

The second part is the events, investments, activities or actions needed to reduce costs. Just because you have identified the costs does not mean that they will automagically go down. We have to make decisions on what things to do to have the costs go down.

This goes back to my old analogy of weight loss. We can determine how much weight we want (or need) to lose, but knowing the target does not make it so. We have to then determine a plan of actions (diet, exercise, omitting something) in our life to make the weight loss happen. This analogy works well with storage TCO reduction.

So some of the costs can be:

• Electricity
• Floorspace
• HW maintenance
• Labor to manage the storage
• Long distance circuits
• SW purchase for licenses
• Recovery time, and the business impact
• Cost of downtime
• etc.

Then some of the activities might include:

• Storage Consolidation
• Thin provisioning
• Implementing storage area management tools
• Defining better management processes and procedures
• Tiered storage and a services catalog
• etc.

There are many, many activites and investments to reduce the cost. New vendor solutions and technologies are being developed/discovered all the time. The mapping of the kind of money to reduce to the activity necessary to achieve those cost reductions is not a difficult task, but needs to be done in the context of a storage strategy or annual plan. Just like losing weight, we need to:

1. Plan
2. Commit
3. Act/implement
4. Measure
5. go to step 2/3 again

Describing this storage TCO element is easy: Companies that do not have a disaster recovery or business continuity plan (documented, tested, verified) are running with a rated risk of business outage or disruption. This risk, whether real, predicted or ignored represents a real cost to the business. Determining the cost of the risk is an important factor in considering different storage architectures.

My good friend Dennis Wenk has written several papers on disaster risk and the cost of risk, speaks at events on this topic, and performs risk assessment workshop for customers. See a sample paper here.  Several years ago he shared with me some work on Annualized Loss Expectancy (ALE), and the concept that various events or catastrophes have a probability of occurrance, and cost of occurrance.

The cost of Risk = Threat x Vulnerability x Expected Loss

When summarized, every IT shop can calculate the ALE based on geography, architecture, provisions etc. This ALE becomes a cost number that should be factored into storage plans, architectures and TCO.

Dennis shared with me this graphic, that helps to illustrate the point of ALE.

When comparing 2 or more different architectures, looking beyond the purchase price is essential.  When looking to factor in the cost of resiliency, cheaper-to-buy-solutions may not have the same recovery rate or tolerance to more expensive systems. The actions or investment needed to reduce the cost of risk can be done by:

• Create a formal DR/BC plan and provisions that include frequent tests, reviews
• Define intermediate plans for on-site (local) recoverability or in-area recovery
• LD circuits, replication mechanisms, secondary hosts and storage

Cost item #29 is the risk (and cost of risk) due to management errors and poor capacity planning. Here is the rationale to these costs - Up to 25% of IT outages can be attributed to storage problems in-general. Of the storage-related outages, about 20-30% can be attributed to poor management, lack of processes and procedures in storage administration. Simply put, about 5-6% of all IT outages might be attributed to storage management, the people and processes around the storage infrastructure.

Taking a hard look at people and operational processes, the right level of capacity planning could reduce IT outage time and overall business risk. In an extreme case, one could calculate the hour-per-outage factor for the enterprise and multiply that number by 0.05 to help justify investments in storage area management tools, IT best practices and processes (ITIL is a good place to start) and capacity planning methods and managements.

SAM software, best practices and defined operating procedures can reduce these storage-related outages, and provide improved up-time (different than the data path or storage sub-system uptime). Optimizing the storage management team, reducing duplicate functions, establishing clear roles and responsibilities and MBO will help in total management effectiveness.