You’ve been hearing a lot about cloud computing these days. If the many definitions of the Cloud seem a tad abstract and esoteric – you’re not alone. Go to any IT tradeshow or consulting conference and you’ll see liberal references and pictures of the Cloud. Many messages are often mixed with cloud enabling technologies like Software-defined Storage (SDS), hyper-converged/hyper-scalable virtual Server SANs, and the many different flavors of Flash. Let’s not forget vendor messages around Big Data, Virtualization and 2nd to 3rd Platform transformation to Mobile Computing. It’s no wonder they call it the Cloud – it’s anything but clear!
Whatever it is you are trying to do with your cloud, whatever use case you are trying to solve, it is likely that SDS, Flash, and ultimately software defined data center (SDDC) platforms will be the key building blocks. Why? As performance requirements of cloud computing overtax the max performance limits of traditional spinning media, silicon based NAND Flash with no moving parts bridges the I/O performance gap between today’s hyper fast CPUs/Cores and much slower spinning media (i.e., 3,500 IOPs/SSD vs 150 IOPS/HDD at system level, respectively). A purely software-defined and implemented control and data services application forms an independent, de-coupled abstraction layer from the underlying physical layer of storage hardware resources. SDS is optimally designed for use in today’s modern heterogeneous, multi-vendor, and multi-hardware platform environment. SDS and the greater overarching SDDC umbrella platforms like ViPR and the Elastic Cloud Storage (ECS) Appliance truly make centralized management of the storage infrastructure more flexible, easier to deploy, and simpler to manage. This provides a centrally managed control/access point among disparate physical devices/resources in the process plus the side benefits of hardware agnosticism and vendor independence. Who doesn’t like an easy button?
These days, all flash arrays (AFAs) share equal billing with hybrid storage and server side flash as being the drivers of Flash/SSD storage usage and cloud enablement. A TechTarget 2014 survey reports 25% of respondents have already implemented an AFA. Why? Simple economics. The cost curve on enterprise grade multi-layer cell (MLC) Flash has fallen to the point where AFAs are more affordable and better alternatives where IOPS and throughput considerations are paramount. In 2010, the average installed flash capacity was about 1.5TB. Today it’s more like 7.5TB with that number expected to grow to 9TB in 2015 (source: 2014 TechTarget survey article). Another point worth making is that really good AFAs (like EMC’s XtremIO) are actually commodity hardware based but come with some really sophisticated and intelligent storage software managing those SSDs with Flash optimized features like de-duplication, in memory metadata management, data protection, etc. As for hyper scale-out/scale-up applications with massive parallel I/O performance from commodity servers, it’s hard to top ScaleIO and ECS Appliance. ScaleIO pools together PCIe Flash cards, SSDs and HDDs into one aggregated Server SAN into various performance or capacity tiers. ScaleIO installs as a standalone software application or as part of a hardware appliance (i.e., ECS Appliance). 10M IOPS from ScaleIO with less than 50 servers? No problem. Simply create a pure Flash only ultra-performance storage tier for highly randomized, non-sequential read intensive workflows.
PCIe Flash cards (for server side I/O cache and I/O acceleration) and SSDs are still more expensive compared to tried and true spinning disks especially SATA and NL-SAS drives. One has to consider other cost metrics beyond dollars/GB when evaluating a Flash energized deployment such as dollars/IOPS or dollars/MBs on the performance side of the storage equation. And how do you put a value on your users experience expecting instantaneous screen response with very low latencies? This depends on the specific user community needs and SLAs. Today’s Datacenter planning and operations must balance fluidly between performance, capacity, cost, flexibility, scalability and ease of use requirements. Storage architectures, solutions, and deployments offering the most flexibility, resource elasticity, painless scalability and requisite performance are key enablers of cloud computing. Judicious use of server Flash, SSDs in hybrid arrays or AFAs along with installation of hyper-converged/hyper scalable SDS systems provides significant, tangible benefits to the on-site datacenter and internet based cloud computing apps.
In the end, it’s all about getting on-demand, fast access to shared resources to drive applications, use cases, and data transfers whether directly accessed from onsite datacenters, remote datacenters, or via cloud based resources. Clouds can be public, private, or hybrid in nature consisting of local and/or not so local physical resources. SDDCs unify or federate multiple vendor, geo dispersed and disparate storage (and other) hardware resources that now typify modern datacenter purchasing and operations. Moreover, the rapid proliferation of virtualization (VMs) in of themselves is a software-defined abstraction layer along with cloud based computing and has created heavy I/O, storage, compute, and network data traffic stresses on traditional datacenter resources (and management) just by their sheer nature of randomized, high volume, multi-user, and VM simultaneous access and high availability demands. Flash and SDS meet these taxing demands.
EMC offers solutions and products that cover the entire gamut of today’s modern IT datacenter requirements including cloud computing. The ECS Appliance and ScaleIO are two such solutions. EMC also offers a complete Flash and SDS deployment strategy. So energize your Cloud with Flash, SDS and EMC!