Cloud vs on-premises storage: Does it really matter?

New data using the SPECsfs2008 benchmark and Amazon Simple Storage Service highlights that innovation and intelligent use of technology is helping to build better storage architectures

Organisations always clamour for better performance for their mission critical applications. The goal is often to make decisions faster to maintain an advantage over the competition while delivering the highest levels of service to their customers. The ability to meet many of these requirements often relies on getting data into and through computing processes as fast and as cost effectively as possible. The current status quo for ramping up storage capacity and performance is to buy bigger storage servers loaded with an exorbitant amount of disks. Add these storage servers to bigger pipes connected to your workflow and hopefully scale performance. However, the real world does not always behave in the way of the theoretical model and, more crucially, organisations clamouring for more performance from mission critical applications are heavily cost constrained.

Normally, this type of article would pause and then go onto explain that “Cloud is the answer!” but the reality is that cloud is just one technology and process methodology that can help to span the gaps between performance, scale and cost. The reality is that different methods favour alternative use cases. The real secret which Software Defined Storage is poised to take advantage of, is the intelligent use of storage technologies that can negate the weaknesses in one method or accelerate underutilized strengths in another.

Two of the biggest emerging trends in storage are flash and cloud, not a surprising statement. Individually, massive game changers while together potential building blocks for creating innovative new data storage workflows and, more importantly, radically new cost dynamics. In very simplistic terms flash is fast and expensive, cloud is slow and cheap. Neither statements are entirely true but for mindshare, they are two common memes held by the casual observer. But a recent spec.org SPECsfs2008 benchmark, the measurement tool for network attached storage (NAS) server performance, highlights how the intelligent combination of flash and cloud storage use has the ability to dramatically change this perception.

The SPECsfs2008 benchmark is developed and managed by a committee with members from all the top-tier storage companies, amongst others. Results are peer-reviewed and approved for publication by members of the committee; it’s as close as you’re going to get to an apples-to-apples comparison of NAS performance of all the vendors publishing results. In a new published benchmark, a technical team from Avere ran its FXT 3800 Edge filer using the SPECsfs2008_nfs.v3 metric with AvereOS FlashCloud providing real-time tiering to Amazon Simple Storage Service (S3.)

The result was 180,171 SPECsfs2008_nfs.v3 operations per second with an overall response time of 0.86 milliseconds. However, unless SPECsfs stat comparison is your hobby, this number means very little by itself. In decoded terms, what this metric means is by using a 3-node Edge filer cluster in front of the Amazon cloud, the team delivered real workload performance comparable to having that Amazon simple storage cloud sitting on premises. The same levels of benchmarked performance is also attained using back-end Core filer storage from another type of cloud based object store or even generic ZFS whitebox storage arrays.

The full plethora of results from different vendors is available at the www.spec.org/sfs2008 website but it is not a case of crowing about who had the fastest speeds. The fundamental point is that these credible performance numbers mark a turning point. Finally, with intelligent use of real-time data tiering techniques, organisations can start to migrate workloads and applications that straddle on-premises and cloud computing, and finally take advantage of capacity based charging and ancillary benefits like disaster recovery and data portability. The discussion needs to move past cloud vs. on-premises, and on to a simple discussion about how do we meet our use case needs with the technology and budget we have available today.

Across the storage sector, innovators are testing new ideas that blend storage workflows and technologies. But the real push coming from enterprises is to spend less and do more. This tennet goes against the legacy storage principle of “increased capacity delivers increased performance,” as the moment this happens, traditional storage vendors begin to erode margin, revenue and potentially market share. By deploying new technologies which leverage software defined storage in combination with the solid-state disk, companies can move files to slower or faster moving storage based on any number of factors.

Organisations are now looking to take advantage of making the transition to this hybrid cloud compute and storage model without heaping investments in “disruptive” technologies. The new disruptor is no longer buying loads of disk to gain performance, but buying the right technologies that allow that disk to be fully utilized from both a performance and capacity standpoint. The bottom line is the right place (hybrid cloud) combined with right mix of media (spinning disk and Flash) combined with the right time (enterprises pushing to do less with more) will improve performance and drive down cost.