Having attended Oracle’s OpenWorld conference in San Francisco recently, one of the things we’ve all learned is that Larry and Co. want to talk about the cloud. When Oracle, one of the archetypical on-premise IT companies, wants to pivot its entire business to the cloud, you know there is change afoot in the industry.
Here at Kaminario, we have been on our own cloud journey. Over the last couple of years, we’ve been fortunate to see an increasing number of customers seek us out to become the storage platform on which they build their cloud architectures. Whether it is end users building a private cloud solution, software companies building a SaaS offering for their customers, or public cloud providers building platform-as-a-service environments — we are now at the point where most of our new customers are using us to underpin their cloud offerings. They tell us that it is our uniquely-scalable architecture, as well as our ability to offer predictable performance for mixed workloads, which drew them to us.
But while Kaminario is excited to be so relevant to the next age of computing, we feel there is one downside of cloud deployments which has been overlooked.
By definition, a cloud implementation offers a consolidated, standardised platform where all workloads can be treated equally. Layers of abstraction such as virtualisation are used to add flexibility and allow movement of data and services. But a consequence of these layers of abstraction is that the storage layer has never been further removed from the applications it exists to support – especially as these platforms scale. For a cloud environment supporting multiple workloads, trying to work out which storage volumes support which applications is like trying to find your way through the San Francisco fog.
Bringing Clarity to the Cloud
Kaminario has always prided itself on having the most scalable storage platform. So to that end, we set ourselves a challenge: how to create a cloud storage platform which can scale without adding complexity. The answer, we believe, is to make storage application-aware.
Kaminario Clarity is a suite of features designed to bring storage closer to the applications it serves. Our customers already benefit from a comprehensive call home system which allows our support team to proactively monitor and resolve potential issues before they become service-affecting. By exposing this data to customers as a cloud-based platform, we can allow them to derive the benefits of real-time and historical analytics across their estate, as well as enabling multi-system management.
Also, included in Clarity is a form of Quality of Service (QoS) with a difference. We believe that traditional implementations of QoS are fundamentally flawed because they introduce complexity into the storage layer. The typical QoS features found on storage platforms today require thresholds to be set at very granular levels: IOPS or bandwidth limits, latency guarantees, and so on… all set on a per volume basis. This means administrators have to understand what each volume is used for and then adjust those settings as environments evolve – a lot of manual work just to configure a throttling feature. After all, QoS essentially makes storage go slower.
Our implementation of QoS is designed to keep things simple through the use of policies. You have a production VM? Set its volumes to use the Gold policy. Test/Dev environment? Probably a good candidate for a Bronze policy. Or how about if you’re running a transactional database? Wouldn’t it be great if you could set an OLTP policy which gives preference to smaller I/Os (typically user transactions) over larger I/Os (typically background processing such as batch jobs or backups)? Or perhaps switch to a Batch policy for the end-of-month billing run? For the first time, QoS can be seen as a feature for allowing applications to go faster, rather than a harness for slowing things down.
Intelligence is Key
Multi-System Management, Cloud-Based Analytics and QoS. All great features, but nothing that hasn’t been done before, right? Even if we think our implementation of QoS is ground-breaking. So where’s the magic?
The most exciting part of Clarity comes from a realization. It’s the realization that as a storage vendor with many customers across many geographies and industries, we have a wealth of anonymised performance information generated by our arrays calling home to report their vital statistics. By running analysis on that data and applying elements of machine learning, Clarity offers prescriptive analytics – advice on how to derive the best performance or value from Kaminario storage. For example, a suggestion that by setting this policy on these volumes you will see superior performance with reduced impact on neighboring volumes.
The more data points we gather from our install base, the smarter our algorithms can get at interpreting what each storage volume is doing and then making suggestions to improve service levels for those volumes. Suddenly, it’s less critical for administrators to understand and track what each volume does, because the storage itself can make intelligent guesses. That’s an important benefit for all customers, but never more so than for those building complex cloud architectures where a top-down view from the application is hard to track.
After all, the larger the cloud, the harder it can be to see what’s happening underneath. Sometimes we all need a little Clarity.