Server-side Flash Scale Out: What to Look For
Everyone and their brother seems to be promising “SSD and flash high availability and scale out” these days. But what exactly is it, why do you want it, and what features should you look for?
“Scale-out,” of course means expanding the amount of work your servers can do. The workload can be either block- or file-based, ranging from application serving to Web serving to database access. Such scale-out is only effective, of course, if it can be done as inexpensively as possible and as quickly as possible utilizing all the hardware and software platforms customers must support.
“Flash” refers to the high-speed solid-state memory deployed within servers, as opposed to the solid-state disks in storage arrays. Using this fast flash storage to create very low-cost, highly scalable and available server infrastructures is an attractive proposition for customers that need to quickly handle many more customers and much more data without extra cost or staff.
But not all the “solutions” that claim to deliver server-side flash are created equal. Some key questions to ask:
Does it create a single point of failure? High performance and scalability are not worth much if the failure of a single server can bring down an entire application or (in this virtualized era) multiple applications. Look for features such as dynamic and symmetrical clustering, load balancing, RAID and a shared file system that allows another server to instantly take over if one fails.
Does it create silos of applications or data? The same architectural constraints that make each server an island to itself can also make it harder for applications or users to share data. In today’s networked economy, such silos can cripple productivity and efficiency. The same shared file system, clustering and load balancing that eliminate single points of failure also allow any server or application to access vital data.
Does it span Flash, SSD and spinning disk? In addition to flash, most customers are also using SSDs as a storage cache layer, and low-cost high capacity JBOD (just a bunch of disks) with traditional HDD to store the vast majority of their data. To cost-effectively scale capacity across block and file storage while benefiting from the low latency and fast performance of SSD and Flash, look for solutions that support all those media types, and can intelligently route data to the most cost-effective media.
Does it provide appropriate data protection? If you’re going to run important applications on your agile infrastructure, you’ll need to protect the data. How well, and how easily, does your server-side scale out solution provide data protection if one of more drives fail? Look for solutions that support features such as snapshot and RAID.
How easy is it to manage? Using your existing server administrators to manage your storage can save you 40-50 percent on your staff costs. Look for centralized management capabilities that give server administrators the ability to monitor and manage your Flash, SSD and JBOD for maximum efficiency.
How efficiently does it use storage and server resources? The whole idea of virtualization and scale-out is to do more with less. Look for solutions that use a shared file system to allow all servers to access the same version of data without the performance and capacity hit of replication.
The flood of announcements about server-side caching and scale out won’t end anytime soon. Let us know what questions you have about what to look for, and if you want to stay informed about the intelligent software layer we’re adding to JBOD, SSDs and Flash.
Follow me on twitter @mvmsan
Experience for yourself why OVER 700 customers globally have chosen Sanbolic over more costly and complex alternatives. Melio for FREE