Many vendors who have scale-out storage solutions claim to have a future proof solution, but in most cases, scale-out storage does not guarantee future proof storage. The fundamental promise of scale-out storage is that if the customer needs more performance or capacity, all they have to do is add another node. The reality is that most scale-out systems are not future proof at all; in fact, these designs are so restrictive and inflexible that they are less future proof than alternatives.
The Motivations to Future Proof Storage
StorONE’s Product Manager, Gal Turchinski, and I recently conducted a webinar, “Future Proof Your Storage – End Storage Refreshes.” In our webinar, we talked about the various motivations for storage refreshes. After all, no IT professional looks forward to a storage refresh. There has to be a specific motivation to embark on such an arduous project. In the webinar, we cover those motivations and provide a list of what to look for in your next storage solution so you won’t have to refresh storage again. The primary motivation for future-proofing storage is to eliminate storage refreshes. You can watch the on-demand version of the webinar now.
The reason that many storage professionals need to go through the storage refresh process is to either meet a capacity demand or a performance demand. Many times they need to achieve both. Gal, in addition to the webinar, recently wrote a blog explaining why “reducing storage infrastructure costs” is also on the list of storage refresh motivations.
Great Question
During the webinar, a question asking “is scale-out storage future proof?” came up, and while at first glance, it seems like scale-out storage is future proof when you dive deeper, you may find it is not. In this blog, we’ll look take a close look at scale-out storage to see if it meets the future proof standard.
Before we determine if scale-out storage is future proof, we need to define what future proof means.
What Does Future Proof Storage Look Like?
First, future proof storage is more than just the ability to add storage capacity. Future proof storage is the ability to move to higher capacity SSD and hard drives while leveraging the full capacity of those hard drives. In most cases, data protection algorithms force customers to format new higher capacity drives down to a size that matches the older drives in the system, forcing customers to buy more drives than they need. In some cases, the scale-out architecture forces customers to use the new drives on separate volumes or even make them create new clusters.
Second, future proof storage is more than just the ability to add the latest high-performance storage media to the storage system. Most storage systems that run on a server with NVMe drive slots can, in theory, now support NVMe SSDs. Most systems can’t, however, deliver the full performance potential of those NVMe SSDs. The software behind these solutions are often Linux based and count on the Linux driver for NVMe support, which is not well optimized for the characteristics of NVMe SSDs.
Our findings indicate that most storage systems extract less the 20% of the per drive potential when they move to these new media types. A future proof storage solution means obtaining 80% or more of the performance potential of each drive placed in the system. Future proof also means having an architecture that maintains the maximum per drive performance promise as new drive technologies come to market.
With the definition of “Future Proof” understood, let’s see how scale-out storage meets these requirements.
Scale-out Capacity Does Not Guarantee Future Proof Storage
Again, the concept behind scale-out storage is that the organization can “just add another node” when they need to add capacity. This process is not future-proofing the storage infrastructure; it is an upgrade. All storage systems, both scale-out, and scale-up can add capacity. Most scale-up systems can add enough capacity to meet the requirements of most data centers.
There are two issues when adding storage capacity to scale-out storage systems. The first is when IT adds new drives to the system. Will the organization benefit from the full size of those modern drives? Scale-out primary storage systems, in particular, often require that each additional node match the existing nodes in terms of capacity, meaning that the extra capacity must go to waste.
The second issue is performance. Scale-out storage’s theoretical advantage is that since the architecture adds processing and capacity simultaneously, performance increases to keep up with capacity. Scale-up storage systems have a theoretical disadvantage as IT adds more capacity; the original CPU processing power has to manage more and more capacity.
Scale-Out Storage Performance is not Future Proof
Scale-out systems need to answer two questions when it comes to performance. The first question is, how is performance impacted as IT adds capacity? Again, scale-out vendors scale capacity by adding additional nodes, not by replacing smaller hard drives with larger hard drives. These vendors claim that because you are adding computing power with storage capacity, you are ensuring that performance scales as capacity increases.
The reality is that most scale-out systems don’t increase performance as IT adds nodes; they, at best, maintain performance. If you are cynical, then it might be more accurate to state that scale-out systems spread mediocre performance equally.
When a scale-out vendor claims linear performance scaling, they are, in most cases, assuming that as you add nodes to the storage cluster, you are also adding additional applications or workloads. If you have a single or handful of workloads that require more performance, then, generally speaking, adding a node won’t increase its performance.
When examined per node, most scale-out architectures don’t deliver anything close to the maximum per drive performance. They are at the lower end of our above statement that most storage systems provide less than 20% of the per drive performance potential. The reason for the lack of real performance increase as IT adds nodes is twofold. First, it is a combination of the scale-out vendor using legacy software libraries and drivers. Second, it is the overhead associated with managing a cluster of storage nodes.
The inability to easily mix media also comes back to haunt the scale-out architecture. When new drives come to market, like Intel Optane, for example, scale-out systems can’t typically leverage those drives with existing volumes and existing clusters. In most cases, the customer has to create an entirely separate cluster, which is no more future proof than any other storage system. The customer ends up with silos of scale-out clusters instead of silos of scale-up storage.
Scale-Out Storage Isn’t Network Storage Proof Either
Another challenge with scale-out storage is they are not networking future proof. Most scale-out designs are IP based and don’t support fibre channel, so if the organization needs support fibre channel infrastructure, they need to buy a separate solution and manage its refresh cycle. Even on IP networks, scale-out systems are not future proof. They, for the most part, don’t support new NVMe over Fabric standards (NVMe-oF). Scale-out architectures with NVMe-oF support do exist, but they are, once again, another, new, separate system that IT has to manage.
The Platform is Future Proof, Without Scaling-Out
Vendor marketing has conditioned customers to assume they MUST have scale-out storage to prepare for the future. That is because the alternatives these scale-out vendors have chosen to compare to are legacy scale-up systems. What if you took a platform approach to the problem? StorONE’s S1 Enterprise Storage Platform starts as a scale-up architecture, with an initial dual node, active-active, highly-available, configuration. Our platform can scale from 15TBs to 15PBs in capacity and over 1 million IOPS in performance. These capacities and performance capabilities should meet the demands of the overwhelming majority of enterprise data centers without ever having to add nodes.
Also, the S1 Enterprise Storage Platform can extract 80 to 90% of the performance of drives placed in it. This per drive performance claim includes high-performance drives like Intel Optane and NVMe SSDs. The platform can support any network protocol, fibre channel, iSCSI, NFS, SMB, and S3. In a few months, NVMe-oF support will be available to customers with a simple no-charge software upgrade.
We believe that S1 is the most future proof solution on the market today. It is more than future proof, it is future proven!
To learn more, watch the webinar below or sign up for a personalized demo here.
