Recently, I found myself working with a customer interested in leveraging a hybrid cloud model. The CIO was aggressively looking to move part of their workload off into the cloud; specifically archive workload. The customer has around 120 TB of usable capacity, about 25 TB of that would be considered archive and is a candidate for cloud placement. The bulk of their environment would remain on premise. When I asked “what business problem does a hybrid cloud solve for you” I got a very typical response:
- Less to manage
- Less environmentals to pay for
- Opportunity cost of data center space
- Availability of the cloud
- Elastic abilities
- Capital expense vs. operation expense
When I asked “how would your application manage both onsite for active data and cloud storage for archive, I got a back the question “Can you tell us how other customers leverage the cloud?”
To answer that question, I thought it best to explain what most cloud providers offer. It can be broken into three categories with an added caged service. Most cloud providers offer any one or combination of these ideas:
Software as a Service (SaaS):
Cloud providers manage all aspects of the environment providing you with a consumable product to meet a specific need. Specific business tools like email, CRM, etc. are examples of this type of service. Yahoo email, Google email, and Google Docs are examples of SaaS that we all probably consume as individuals (or businesses).
Platform as a Service (PaaS):
Cloud providers manage both hardware and software layers. They provide compute platform complete with any backend software required to run your platform (platforms include technology ideas like data bases, web servers, backup targets, etc.). In most cases, they build all the backend technology you need, you manage the overlying application. Some providers break this into further boutique offerings. An example would be Backup as a Service (BaaS – not to be confused with backend as a service, a completely different offering for mobile devices).
Infrastructure as a service (IaaS):
Typically, a hypervisor runs VMs in the cloud, you pick the hypervisor you want/need and the cloud provider builds an elastic pool of resources available to you (sometimes tailored to your needs). Your ability to manage the actual hypervisors vary from provider-to-provider, however, in most cases, you have access to the operating systems that run in this type of environment.
The provider gives you physical floor space to place your own hardware. The provider takes care of environmentals and physical security. In some cases, the provider offers to manage your technology for you being your onsite support mechanism for the technology. This is great for customers that want to own their own environment without owning data center responsibility.
In almost all of the “X as a service” cases, the provider manages ideas like network, storage, firewalls, and most physical aspects of the environment. They hand you an application (SaaS), platform (PaaS), a complete compute environment (IaaS), or some hybrid combination thereof, to use with a commitment to provide you this service and availability (SLA).
There is no magic in the cloud. 10TB disk drives don’t currently exist; the same technologies available to you today are what they use. Cloud providers look to leverage economies of scale you cannot as an individual business.
The SaaS email offering is an easy offering to discuss. Most businesses have at least two individuals that manage messaging. It may be a full time job, or part of their job, but there is usually at least 2 people caring and feeding email. Most companies consider email a mission critical component to their business. Further, most companies have governance and compliance requirements around messaging. SaaS providers have figured this out and targeted this market. They can build a SaaS offering to give meet your business requirements with an aggressive SLA at a cost near or equal to your cost to maintain resources, licenses, capacity, compute, etc.
Understanding the cost of data in the cloud
Back to our example:
The CIO stated they can get cloud storage for about $.65 per GB/year. Their current 25TB need at $.65 per GB equals about $16,250 per year (forever) and will likely grow year-over-year as they archive more data. The cost may drop in the future as more dense drives find their way onto the market, but it’s not likely. It depends how much capacity they consume and that is dependent on the underlying business requirements.
I fully expect emails stating the $.65 number is high. You can do a quick search and find providers out there that will charge $.10 per GB/year. If you continue reading on, you will find that those providers also charge you for data ingestion, access, every transaction, and that transactions size. Also, you will find some charge you by GB, and others by GiB. I’ve found the cheapest up-front cost/GB providers are usually the most expensive in the long run. Additionally, there is a different rate and SLA when you want to move away from these providers. Be careful what you sign up for, and understand exactly what you are getting into when you sign a contract. Based on this customer’s example, the estimated cost for their 25TB is around $.65 per GB/year
It turns out, our example customer’s application does not natively support archive functionality. They would require a 3rd party package to manage the data mobility effort between on premise and the cloud. An additional expense usually licensed by available TB to the underlying application or system.
They are considering a new disk array, capable of auto-tiering. This functionality will automatically move stale data to the most cost effective tier over time. Their current array is dated, and needs to be refreshed. The current array does not have auto-tiering functionality.
To meet their 25TB, we determined we could build a single RAID6(12+2) pool of storage within the proposed array and add a hot spare. These 15 disk drives configured like this would take 3U of space and provide about 32TB of storage. The disk array could manage data placement and would not require any care and feeding. The new disk array includes 3 years of hardware and software support.
The list cost per to add the 32TB of archive data capacity to the proposed array is nearly $31,000 (without environmentals). This equals $.93 per GB for the 3 years of hardware and software support. Or, roughly $.31 per GB/year. Nobody buys technology at list price. I won’t work backwards through that number to get to the actual spend for a GB/year. I’ll assume that installation effort, migration, and power/cooling for 3 years for the 15 hard drives, disk shelf, and capacity licenses will make up the delta between list price and actual cost. Reality is the delta between the list price of the technology and the actual cost of the archive capacity exists; the complete cost of the archive capacity is less than list price.
The $.31 per GB/year is a long ways away from the $.65 per GB/year they are considering. Keep in mind, the $.65 per GB/year does not include software purchase necessary to manage on premise and could storage as one contiguous space.
This model might not always favor hosting your own solution. If this customer was not considering a disk array refresh, the math may not work out. Keep in mind, there is no magic in the cloud. The high capacity drives we are considering adding to their proposed disk array are the same drives the cloud provider uses. It will still require the same quantity of drives to meet that capacity requirement regardless if it’s in the cloud or on premise.
If you are interested in going through this type of study, let us know. SIS will be happy to assist you as you figure out what your journey to the cloud looks like.
Let us know if you have any questions….. – Contact Us Today