How to Put a Price on Storage: The Danger of Cost Per GB

IT Storage Standoff

It is 4:45 in the afternoon on a Friday as John and Travis sit across from each other.  A heavy oak table sits between them, but it may as well be granite wall.  After a series of uncomfortable discussions, they are no closer to agreement than when they began reviewing the storage proposals nearly a week ago.  Sunlight beams into the dim room, creating rays as it reflects off the suspended dust. 

Travis wipes the perspiration from his forehead and glares at John.  “I’m the Storage Manager, it is my responsibility to select the appropriate technology for my environment.  I’ve looked at all the options and I know Hi-T-Stor is the right choice.”

“We’ve been through this.” John says, falling back into his chair in frustration. “Hi-T-Stor is twice as expensive per Gigabyte than EconoStor.  As IT Procurement, I can’t justify spending that much more just for data storage!”

Travis closes his laptop. “You still don’t understand!  EconoStor doesn’t include dynamic tiering, deduplication or thin provisioning!” Travis continues to list off technical differences between the solutions. John responds by unenthusiastically looking at his watch.

“I don’t care about all the dedupification—or whatever it’s called.” John says as he stands up from the table.  “We just need capacity to store our data and EconoStor is the most affordable option.  It will be my recommendation when we meet with Jeff Monday morning.”

Travis nearly leaps from his chair and points a shaky finger at John.  “I’ll be damned if that gear ends up in my datacenter!” he snarls.  Red faced and fuming, he storms out of the conference room.

Why is Putting a Price on Storage so Difficult?

Storage Admins, Engineers and Directors frequently have a difficult time when communicating the value of storage to IT procurement teams.  Procurement teams have to take an ever-evolving variety of technologies and boil them down to simplified cost metrics that assist them in sourcing technology.  Cost/GB can be helpful in comparing competing storage solutions, but relying too strongly on this metric can open up your organization to some serious disadvantages and missed opportunities.

There are several aspects of storage that should be considered when comparing solutions. In this post, I’ll share the approach I use.  I hope it also helps you to improve your accuracy in storage solution valuation.

3-D Storage Valuation

Historically, the price of storage has been determined by what is paid for capacity in static tiers.  New innovations in storage pooling and sub-lun tiering have changed the way we manage storage.  Instead of static tiers, there are dynamic pools that share capacity and performance.  Engineers now spec out storage systems with aggregate capacity and IOPs in mind much like server virtualization farms are spec’d for aggregate memory and compute power.  As a consequence, storage cost metrics must also be modernized.

I have adapted to use a 3 dimensional approach that considers the cost of capacity, the cost of performance and the additional savings that efficiency capabilities can potentially provide for the solution.

1. Cost/GB

Cost/GB must be based off usable capacity—not RAW.  I consider usable capacity to be the remaining, host presentable capacity after taking into account RAID overheads and any array or virtualization overhead.  This doesn’t include potential logical capacity when estimating any dedup or thin provisioning savings.  Keep it simple. Vendors should provide the usable capacity for their own solution.  It should also include the cost of the array, all software and maintenance required in the solution.  Sometimes I breakout a Cost/GB metric just for the disk capacity and a separate, all inclusive or “loaded” Cost/GB that includes the array software cost, but the calculations can be tricky unless you understand the vendor’s pricing model very well—so it is generally best to go with the loaded costs.

Key Factor:  Ratio is primarily affected by the number and capacity of drives.

Best Used: This number should be weighted most heavily when considering solutions requiring large amounts of storage with light IOP requirements.  Archive or backup capacity are good examples.

2. Cost/IOP

This is aggregate IOP potential for the solution.  It must be limited to the weakest performing link in the solution.  Meaning if the drives can do 100K IOPs, but the controllers max out at 25K, the cost is calculated at 25k.  An important consideration is your I/O profile (reads vs. writes, random vs. sequential and cache hit rates).  There is a huge performance delta between varying workloads and if you don’t provide your profile to the vendor, they cannot appropriately size the solution.  Worse case, they may size the IOPs with a more favorable profile in mind leaving you unpleasantly surprised later. The cost is calculated similarly to Cost/GB with total potential IOPs the vendor claims (with I/O profile assumptions) against the total, loaded cost of the solution.

Key Factor: Ratio is primarily affected by storage system compute resources such as CPUs, Cache, and high performance SSD or FC disks.

Best Used: This number should be more heavily weighted when considering mixed workloads with moderate to high performance needs.  VMware or database workloads can be good examples.

3. Efficiency (Utilization) Modifiers

These are technologies that boost the baseline utilization of the usable capacity.  They can include benefits provided by storage pooling, thin provisioning, deduplication, thin clones or compression.  For example in a traditional storage environment, I estimate a utilization rate of about 40%—meaning that 60% of the storage is either segmented and cannot be provisioned or is already provisioned but unused. Customers can typically boost storage utilization from 40% to 60% just by taking advantage of thin provisioning.  Another example is deduplication, where storage is relatively much more expensive per GB.  However, when factoring in estimated utilization increases of 500% to 3000% through deduplication, the storage becomes more efficient and, therefore, more affordable.  Like IOP potential, efficiency increases depend on the type and structure of data—so make sure you share this information with your vendors!

Key Factor:  Modifiers are primarily impacted by potential savings through software virtualization capabilities.  You frequently have to take the vendor’s word for it, but ask for details and references to back-up their claims.  Be sure to weigh the cost of the software against the projected savings!

Best Used:  These adjustments can either be used to show potential storage savings going forward or you may allow vendors to bid a lower usable capacity than required while giving reference to the expected savings.  In most cases, I prefer to grow into potential savings as opposed to cutting what may be on the floor today.  There is less risk in that approach.  Be sure to contractually hold the vendor accountable to their claims either way!

You may only use one or two of the metrics above, or all 3 together to assist in your valuation.  It really depends on the requirements of the solution.

A critical mistake I often see are organizations requesting a solutions without providing all the required information and then blaming the vendor later on because of missed expectations.  Capacity is not the only requirement!  Be sure to give the vendor as much information about the performance requirements, I/O profile, applications and data types as possible.  Vendors may over-look potential savings you could have benefited from, because you didn’t provide enough information about your environment.

Other Thoughts

• The relationship between IT Procurement and storage teams is a two way street.  The technical teams must gather the capacity, performance and utilization metrics so a valid comparison can be made across all options. 
• There are many, many other considerations in regards to High Availability, Data Protection and Disaster Recovery requirements that also should be weighed.  The 3-D Storage Valuation approach is simplified way to help compare the cost of storage—it doesn’t compare technical feasibility of the solution.
• Any vendor worth doing business with will ask for the required performance and data profile information.  Be leery of any vendors who throw over configs based solely on capacity—this is a major red flag.  Good vendors will almost always do an analysis of the existing environment so their solution can be sized appropriately.

I hope this post at least encourages you to think about storage costs in a new way.  As always, your feedback is welcome!