Last weekend I enjoyed what surely has to be the final ride of the season. The weather was a balmy 65 degrees, and in the past 25 years of being an avid motorcyclist, I cannot remember getting out so late in the year. My buddy and I had a blast carving up the empty roads of western Massachusetts and enjoying the scenery.But once the ride was over, I quickly shifted into the year-end ritual of shoehorning multiple bikes into my garage. This annual ritual inevitably re-opens my wife’s question of justifying three bikes. Just like the question, my response is pretty much same every year – although personally I feel it, just like a fine wine, is getting better with age:
“Yes, the three are all basically modes of transportation designed to take the rider from point A to point B – but they are very different in terms of performance and other attributes. The road bike provides comfort and cruising performance for those long summer trips up north. The off-road bike has highly optimized suspension and overall durability for trail riding. The track bike is lightweight and powerful with fantastic brakes and superb handling.”
I like riding motorcycles, and I cannot reasonably expect one bike to perform well when the riding environments demand such different skills and machine attributes.
Okay, I could purchase a super high-end dual sport from BMW or similar, but my pockets aren’t deep enough to justify the expense, and I’d be buying way too much machine for any two out of the three standard riding environments.
Reflecting on my bike situation, I realized the similarities between it and the world of data storage – another topic I am passionate about, which consumes even more of my time than motorcycles.
Different storage use cases demand very different storage performance attributes: throughputThroughput is the measure of the amount of data that may be written to or read from a system. A simple analogy is to compare throughput to the number of lanes on a highway – generally speaking, more lanes mean more traffic with less congestion., IOPSIOPS (input/output per second) is a measure of the number of transactions a storage solution is capable of handling. Think of the tollbooth, which is generally architected for either high volumes of passenger cars per minute or a lower volume of eighteen-wheeler trucks. It’s tough to optimize those booths for both transport types. and latencyLatency is a measure of how quickly the storage solution can access data. If you think humans are intolerant of slow response times for data requests, you simply cannot imagine how impatient software applications can be – most require no more than a few milliseconds before they start complaining or even give up..
No single storage solution is optimized for all three dimensions, despite vendors’ data sheet claims. For example, a solution optimized for online transaction processing may be a horrible match for big-data applications – the storage demands for each use case are very different.
Online transaction processing applications require low-latency response times and high IOPS numbers. These requirements are driven by hundreds of users and applications, accessing querying databases and updating records. When the transactions involve purchases, response times can make the difference between a purchase and visit to a competitor’s website – meaning response times and revenue can be directly linked.
Big data places different demands on storage; it often drives bandwidth-intensive solutions, as data volumes can be enormous when multiple datasets are analyzed using parallel processing. Because the computation takes time, throughput is weighted higher than storage latency.
In other words, to determine the ideal storage performance needed for your system, your first priority is to figure out the types of apps you’re using, their respective requirements and how to match them with the right storage solution attributes.
Storage purchases are complex even when the application requirements are well understood. The myriad of solutions on the market all appear to satisfy the business requirements but at vastly different price points. The buyer could opt for an all-flash array (AFA) solution, meeting application performance criteria – but at what cost for the required capacity? This feels like the aforementioned BMW motorcycle option where one could easily over spec a solution. A lower-cost alternative could be hybrid storage with a flash tier to meet the high performance demands of certain applications. But how much flash versus spinning media is needed to ensure the optimal balance between performance and capacity?
Storage buyers are typically asked to focus first and foremost on performance attributes, but as anyone who has owned and operated storage solutions knows, other attributes are equally important in any buying decision: scalability, service and manageability, ecosystem support, and data integrity/security. These attributes typically require a deeper level analysis and detailed cost modeling beyond the information that’s easily mined from the back of a data sheet.
These are non-trivial challenges and the problems are compounded when the business requirements change or when that new application is introduced, or that in-house application is pushed to the cloud.
ClearSky offers a refreshing approach to all this complexity with the introduction of our global storage network. As a fully managed storage service, ClearSky ensures the solution is optimized for the environment with configurable connectivity choices, expandable cache configurations and other performance tuning options. You no longer need to worry about replacing failed drives, configuring RAID, building duplicate systems for backup and disaster recovery and all of the other endless performance and availability headaches. ClearSky supports multiple workloads with a mix of IOPS, latency, and throughput requirements in the most optimized manner.
So back to the annual discussion about why three motorcycles, and how much easier life would be with that extra garage space. Well, I am waiting for a global storage network equivalent to introduce a MaaS (Motorcycle-as-a-Service) solution so I can pick and choose the optimal bike for the chosen riding environment – and I am free to change machines as the mood takes me.
Learn more about meeting performance requirements in a rapidly changing storage landscape.