Peter Bookman, Global Strategist, Sphere 3D
At the turn of the century (I feel old just saying that) Microsoft commoditized hardware so it could sell software. Bill Gates was on the cover of magazines and everyone with half a brain agreed wholeheartedly that hardware was dead and software was king. Fast forward a decade and Apple is giving away software for free so they can sell hardware at a premium. Apple and hardware is now king of the hill. What’s going on here?
The same thing happened with centralized and decentralized compute. Mainframes to Windows desktops to data centers. We have gone from ‘obviously compute should be centralized’ to ‘compute obviously shouldn’t be.’ Orange is the new Black.
Things aren’t that simplistic. Things don’t go from black to white to black again. Usually they evolve and become more mixed or complex than the absolutes we started with. Let’s look at Cloud computing from this perspective. Some fanatics believe that the world is headed back to centralized compute with the public Cloud as the catalyst. They claim that on premise servers are dead.
I disagree. The public Cloud – while a great innovation and an enabler of many new services – will not be the death of on premise but rather its savior. I say this not for on premise equipment that is sitting in a corner unwilling to play with new friends, but rather for on premise boxes that are extroverted and play nicely with their public Cloud counterparts. The vanilla server or servers that sit in supply closets in hundreds of thousands of offices today will continue to be the backbone of workplace productivity far into the future…just with some changes (we like the term ‘cloudification’). On premise equipment needs the public Cloud to be successful just as the public Cloud needed the foundation of on premise equipment to be as successful.
Let’s start by exploring some of the limitations of the public Cloud. We’ll focus on cost because it’s easiest to analyze.
A perception worth reconsidering: public Cloud will take over the world because it saves money. On premise is DEAD. I don’t know if people trumpet this because they believe it or are just exaggerating to prove a point. It’s clearly not true. Certainly public Cloud can save lots of money in certain use cases but being 100 percent of your infrastructure is not one of them.
Let’s look at the environment to see why. If you move a server from a supply closet to a data center, you seriously increase the operating costs (air conditioning, bandwidth, etc.; I mean, it even costs money to have one server in a data center speak to another server in the same data center). You can see this in the prices charged by public Cloud providers. Recently, I did a quick Google search on external hard drives to see what a 1 TB drive would cost at typical electronics retailer and then looked at pricing for popular Cloud storage services.
AWS S3 Storage
3 year cost of 1 TB
$85 (includes electricity)
N.B. Prices as of July 6, 2016. Dropbox Pro is $8.25 per month. AWS S3 from Virginia is $0.03 per GB per month.
For this comparison, let’s say it costs $10 a year to power a local external hard drive (it doesn’t; it’s much less). It would still cost 3x as much to use Dropbox Pro and more than 10x as much to use Amazon’s S3 storage over a 3-year period. Though it’s harder to make an apples-to-apples comparison of compute, comparing physical servers to virtual servers on the public Cloud show price differentials just as stark. In other words, relying on 100 percent public Cloud infrastructure for your storage could be very expensive relative to on premise alternatives.
Conversely, on premise doesn’t work for all use cases either. What happens if you have wildly fluctuating needs for storage or what if your business changes so rapidly that you often need substantial storage in a hurry? The beauty of the public Cloud is that if you don’t use it, you don’t pay for it. Offloading computing resources during peak times is a great way to save money because during off-peak times you can turn off the public Cloud. The public Cloud can also spin up in a second to deal with a change in your business. Isn’t that better than stockpiling on premise equipment to meet business forecasts that may or may not come true?
It’s analogous to a retail business during Christmas, an accounting firm during tax season, or a Fortune 500 company that has in-house counsel but still keeps an outside law firm on retainer. In all these cases, during a good part of the year, the business uses its inexpensive, in-house resources. However, when it needs the extra resources (Christmas rush or tax season) or the extra expertise (outside law firm which has specialists), the business turns to higher priced, outside resources. Both in-house (on premise) and external (public Cloud) have their uses and both profit from integration with the other.
Here are several other use cases for when on premise equipment is preferable to the public Cloud:
- When you need speed. That local box is fast while the public Cloud is slower because the data center is farther away.
- When you don’t want to be internet dependent. My internet connection can go down and depending on the app, I can still be productive. I also don’t need a huge bandwidth pipe.
- When it needs to be simple to explain. I don’t need to ask a lawyer if my tech infrastructure breaks the rules of my regulatory agency (HIPAA for example). I can tell my customers exactly where their data resides.
Here are use cases where the public Cloud has the advantage:
- Ubiquitous access. Everyone wants to use mobile devices to access their files and use their apps, and the public Cloud lets you do this.
- Central management. In the Cloud, you don’t have 10 servers with compute or 10 storage servers, you have one pool of compute or one pool of storage. That makes management easier.
- DevOps. The public Cloud is a great way to develop new apps particularly ones where ubiquitous access (see above) plays a major part. With the world and business changing so rapidly today, new apps that deal with new workflows is essential to success.
- No obsolescence. If the world changes, there’s no risk that I will need to rip and replace my infrastructure.
- Bursting. Who can argue with only using the computing resources you absolutely need, when you need them.
Why can’t we marry the best features of on premise with those of the public Cloud? What is the definition of hybrid Cloud after all? Can I propose that the hybrid Cloud is simply integrating the physical servers we have on premise with the virtual servers in the public Cloud, while providing the best features of both. At Sphere 3D, we define on premise that has been cloudified as private Cloud infrastructure. Cloudifying to us is taking boring on premise servers and giving them public Cloud features like ubiquitous access, central management and bursting.
We call this the next generation infrastructure, where on premise has been transformed into the private Cloud and then integrated with the public Cloud as the complete cloud experience.
In my next blog post, I will discuss the necessary characteristics of private Cloud infrastructure.