Regardless of what anyone says, if you’re supporting a relatively complex IT environment there is not a single cloud that can successfully, efficiently, and safely solve all your IT workload demands.
Why One Cloud isn’t enough
No, I don’t mean that one cloud has run out of capacity, I mean one variety of cloud isn’t enough. I wrote a blog a few months ago for GigaOm that focused more on the DARP (Disaster Avoidance & Planning) side of using more than one cloud, but in this case I’m going to focus on the usability and right placement of workloads.
Three primary criteria for workload placement
- Best/most appropriate infrastructure and cloud design to support the performance requirements of the workload (I/O, CPU, Memory, Latency, Network, etc.)
- The ability to split workloads across diverse cloud providers with similar design strategies
- Proximity to customers or other beneficial partners and services
As an example, at the Switch SuperNAP we have roughly two dozen different cloud providers; among those providers each has created a unique architecture of hardware and software, design to better suit the needs of specific workload/customer requirements. The benefit to customers utilizing a cloud in the SuperNAP is that they can diversify against the first two bullets, without ever leaving the premises (proximity).
Workload requirements were not created equal
Many cloud environments will likely be designed with characteristics similar to HPC (High Performance Compute). In other words many cloud environments will designed with lots of compute connected by high performance and high capacity networking to storage. Part of the reason for that is the efficiency and performance gains possible through concentration and proximity of compute and disk. Amazon’s cloud is still by far the most successful offering, few can argue that point. However, while Amazon is coming out almost daily with new features to make their cloud more usable, by and large they are still commodity cloud or single size (attempt to) fits all. Commodity cloud in today’s market is similar to virtualization acceptance in 2007. Everyone had finally decided to start using it, but they still largely only expected to use it for non-critical or low overhead workloads (print servers, web servers, Active Directory, DNS, test and dev, etc). As I’ve already stated the Amazon cloud environment is amazing, but it’s still not a fit for all jobs, and even if it was, my suggestion is that it wouldn’t be wise to place everything there.
In the SuperNAP there are several providers that are offering PaaS (Platform as a Service), and or IaaS (Infrastructure as a Service). While each environment is designed differently (some use SSD, others use Infiniband, etc), the main point is that there is a variety of service capabilities so that you can safely diversify your risk, while also right placing your workloads.
The security and ease with which you can share across clouds or move to public from private and back is a key characteristic of the ecosystem we’ve developed at Switch. Nowhere on the planet can you find the same number and quality of services, to create almost any IT opportunity like those found in the SuperNAP. When you consider your high I/O workloads or maybe the ability to move data to and from an environment quickly, safely, and cheaply by using in facility cross-connects you come to the conclusion that this is the type of environment you should be in, even if you don’t need it right as this moment.
Concern over Complexity
With any change in IT design the assumed future complexity should be taken into consideration. If you’re concerned about the complexity of having more than one cloud supporting your environment, that’s understandable, but not insurmountable. There are many consulting groups, and software solutions that can help you safely and effectively create the hybrid cloud environment best suited to meet your enterprise needs. Examples of some of the outstanding products and services that are part of our ecosystem and whom can help you safely make a cloud transition include Joyent, Trace3, Taos Mountain, HP, Cloudsigma, Profitbricks, Fusionstorm, CloudprovidersUSA, ACS, ServiceMesh, VMware, Cisco, EMC, Greenplum, Infochimps, Secure-24, etc., etc.. These are but a few of the 30+ cloud oriented services and products offered out of the Switch Technology Ecosystem.
Last but not Least
Don’t think of your future technology needs and how they’ll be solved based on what is economically or technical available and doable today. As you go through your consideration of what makes sense for your business you should consider some of the formerly accepted statements following:
“Virtualization isn’t for CPU or I/O intensive critical workloads” Circa 2007
“SSDs will never be in large scale production environments” Circa 2010
“Public Cloud will never work for critical production environments” Circa 2011
“Private Cloud isn’t and won’t ever be real cloud” Circa 2011
“Moving between clouds is too much trouble” Circa 2011
“I can’t get the power density I need to build the cloud/bigdata environment I want” Circa (for most) the present
All of the above have proved false, or at best temporarily true. So, as you go about considering your options of using a specific provider or diversifying across providers keep that in mind and be sure to challenge assumptions you’re making about what’s possible.