Around the Storage Block
Showing results for 
Search instead for 
Did you mean: 

Lies, damned lies and statistics! Making cents of on-premises and cloud storage


SWjpg.jpgBy Simon Watkins, WW Product Marketing Manager, Backup, Recovery & Archive (BURA), HP Storage


“Lies, damned lies and statistics!” Mark Twain popularised this phrase in the United States to describe the persuasive power of numbers, particularly the use of statistics to bolster weak arguments.  


This quotation came to mind while reviewing a spat in the comments section of a recently published article on tape in the online publication The Register. The story quoted extensively from one of our blogs entitled: The opportunities for LTO tape in the era of Cloud and Big Data. But what appeared to generate most controversy among readers were the conflicting Total Cost of Ownership (TCO) benefit claims for tape and disk.  


So, in an attempt to put the record straight—and to perhaps to encourage more debate—I want to share with you the findings of a TCO model that uses published numbers to compare the economics of on-premises tape with off-site cloud for long-term archive.


Total Cost of Ownership (TCO) defined

Any credible TCO discussion starts with a specification of its components. TCO can be simply defined as the initial purchase costs of an asset that are capitalized (CAPEX) plus the on-going costs of ownership and operation (OPEX).  


CAPEX for on-premises storage solutions typically include one-time buy items such as hardware, software licenses, planning, installation and training. Operating expenses for on-premises storage comprise periodic costs incurred over time such as floor space operating cost, hardware maintenance contracts, software support contracts, power consumption (power, cooling) and personnel expense.


The initial capital expense (“short-term price”) of a storage system often gets an excessive level of focus from most buyers when choosing among alternatives in a purchasing decision. This is despite the fact that operating expenses can often contribute most to the TCO (“long-term price”) of many storage solutions.


TCO1J.jpgThis is amply demonstrated by a Clipper Group analyst report that compares the TCO of an on-premises disk array and tape library for long-term archiving. Using a nine-year TCO model, the study concluded that the average disk solution (including CAPEX and OPEX costs) costs more than 26x the average tape solution and requires 4x more floor space. The average disk solution was found to consume 105x more energy than the average tape solution, with the cost of energy alone for the average disk-based solution exceeding the entire TCO for the average tape-based solution!


Cloud storage avoids many on-premises CAPEX and OPEX costs (including hardware, software and storage management) but adds service costs such as data storage costs (monthly based on GB used), request costs (the number of transaction requests for data to be stored or retrieved from the cloud) and data transfer costs (into and out of the cloud).  


Comparing the economics of on-premises tape with off-site cloud for long-term archive

The objective of our TCO model is to compare the cost of storing 2PB of archived data on Google’s Nearline cloud-based archive service versus on-premises tape over a 2-year period. We assumed a 12% annual data growth rate with a monthly access requirement of 1.5%. Our model also posited that stored data had been optimized for archive (i.e. compressed and deduplicated) and that the use case was "deep archive" (i.e. large stores of infrequently accessed data, with low requirements on speed of access).


Our Tape TCO calculation included the published CAPEX costs (with industry-standard discounts
and price erosion) of 2 x HP StoreEver LTO-6 tape libraries with tape media as well as the OPEX costs of maintenance, depreciation, power, cooling, floor space and labour. The Cloud TCO model was based upon the recently published Google Nearline price structure and assumed data storage costs of $0.01 per GB/month with data out costs of $0.12 per GB and no charge for data in costs.


The resulting TCO for the Google cloud service solution was $640,000 USD over two years. UsingOnPremisesJ.jpg
published HP costs, a redundant tape library deployment could cost less than $450,000 USD over the same 24-month period. 


The delta in cash output scales as the size of the storage requirements grow. For example, a 3PB initial requirement with the same growth and access requirement could cost nearly $1M USD for the cloud-based solution verses $570,000 USD for an on-premises alternative.


Using tape AND cloud for business value

It’s clear that cloud-based services such as Google Nearline and Amazon Glacier will not cast a cloud over tape for big data archives. Depending upon several factors including the amount of data stored, the number and size of data retrievals, and the response time for retrieval, LTO based solutions will continue to offer a lower cost alternative to disk-based or cloud-based long-term data retention solutions. Additionally, we anticipate tape will continue to be highly utilized because many cloud providers do not offer the five-, ten- or 15-year retention capabilities that many organizations need.


Leveraging a cloud infrastructure makes good economic sense for any IT organization looking at TCO. But tape also makes economic sense for any public or private cloud providers of “at-scale” storage services who are looking at ROI. Intelligent tiered solutions with integrated tape storage have the potential to support service level agreements (SLAs) while significantly lowering costs.





Twain1.jpegI opened this blog with a quote from Mark Twain so let’s end it with another (slightly modified!) quote from the same literary icon: “Reports of tape’s death are greatly exaggerated!”


Learn more


Revisiting the Search for Long-Term Storage - A TCO Analysis of Tape and Disk

The opportunities for LTO tape in the era of Cloud and Big Data

0 Kudos
About the Author


Our team of Hewlett Packard Enterprise storage experts helps you to dive deep into relevant infrastructure topics.


It's "on-premises" not "on-premise." There is no relationship between "premise" and "premises"


Dan0 - totally get it and in fact I noticed this after it posted and asked Simon to fix it. 


That said, I think this is one of those pet peeves that some folks have that is made into a bigger deal than it needs to be. I've seen arguments about this on Twitter that you'd think a war was about to break out. 


For lots of people - me included, using "on-premise" makes sense because it's used as a noun. "My tape backup is on-premise";  It's only in one place so using a singular noun makes logical sense. But alas, the English language gods decided the the grammatically correct form had to be "on-premises". Even though I'm talking about a singular place, a plural noun has been deemed as the standard. For the next 100 years I would have continued to use "on-premise" if I hadn't been corrected (and within about 13 seconds after posting a blog) that my usage was incorrect. 


Personally, I think we should all be more focused on the meaning of what is being said and a bit less worried about premisses - say that in your best Smeagol voice! Smiley Happy