BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
Cloud storage performance matters across all types of environments, but are we looking at performance in terms of IOPS, such as Gets, Puts and reads? Is it about bandwidth? Is performance tied to response-time latency, or is it tied to the number of transactions and amount of file movement?
Whether you're looking locally, remotely or in a hybrid cloud environment, you need to understand how your resources are being used. For example, if you use solid-state locally or in the cloud, how do you think that solid-state or regular storage is performing? Oftentimes, somebody will buy storage expecting to get the 50,000 IOPS, 100,000 IOPS or 1 million IOPS they have read about somewhere, only to find they are getting a fraction of that. The reality is that their applications might be doing much bigger, more substantial work and not the small, marketing-sized jobs that produced those benchmark numbers. They may be getting the effective results they need, but they are looking at the wrong metric. Keeping the performance in perspective, you need to decide, "Is the application actually running better as opposed to what the storage system is doing?"
The good news is that focus is shifting from cost per gigabyte to cloud storage performance availability and durability. But we have to keep in mind that with performance, there is the issue of local, wide-area and cloud networks. For example, with cloud providers now offering solid-state storage, solid-state is fast, but are you doing it over the network? If so, the speed of your network matters.
We need to keep the metrics that matter in mind. Are we looking at activity, such as the amount of data being moved, or are we looking at response time? Most importantly, consider the applications you're supporting -- how are they going to benefit and what are you getting for your given cost? You need to look beyond the price per capacity and at the effective amount of performance you are getting. That's what matters.
Improving the performance of large cloud deployments
Guide to cloud application performance monitoring