Scale-Out File System (SOFS): A Unique Scality Strength

Survey the storage industry, and you’ll find no lack of varying opinions, viewpoints, and claims. But you’ll also find broad agreement that Gartner industry reports are carefully researched, reliable, and bias-free. This makes the Gartner Magic Quadrant ground zero for a deep dive into Scality’s overall strengths, and our SOFS advantages in particular.

In 2017, as in 2016, the Gartner leader’s quadrant contains only one software-defined storage vendor. It also names just one whose platform supports both object- and file-based data simultaneously. This vendor is one and the same: Scality. Let’s explore the reasons behind our coveted leadership designation.

The Scality RING has an integrated scale-out file system (SOFS) that runs natively on top of our underlying software-defined object-based platform. Our SOFS behaves just as you’d expect: exactly like a file system. But everything underneath it is an object store — extending the many benefits of object architecture to your file-based applications.

Object Storage Economics for File-Based Data

As a software-defined system, the RING performs independently of the hardware back end. It runs on standard, commodity x86 hardware — virtually any brand or model. This creates a two-fold advantage: One, you can use off-the-shelf servers of your choosing. Two, you gain the freedom to add the densest, cheapest hardware as soon as it becomes available. Say good-bye to expensive proprietary appliances, complex integrations, and hardware over-provisioning (we’ll explain this last point in a moment). With Scality, you only buy what you need, when you need it.

Scale Capacity Without Limits

Scality can support billions of files, with file sizes up into the petabytes — easily the most unrestrained scalability in the industry. We have a number of customers running real-world Scality-based production systems with SOFS capacities between 5 and 10-plus petabytes.

When we say “without limits,” we’re also talking smooth, non-disruptive operations. To expand capacity, you simply add servers. The RING software does the rest, with no downtime or manual load balancing required. The same goes for new software releases: Scality upgrades happen transparently, with no interruption to system availability.

Linear Scale-Out Performance: the Industry’s Best

Scality provides scale-out throughput through our connector layer. The single-connector throughput numbers are among the fastest anywhere: 500 MB to 1 GB/second with SMB — as much as 5 times speedier than our nearest competitor. With NFS, we’ve clocked up to 1.5 GB/second over a single 40-gigabit Ethernet link.

As with capacity, Scality performance also scales independently from hardware, so you don’t have to over-provision servers to increase data throughput. All you do is add connectors. As you can see in the chart below, throughput scales just about linearly with each added connector.

All this is in stark contrast to traditional NAS-based systems, which intrinsically don’t scale well in terms of throughput. You are bound by the constraints of a single server which limits both capacity and performance. Scality frees you from that constraint, providing a platform which horizontally scales both capacity and performance.

Diagram (Click to Zoom): Scality RING with Integrated File System and S3 Storage

 

The Most Comprehensive Disaster Recovery

The best remedy for disasters is to keep them from occurring in the first place. Scality does this in a number of unique ways. The system has no single point of failure (SPOF). The RING architecture is completely distributed, both for data and metadata. Our design is “shared nothing,” meaning there is no central “anything” on the system. Every node contains a portion of the data, with real-time erasure coding to ensure exceptional availability and durability. Your original data is kept intact, an achievement that is unmatched in the industry.

Another unique Scality design feature is how SOFS data is accessed. Other solutions make use of an external, dedicated gateway node — performance-wise, this is a classic bottleneck. And disaster-wise, it’s the very definition of a single point of failure. If you lose your gateway, you lose access to your data.

Not so with Scality. As our SOFS system runs natively on an object store, it needs no external gateway. Therefore, no throughput bottleneck (several connectors can transfer data simultaneously), and most importantly, no single point of failure. The filesystem namespace is distributed and protected across the nodes.

Even with all this going for you, let’s say one or more of your hardware elements fail (and let’s face it, disks and servers malfunction now and then). In such an event, Scality can give you zero RPO, or close to it, thanks to all the factors above plus another singular benefit:

The Best Geo-Distribution Options

If you have 2 or 3 sites in fairly close proximity — one in Paris and a second in Amsterdam; or one in New York, another in New Jersey, and a third in Boston — you can stretch the Scality RING across them, distributing the data with erasure coding. The explanation is a bit involved, but the upshot is that the loss of any one site will not take the RING down, so 100% of your data will survive the event. Hence you get zero RPO. The RTO depends on the rest of your infrastructure.

If your datacenters are more remote from one another, which is often the case, your DR strategy will be asynchronous data replication. Scality performs this very well. At petabyte scale, we do it better than anyone else. All in all, any which way your data centers are geographically positioned, Scality offers you a better DR solution.

Taken together, these advantages explain why we’re positioned in Gartner’s leadership quadrant. And why — if you’re looking for a storage platform to run your digital business — Scality deserves to be on your short list.

Leave a Reply

Your email address will not be published. Required fields are marked *