Scale-Out Storage Provides Performance Edge As Capacity Increases

Scale-out storage, or the ability to increase storage performance even as capacity is increased, has in the last couple of years taken its place on the list of "must consider" if not "must have" technologies for managing storage.

It is being embraced both by vendors who have scrambled to acquire the necessary intellectual property, and by customers concerned about the performance bottlenecks that come from their ever-increasing data stores.

Scale-out storage is a way of non-disruptively increasing the performance throughput of networked storage even as capacity is increased. It uses clusters of storage nodes, each with its own processing power, storage capacity, and I/O bandwidth so that as capacity is added, the processing power and bandwidth increase at the same time.

However, when presenting the technology to customers, just remember not to call it "scale-out storage," warn solution providers.

Sponsored post

Customers are increasingly asking for the capability, but they don't yet know the term "scale-out storage," said Greg Knieriemen, vice president of marketing at Chi, a Cleveland, Ohio-based solution provider focusing on the storage industry.

"They call it high-performance storage, or high-performance file storage," Knieriemen said.

It is a technology which customers ask for every day, even if they don't know the term "scale-out storage," said Jamie Shepard, executive vice president of technology solutions at ICI, a Marlborough, Mass.-based solution provider.

"Customers don't know they're asking for it," Shepard said. "But every time they speak, they're asking for it. They're asking for storage that can grow and expand without limitations."

Scale-Out Vs. Scale-Up Storage

The interest in scale-out storage stems from limited performance and network capabilities of traditional "scale-up" storage architectures as customers look to handle either larger data volumes or larger amounts of data as a single system.

As the amount of data stored within a scale-up storage system increases, performance of that storage eventually starts to decrease. This is because increasing the capacity of a typical storage array means pushing and pulling more data through a storage controller which eventually reaches limits in terms of its processing capabilities and network bandwidth.

As the processing and bandwidth performance become constrained, the overall performance of storage becomes a bottleneck. This can impact not only the speed at which data is read or written, but also the performance of several common storage services.

For instance, declining processor performance could lead to limits in number and frequency of snapshots of the data for setting recovery points and adversely impact replication performance. And, as network bandwidth reaches its limits, server access to storage across a storage network starts to slow.

With traditional scale-up storage architectures, most arrays can be upgraded either with more powerful storage controllers or multiple controllers within the same system, which provides performance and bandwidth relief until capacity grows to a certain point. Customers can also connect multiple arrays in a cluster configuration to increase performance and bandwidth. However, in both cases, the boost reaches a plateau based on either the relatively few controllers that can be added to an array or the limits in number of arrays that can be clustered.

With scale-out storage, however, performance and network bandwidth both increase as new capacity points are added to a storage system. This is because the additional capacity comes from the installing of more storage nodes, each of which has its own controller with one or more processors and its own network connections in additional to the hard drive or SSD capacity.

As a result, increasing the capacity of the storage systems by adding new nodes actually decreases performance and bandwidth bottlenecks.

Next: Storage Capacity: Cutting Back On The Bottlenecks

Scale-out storage also increases storage system reliability, as the data is spread across multiple storage nodes in such a way that the removal of one or more nodes has little or no impact on overall storage manageability or performance. Contrast this with scale-up storage where redundancy in terms of controllers, fans, disks, and other components are required to prevent a failure in one component from taking down the entire storage system.

Software for scale-out storage architecture is designed to allow the non-disruptive addition of new storage nodes or replacement of storage nodes which suffer from planned or unplanned downtime.

ICI's Shepard, who's company works with EMC's Isilon scale-out storage line, said that customers files are continuing to grow. "They are asking for scale-out more than they are scale-up," he said. "Isilon allows us to provide for their requirements."

Chi's Knieriemen said his company works with Scale Computing, an Indianapolis-based developer of technology which offers multi-protocol scale-out storage technology.

"If customers are looking for something more performance-driven than capacity-driving, we point them to Scale," he said. "But this is still a niche market. We don't see a lot of activity from our customers. But it's a growing business."

Increasing Vendor Interest

Storage vendors have responded to the increased need for scale-out storage with several acquisitions in the last couple years.

EMC late last year acquired Isilon, a specialist in the development of scale-out NAS technology. Isilon first developed its scale-out NAS technology in 2001.

Sam Grocott, vice president of marketing at EMC Isilon, said scale-out NAS is already being adopted as the primary storage solution for big data, or data which scales to multiple petabytes of capacity and is created or collected, is stored, and is collaborative in real time.

"In the future, as big data continues to proliferate throughout the mainstream enterprise in the form of massive home directories, large-scale virtualization deployments, rapidly growing general-purpose file shares and more, scale-out NAS will increasingly become the primary NAS solution for the enterprise data center," Grocott said.

Scale-out NAS is also being driven by continued pressure to do more with less, and as corporate resources become strained, the increased performance and capacity from scale-out NAS, along with better automation and utilization, will make scale-out NAS the better choice for storage architectures, Grocott said.

NetApp, which early last year acquired scale-out storage developer Bycast, has made scale-out capabilities a core part of the Data ONTAP storage operating system it uses for its entire unified SAN-NAS storage line.

Dell early last year acquired Exanet, an Israel-based developer of scale-out technology which Dell has already started integrating into its other storage technologies starting with its EqualLogic iSCSI line.

Dell also offers a scale-out NAS appliance in its Compellent line based on the open source ZFS file system, as well as a new scale-out object storage platform. Also available is the Dell Storage File System, a scale-out and scale-up file system which Dell plans to integrate across its Compellent storage line.

HP's 2009 acquisition of software developer IBRIX gave it the technology behind its scale-out HP X9000 NAS offering. HP also offers scale-out technology in its HP P4000 LeftHand storage virtualization offering, as well as scale-out capabilities in its 3PAR line.

Sean Kinney, director of product marketing for HP storage, said about 50 percent of his company's midrange and enterprise customers currently use scale-out storage technology, and that that number is growing rapidly.

NEXT: Wide Range Of Scale-Out Solutions

IBM developed its own scale-out NAS offering, the IBM Scale Out Network Attached Storage (SONAS), to support multiple petabytes of storage in a single file system. For scale-out SAN, offers its XIV line, which it acquired in 2008. XIV and SONAS can be used together via a gateway that lets SONAS customers use XIV disk for back-end storage.

Hitachi Data Systems' Virtual Storage Platform (VSP) array provides both scale-up and scale-out capabilities, as well as what the company terms "scale deep," or the ability to virtualize heterogeneous storage arrays connected to the VSP, said Christophe Bertrand, senior director for HDS's product marketing platforms and business continuity.

Bertrand said he scale-out storage will become pervasive as it is something enterprises will be looking to implement to make their data centers run more efficiently. However, he said, basic implementations of scale-out storage such as loosely coupling whole storage nodes across a switch network like Ethernet, RapidIO, or Infiniband, will not accomplish this goal. Instead, he said, a more efficient way to scale-out storage systems is through the tight coupling of storage resources through an internal switch or network where individual components can be added as needed without having to add an entire storage node.

Xiotech's ISE and Hybrid ISE storage systems provide scale-out storage capability by building redundant storage capacity, controllers, and power supplies into a sealed 3U enclosure that lets customers start out with a single system and expand capacity and performance together by adding new units. About 90 percent of Xiotech's customers are using its ISE and Hybrid ISE technology.

Scale Computing's Intelligent Clustered Operating System (ICOS) technology provides scale-out storage capabilities to 100 percent of its customers, particularly in the SMB market. The company said that the scale-out storage model provides room for repeat business as a primary revenue driver by allowing customers to buy only the storage they need and add to it as requirements grow.

BlueArc provides scale-out capabilities as an integral part of its high-performance storage offering, and provides customers the ability to select how they scale, including bandwidth or IOPS performance, capacity, and how they manage a system at scale. Gridstore offers a scale-out NAS software for SMB customers which is delivered primarily by MSPs. The software can be used in conjunction with existing industry-standard servers to turn idle servers into a scale-out NAS platform and managing their combined capacity as a single pool of storage. Coraid's EtherDrive technology provides scale-out capabilities via its EtherDrive nodes, which are based on industry-standard components. EtherDrive scales up to 65,536 nodes per network, each of which can feature a mix of SSD, SATA, or SAS drives and multiple RAID configurations.

Quantum's StorNext data management software provides high-speed content sharing combined with data archiving. Its Distributed LAN Client (DLC) provides NAS-like scalability to thousands of servers and compute nodes, the company said.

Avere Systems offers its Avere FXT series of high-performance, scale-out NAS appliances for adding NAS functionality to both existing and new NAS deployments and both single-vendor and heterogeneous NAS environments.

DataDirect Networks provides three different scale-out storage solutions in its GridScaler, ExaScaler, and NAS Scaler lines. About 85 percent of the company's customers use scale-out solutions.

Symantec provides a scale-out storage solution in its Symantec FileStore software, which is used by about 25 percent of the company's customers, especially those dealing with big data workloads, including media repositories, large archives, and Web content storage. The company expects scale-out technology will be the foundation of all enterprise file-based storage platforms in five years.