In this blog I look beyond the obvious and try to find out why people and companies do what they do - and what it means for the rest of us. 
Popular belief would have you think that the current economics of cloud based infrastructure services would either be linear or improve the more you shove out into the ether, but suprisingly (at least for now), that’s not true.
Pulling from my previous scarcity rants, and some anectdotal data collected at the CIO conference last week, I see where at some point – the economics work the wrong way.
Case in point: If it cost $.15/GB/month to rent cloud storage capacity from Amazon S3 and if the SLA and attributes (or lack thereof) work ok for you needs, no one really will question the math until the volume requirements get much bigger than GBs.
1 TB will cost $150/month. Still not that big a deal – but it starts to get interesting.
20 TB will cost $3,000/month – or $36,000/year. Now it seems we will hit a question/inflection point where someone is going to say, “wait a minute, can’t I buy 20 TB in an array for $5,000?” The answer, (regardless of W. Curtis Preston’s mathematical logic) is yes. You can.
“It’s cheap stuff” – who cares? So is the cloud. ”You can’t buy good RAID stuff for that money!” – really? What are they storing your data on in cloud? Symmetrix? I think not.
So the argument needs to be something much more attractive than $$ per GB or people are going to figure it out it’s really costing them a ton more than they can buy it themselves. Note, I’m not saying MANAGE it themselves or OPERATE it themselves – I’m saying BUY it themselves.
At a certain economic reality, no one will care. But it seems once you get up to 25-100 TB, it gets ugly in the current model. And I’m using AMAZON numbers.
What am I missing?
Related posts:
blogs
Certainly the opex is key. In addition, I think you’re undervaluing the tech on the cloud side with <>
Last I looked amazon was replicating for you, so you have automagic data integrity greater than the cheap raid.
Steve; Spot on amigo, and I run through this financial and operations analysis with all of my Service Provider partners.
Single storage services (either primary or backup) via Cloud is difficult to quantify in its own element. When additional services are layered on top then justification can be quantified more easily. Two examples:
1) Hosted exchange + primary storage = operations justification than managing 10000 mailboxes for your organization.
2) Remote/Offsite Replication + Archive (to meet compliance) + BMR = cost and operations justification than on-site backup.
SLA and robustness of service mitigates physical infrastructure your Cloud Service providers uses.
> can’t I buy 20TB in an array for $5,000?
> The answer is yes. You can.
The answer is no, you can’t. To match Amazon-provided 20TBs, you’d have additionally buy floor-space, energy, and staff. You need another math
, raw-capacity pricing doesn’t buy one production-ready storage.
—-See previous reply.
Steve
Excluding management and operations considerations makes this an unfair comparison. As I read this, you are comparing a service to hardware and by choosing to ignore the value added by the service conclude that the economics of the service are untenable compared to the hardware alone.
—Yes, I am. All I’m suggesting is that someone – at some point – will go back and look at the Cap-ex. If you are an already established IT organization – even if you suck – you have boxes, power, a data center, and some boss somewhere is thinking “don’t we already have people who know how to deal with this stuff?”. When the numbers get WAY out of whack they will draw attention and scrutiny – that’s all I suggest. Below that, it can sail under the radar, but once it looks on paper like I’m paying 10X more than I need too – things will get ugly.
Steve
OPEX vs. CAPEX ?
http://blogs.smugmug.com/don/2006/11/10/amazon-s3-show-me-the-money/
Don’t forget manage/operate/power/bandwidth etc all cost money as well…
But obviously at some level of scale – there is benefit in doing it yourself (otherwise amazon wouldn’t be making any money on it)
Thanks for highlighting this Steve. FYI – You may recall, Will Forrest over at McKinsey & Co. published a related report in the same vein last April, which stirred up quite a hornet’s nest of commentary from public Cloud proponents. I think the truth is somewhere in the middle today, with the balance of power tilting towards the public providers in roughly 5 years.
http://www.businessinsider.com/mckinsey-cloud-computing-overhyped-still-too-expensive-2009-4
I don’t think you are missing anything, just choosing not to look at the full picture. $$ per GB is a weak argument for moving to the cloud and anyone trying to justify doing so with that argument won’t make the move. The savings come in the areas you touched on but did not explore – management, power/cooling and the switch from a CapEx to OpEx. The TCO of cloud is illustrated well in “Above the Clouds: A Berkeley View of Cloud Computing” – found here http://www.eecs.berkeley.edu/Pubs/TechRpts/2009/EECS-2009-28.html.
Your missing the additional cost of the great big fat pipe some industries & individuals will need to effectively pump data to and from the clouds.
http://www.matrixstore.net/2009/09/21/big-fat-pipes-to-the-cloud/
With more “home shop” freelancers hitting big data in creative content creation the cloud is not yet the answer for them..
Steve,
that’s over simplified, what about power, cooling, bricks and mortar etc. How does it work out when this is factored in?
Dave O’.
[...] Duplessie posts on Cloud Economics and especially the economics of Cloud Storage, 20 TBs of storage from Amazon’s S3 cloud will [...]
I’m with Tomky – you already know what you’re “missing” and choosing not to include it.
Well, maybe there is one thing you overlooked: Cloud storage only charges for used space, not raw or usable. So using typical utilization values it’s 3x cheaper than buying stuff even at the same “cost per GB”. In other words, if you’re buying a 20 TB array and only putting 6 TB of data on it (as is the case often) then it’s 3x more expensive than you thought. You’ve gotta fill it to 100% to compare it to Amazon (or Nirvanix, where I work).
If you assume the following are true, then you’re absolutely right:
You know what you need ahead of time and so changing demands aren’t an issue
1) You only think about raw hardware cost (where’s maintenance? software?)
2) You want cheap featureless junk
3) Your labor is free
4) Environmentals (space, power) are free
5) You don’t want/need your data off-site
6) You don’t want/need your data to be sharable outside the firewall
7) You don’t mind buying it all again and migrating in 3 or 4 years when everything fails
Apart from that, you’re brilliant!
Steve, I agree with you that there are still issues that prevent large scale cloud deployments. Your cost points agrees with some back of the envelope math we did when looking at some projects, namely that at scale getting data in and out of the cloud from cloud on-ramps at scale really leads back to the question of buying more storage. However I think that there are other issues thwarting cloud adoption and I explore them in these three articles.
http://blogs.hds.com/michael/2009/09/chapter-3-cheap-cloud-storage-but-what-about-tco.html
http://blogs.hds.com/michael/2009/09/chapter-2-the-millennials.html
http://blogs.hds.com/michael/2009/09/what-is-the-challenge-to-web-applications-in-the-cloud.html
Even rolling in opex costs… for the level of service (which let’s admit is bottom of the barrel) I still don’t think it’s cheaper for any of the cloud storage providers currently for anything longer term.
Throwing some numbers together off the hip… let’s use a period of 4x years in Amazon and 4x years in a colo with purchased equip
Let’s give Amazon a grow-into advantage, and assume the array we buy is 20TB up front and can’t be grown so in year one we buy 5TB of storage, year two we have 10TB, year 3 we have 15 and year 4 the whole 20TB.
Amazon Cost
Year 1 = $9,000
Year 2 = $18,000
Year 3 = $27,000
Year 4 = $36,000
Total = $90,000
http://www.netstarcolo.com/prices.html was the first place I found with listed price per U on the web, I don’t know if its a crapy place or not(found cheaper/pricier ones)
14U of space is $300/month assuming I can get 20TB into that 14U = $3,600 year in colo fees (+one time $100 setup fee)
Let’s assume my acquisition storage costs is double so $10,000 for the 20TB just to pad up the numbers a bit
Year 1 = $10,000 + $3,600 + $100
Year 2 = $3,600
Year 3 = $3,600
Year 4 = $3,600
Total = $24,500
$90,000 – $24,500 = $65,500 / 4 years = $16,375
$16,375 is what I’ve got in admin, replacement, etc costs to have the same average cost as Amazon is each year (which for this type of storage I think is pretty easy to achieve: no performance tweeking, no backups, no space management, etc just let it sit there and someone reads/writes) it’s the equivalent of 109 hours billed at $150/hr per year (or 2.38 hours every work day for an entire year)
What I found interesting is that if I say my admin costs are cheap I could buy a second one for a redundant site ($49,000) for 4x years and still have $10,250/year in extra costs before the prices are the same and actually be redundant (or I could go crazy and have 4x of them totaling 80TB up and online for 4x years and cost only $2,000 / year more than Amazon assuming I have no admin costs)
That’s ruff and tumble thrown together numbers that may be wrong (please correct as it’s off the cuff)… but it sure seems like Amazon isn’t massively cheaper, or even marginally cheaper (or even cheaper) to take a business gamble on something that gives you no SLA guarantee and when it goes away there is no IT guy to go yell at to give the upper mgmt a warm fuzzy but simply a status page you click “refresh” on until it comes back sometime later.
Amazing how polarizing discussions around clouds, regardless of if storage, compute, application, product or services can be (see this post http://storageio.com/blog/?p=657 ).
Steve (Duplessie) brings up a good point about the carriers (e.g. telcos) and their place in delivering services. A variation is what some of the banks like Wells Fargo with vSafe are doing in terms of providing a service. Initially, the service is marketed like a virtual safe deposit box when in reality it is a backup managed service.
In reality if a bank like Wells Fargo or any other for that matter is serious, they need to provide a service that is more in tune with being archive not to mention those value adds that go along with it. That is, provide audit tracks/trails/reporting/alerts in addition to table stakes integrity/availability/security type functions.
If the banks can provide a real service, then they can cater to a specific market, those that trust banks with their digital assets instead of someone else. If the banks don’t take a service and value add or revenue generating/protection approach, then the current trend of backup as a service will fall into the category of traditional banking gifts as a thank you for opening a checking account.
Hence as Steve D points out, no surprise why backup as a service ( BaaS
or cloud backup is having success this time around compared to the last SSP go-round. It’s addressing a core issue, similar to some organizations such as universities letting some email accounts go to Google or, perhaps hosting faculty and staff on Microsoft based cloud services as a way of shifting some burden to free up resources for other tasks.
Steve Foskett brings up utilization as an enabler. Utilization could and should play well into the hosting of digital archives potentially even at a lower cost than internal hosting, maybe even lower than traditional off-line, off-site archives. Yet, that raises the concern or question of who are you going to entrust your digital archives to, a well funded startup, Google, Amazon, Microsoft, a bank, ATT, EMC, HP, IBM, Oracle, SunGard or Iron Mountain among others?
Even though clouds, public or private are built on binary technologies, the business and as Steve D cites, emotional and philosophical aspects are not so clear cut, black and white let alone binary. However one thing is for sure, discussions around IT clouds can get rather polarizing.
Cheers gs
[...] may have read a recent blog entry by a leading industry analyst addressing cloud storage economics: 20TB of public cloud storage as a [...]
[...] in the blogosphere when he re-visited the cloud economics issue in two very interesting posts (see http://www.thebiggertruth.com/2009/09/reverse-cloud-economics/ and http://www.thebiggertruth.com/2009/09/cloud-economics-continued/). He points out that while a [...]