Update — Reader Justin Fischer points out my math below is off. Interestingly I wasn’t the only reporter to make this mistake. Sorry.
If you have an entrepreneurial bent it’s hard not to see an opportunity to start the next big cloud storage company in last week’s Nearline Storage announcement by Google. I saw it immediately. So did Google make a big pricing mistake? Probably not.
Nearline storage usually means files stored on tapes in automated libraries. You ask for the file and a robot arm loads the tape giving you access to your data in a couple minutes. Google’s version of nearline storage is way faster, promising file access in three seconds or less. They don’t say how it works but it makes sense to imagine the data is stored on disks that are powered-down to save energy. When you ask for the file they spin-up the disk and give it to you.
The big deal here is the price difference between Google’s normal cloud storage and Google Nearline Storage, which is 26 times cheaper per bit. Google Nearline costs 40 percent (not the four percent my ancient eyes originally saw) of the online storage price.
The chance here for storage price arbitrage screams out (or did, I guess, so kick me). Start your own cloud storage business for large files (video is perfect) based on Goggle Nearline. Keep a giant File Allocation Table (FAT) of nearline files in RAM on your web storage server. This is exactly how Novell revolutionized the LAN file server business in the 1980s with its Indexed Turbo FAT speeding file access by 1000X. Hold the most recently viewed files on a RAID array that also includes the first few megabytes (three seconds worth) of all other customer files. That way when the file request comes in the location’s already in RAM and the first few seconds can be delivered from RAID then the nearline data kicks in smoothly.
The result would be a cloud storage system for large files that’s as fast as any other — as fast as Google — for maybe 10 percent the cost per bit. I’d bet that’s exactly how Google manages YouTube files.
With such an obvious opportunity to build a business at their expense, why didn’t Google price Nearline higher? Maybe they have patents to cover some of this and think they can keep competitors like me at bay. But more likely they just want Nearline to succeed and making it darned cheap should do that. In fact this could in the long run be the death of tape storage.
However a very smart friend of mine sees the market differently. He thinks cloud storage price competition is going away and the price per bit at every cloud storage vendor will shortly be the same — zero. They’ll give sway the storage space and make money instead on the extras —security, encryption, redundancy, geographic distribution, and integration with many operating and file systems.
Having given it some thought I agree. I’d go even further to suggest that some consumer cloud storage companies don’t care about the money at all. Their game is to be acquired so what really matters is the size of their user base and the best price for getting lots of users is free.
These business lessons were learned 20-30 years ago in the storage locker business. I have some of my meager retirement money invested in limited partnerships at Public Storage Inc., the national storage chain. That business is financed with hundreds of such limited partnerships. The PSI value proposition for limited partners is interesting. They build in seedier (cheaper) parts of town. They are able to charge rents equivalent to apartments on a per-square-foot basis without requiring the amenities or management labor of apartments. Even better, they have your stuff, so losses for deadbeats are near zero. If you don’t pay the rent they just auction your stuff. All very interesting, but not the major point. PSI makes its real money by building or buying storage units in the path of growth. They think 20 years ahead to what that seedy neighborhood will eventually become. Sometimes they build structures that can be converted to specific other uses when they are sold, increasing their value in a growing market. The actual storage business only has to run at break-even with the big bucks coming when the property is redeveloped or sold.
Now consider how this model might apply to cloud storage. They have your stuff, making most customers long-term captives. The value here lies in the customer relationship. The cost can go to zero because the real money comes from ultimately selling the customer relationship to a larger company. This makes every cloud storage company either an acquisition candidate (most of them) or a company that intends to become enormous then realize that customer value through selling value-added services down the line. It’s grow or die.
So the actual cost of providing cloud storage isn’t going to zero. Disk drives and bandwidth will continue to cost money. That’s just the cost of being (temporarily) in business.
[…] Full blog here […]
From this link: https://cloud.google.com/storage/pricing
Regular storage is 2.6 cents per gigabyte per month.
Near line storage is 1 cent per gigabyte per month.
That’s 2.6 times cheaper — not 26 times.
Facebook does this already for all of those cat videos that people look at for a couple of weeks then forget about. They might even call it write and read once, if I recall correctly. They use some sort of fancy disk, but the catch is, only one disk on the rack can be running at a time because of vibrations.
You’ve got it wrong, cringe. It’s “flash server” based. Just like flash drives as portable storage. No spin-up, spin-down, nothing rotating. My web hosting company uses flash servers.
Bob is not talking about servers in general, but nearline storage (using disks or tape) as defined here: http://en.wikipedia.org/wiki/Nearline_storage. Of course, in principle, there is no reason why SSDs couldn’t be used for this purpose. The question that occurs to me is whether you are talking about servers or nearline storage.
The white paper is marketing it as an alternative to “traditional IT infrastructure: cold storage and disaster recovery.” It would be interesting to know how accessible is the data.
This is an interesting news story that deserves some attention.
.
The most important thing about backup’s is the RECOVERY process. If you were to lose a whole server, or all your servers how long could your business run without them? What is the most important information you need to run your business? How much data? How long will it take to restore it? While cloud backup services offer you a new level of service at a lower price, can you recover your data fast enough in a crisis? This is the first and most important question you must determine.
.
Please do not get too caught up on whether a backup service uses tape, disk, or SSD. For a firm to offer you an aggressively low price for backup services, they need the best cost of storage. Looking at Google’s numbers 1 cent per GB, that is roughly $10 per TB. Checking today’s prices I see I can buy a 1TB drive for $50. If I double that price to cover the cost of the server, network, etc that comes to $100 per TB. If my fee of 1 cent per GB is a monthly fee, it will take 10 months for me to break even. This is a hardware only evaluation, I haven’t paid anyone a salary on this yet. My point here is the economics of the service are vitally important. If you service cannot stay in business, what good are they? Google is clearly pushing the boundaries in price.
.
These days you absolutely, positively MUST encrypt your backup data! And encrypt it well. It needs to be encrypted while on your network and stay that way through the whole process. Don’t fall into the trap of thinking your backup service is a giant NAS in the sky that you can simply copy files to and from.
.
Other things a good backup service should have — multiple versions of files, data compression, data de-duplication, etc.
.
Most backup services need some sort of software installed on every server and workstation. Some of the cheap or free stuff may not do everything you will need, or may not work correctly. It may not handle databases and email post-offices well. You generally have to pay for the good software. Many backup programs cannot restore a whole server from “bare metal.” You have to install the OS and then the backup program, then start the restore process. Often you have to install all the applications too. So it is a good idea to keep track of all your license keys and keep copies of your software ISO’s too. Those will be the first files you’ll need in a full disaster, so it would be wise to make sure you have a backup copy of everything you need to build a server from “bare metal.” (Note — there are programs that can copy the whole image of a server which can be used to restore it later. These work best when you can restore them to the exact same model of equipment. If you restore an image onto a server that is 2 or 3 years newer, it may not work. This is one of the big advantages of virtualization.)
.
If I were to run (or support) a small to medium company, I’d create a hybrid backup system. I’d set up a NAS to keep a local backup of all files. I’d then use a cloud backup service as a second, off-site backup. If I lost a whole disk or a whole server I could restore it at LAN speeds from my NAS. If I lost my whole computer room, the cloud backup would save me. This combination gives me the recovery speeds I’d like to have for the most common type of problems.
.
IBM has a pretty good cloud based backup service. Google “IBM Smart Cloud Managed Backup.” There you will find information on most of the features a good backup service should have. You won’t find any price information on IBM’s service on the web. You’ll have to fill out a form, talk to someone at IBM, and go through a quote process. You’ll find IBM’s service is probably not optimized for very low cost.
.
Looking at IBM’s service and comparing it to Google’s can be an instructive exercise. You’ll quickly discover IBM does not yet fully understand the BUSINESS of operating in the cloud. IBM will have 10x more people involved in the process than Google. Google service will be optimized for the most economic storage of data. IBM’s service will be optimized to use as much of IBM’s products and services as possible.
Goggle Nearline? Goggle? I like it. However, Google’s attorneys will undoubtedly file suit.
My comments will invite those who’ll label me as certain brand of theorist. So be as it may.
This is all about acquisition of data. This is a no-brainer in my opinion. Offer something of perceived value for free or for next to nothing. In return, glean every last keystroke and data bit the user voluntarily submits. And, the public does this with alacrity.
Nobody in business (in or collusion) does anything for “free.”
Facebook, Google, Amazon, et al are simply data miners. Is there value provided to the end user? Sure. But, it comes at a price. Not all costs are monetary.
That’s what’s funny about this. When Google took over search it was to find small, specific things out about users that could be converted into actionable advertisement placements. The things they learned could be used to make money. Mail was another product that played a similar role, converting small specific facts to ad revenue. Youtube content was mined to place ads too, but it’s primary role was to draw eyes to see the ads.
But with this foray into huge free-ish storage I’m not clear on how they intend to monetize this content. The only way this works is with huge network capacity, but that same pipe makes it fairly easy to move your stuff to a different host whenever you want. The notion that I’m totally screwed because it’ll take a couple weeks to move my backups is ridiculous. There are very few situations that come up in real life where I can’t foresee a vendor change coming months in advance (and most of those pop up from business failures on the other end, something unlikely to happen to Google.)
I think this has more to do with Amazon’s AWS cheap Glacier storage option where you can store very cheaply as long as you don’t need the data often. (Restores cost more depending on how quickly you retrieve your data.) If Google can knock the bottom out of an AWS revenue stream they might be able to delay the universal adoption of that computing and business model and Amazon’s potential dominance there.
Yes the Nearline Storage price is 2.6 times cheaper than the regular Google storage price, but the network egress price is still $0.08/GB at the cheapest price point (ingress is free), which is 8 times the storage price itself.
Meaning for any cloud application where you actually plan to access the data frequently, you will be paying so much for the data egress that the cost savings in the actual storage will matter very little.
See https://cloud.google.com/storage/pricing#storage-pricing for more details.
Interesting comparison with warehouse storage to cloud storage. I guess the obvious question is can Google auction off your data?
If you fail to pay for your storage, and you don’t delete your data, then Google can auction it off. No doubt this will be the basis of the next reality TV version of “Storage Wars.”
Re: “and you don’t delete your data”. Why would that make a difference? Deleting it makes it invisible to you, not necessarily Google as well, unless there is something specified in their TOS where they promise to make it physically unrecoverable.
Cloud can be good for other things. But not for file sharing. Binfer does not store files anywhere. It is a better “cloudless” way to share data securely. About cloudless sharing.
the storage techniques between warehouse and cloud storage is very impressive.
Reading all the comments above, looks like no one actually understand storage tearing. Clearly most are Comments are coming from Tethys who usually lack business acumen. Nearline storage should be used for archival purposes not for backups. Unless it’s an annual backup. A typical business can archive much of its historical data eg. Older than 1-7 years depending upon their business and industry requirements in Google nearline and recent data in Google docs, share point on-prem or cloud or even filers. Of course Google will break even in 12 – 24 months if you don’t retrieve any data but if you do then you pay a premium, that’s a sustainable revenue stream without ads.