Cloud performance. It's perceived by many to be the soft underbelly of the cloud service provider industry. And...
it's one of the main reasons that businesses today are resisting the move to deploy their mission-critical applications into the cloud.
In this tip, you will learn what some cloud storage providers are doing to help guarantee application performance while mitigating the risk of resource contention within their multi-tenant infrastructures. You will also learn the types of questions you can ask a prospective cloud storage provider to see if it's up to the task of managing your business systems.
All flash, all the time?
Some cloud storage providers have designed their storage infrastructure with all-flash storage arrays to, in effect, guarantee that all application workloads will have access to blazing-fast storage resources. But the obvious question is one of cost. Compared to conventional hard disk drives (HDDs), flash carries a significant premium. And if only a relatively small percentage of data needs flash access, isn't it likely that businesses would be overpaying for the luxury of deploying into an all-flash environment?
Interestingly, to attract clients that might otherwise be reluctant to commit to the higher premiums of operating in an all-flash storage infrastructure; some cloud storage providers are offering flash capacity at the same cost as conventional HDD space. The provider, in effect, uses flash storage as a loss leader, then makes up for it in CPU consumption. So, those businesses running heavy application workloads in their cloud will accrue higher monthly costs than a business will that has sporadic application activity.
Noisy neighbor nuisance
While flash is capable of driving much higher I/O throughput rates than spinning media can, all-flash arrays are equally vulnerable to the noisy neighbor issues that often plague heavily virtualized environments. A noisy neighbor is when a rogue virtual machine monopolizes the storage I/O resources to the performance detriment of the other VMs in the environment. To circumvent this problem, some cloud storage providers are deploying all-flash arrays, such as SolidFire's platform, that have built-in storage I/O quotas that can be set at an individual VM layer.
Large enterprise environments and cloud storage providers, in particular, need this level of management detail to ensure that storage I/O resources are assigned to those applications that have the highest demand for them. This can help ensure that each virtualized application tenant in the environment will always have access to a predetermined number of storage IOPS and prevent rogue VMs from upsetting the virtualized infrastructure applecart.
Variable service level support
These types of flash systems allow the provider to set up various service levels -- platinum, gold, silver, bronze, for example -- and then allocate resources based on the customer's application storage I/O throughput needs. But predicting application performance needs can be an inexact science. As is stated in any financial prospectus, past performance is no indicator of what future performance will be. So, a good question to ask a prospective cloud storage provider is whether they can dynamically handle moving between various service levels on the fly. In other words, if your application is set to a gold service level and is assigned a fixed number of IOPS, what happens when performance spikes and you need to suddenly change to a platinum service level? Likewise, how can you drop back down to gold or silver, when application performance demands vacillate over time?
Testing the waters
Some of the early adopters of heavy computational workload computing in the cloud are big governmental and scientific agencies that need auxiliary computing and storage space to process enormous volumes of data. Think of the Large Hadron Collider in Geneva, where simulations of the Big Bang are taking place. Processing the output from these experiments requires enormous computational power and as a result, institutions like these are bursting workloads into the cloud to help speed the time it takes to process the output from these experiments.
Bursting application workloads into the cloud is a good way for businesses to test-drive cloud application provider services. For example, application owners can deploy a nonproduction instance of a heavily utilized database in the cloud and simulate peak activity to see how well the application performs. In fact, some companies like Load DynamiX have virtual testing appliances that can simulate application workloads and run them in the cloud to help IT planners determine exactly what their storage I/O throughput rates need to be. Then the virtual appliance can generate the load to see if the cloud provider's infrastructure is up to the task.
The all-flash cloud storage service provider could offer some interesting capabilities for those businesses that need a storage performance pressure relief valve. With many businesses mandating their IT organizations to cut costs and reduce capital spending, IT planners need to find creative ways to satisfy business needs without breaking the bank. By selectively bursting critical business application workloads into the cloud, it is possible to meet application service levels and keep costs in check. The key, however, is to make sure your provider has the ability to meet your needs as they change over time.
Cloud storage provider comparison
Attrition of cloud storage providers wakes up channel
Are cloud storage providers good for primary data storage?