In-Depth

Can Networking Guys Do Storage?

Acopia is a network-turned-storage-focused company with a new product for managing the network resources that support file access and data movement.

• By Jon William Toigo
• 03/25/2004

Brendon Howe, Vice President of Marketing and Business Development for Lowell, MA-based Acopia Networks, is quick to point out the pedigree of founder Cheng Wu and current CEO Chris Lynch. The two men were previously honchos at ArrowPoint, whose IP switch-qua-load-balancer business was bought by Cisco Systems a few years back. To hear Howe refer to the resumes of his bosses, you would think that his emphasis is on their ability to build successful companies that can be sold at a profit—no rare ability against the backdrop of so many failed tech firms.

But, Howe insists that Acopia is not another venture capital-funded startup with a “three-step business plan” (you know: create technology, attract interest of big vendors, sell company). He says Acopia is here to stay.

In fact, Howe emphasizes Acopia’s network technology heritage because he believes that it is what sets the company apart from contenders. Acopia is part of a growing set of network-turned-storage-focused companies that includes Dallas, TX-based Tek-Tools, whose savant CEO, Ken Barth, comes out of a network management software company, MicroMuse. Howe believes that this network perspective will help Acopia solve some knotty file storage problems the way that Tek-Tool’s Storage Profiler SRM package raised the bar for network-based storage management. He may be right.

Acopia has been operating in blackout mode since 2002, and avoiding the question from storage folk, “What do a bunch of ex-network guys know about storage?” Apparently, Acopia now believes that times have changed in the market, and the network is increasingly perceived as the place where the important work of storage gets done.

Actually, certain aspects of storage have been “networked” for a long time. Almost since the dawn of TCP/IP, you could mount a storage volume located on another server using protocols such as Sun’s Network File System (NFS) and Microsoft’s Common Internet File System (CIFS). Anyone surfing the Web today knows about Hypertext Transfer Protocol (HTTP) and File Transfer Protocol (FTP).

Collectively, these networking protocols have provided mechanisms for accessing data on remote servers and for transporting copies of data across a network infrastructure. That is networked storage by anyone’s definition. Now, the folks at Acopia are building network concepts, such as load balancing, into their otherwise storage-facing solutions. The company shortly intends to release its ARX Resource Switches to enable enterprisewide file/object virtualization.

We’ve been down this path before in this column, with discussions of global namespace servers such as StoreX, from NuView, and data movement and migration managers such as Rainfinity’s GridSwitch. ARX claims to deliver all of this functionality and more.

Like StoreX, the ARX Resource Switch will “virtualize” mountable NFS/CIFS volumes in the network. That is to say, it will present a customizable directory system to the client systems, while masking the actual NFS volumes from client view. This enables files to be moved from one place to another in the infrastructure without impacting client operations.

In this role, says Howe, ARX addresses the “pain” of ongoing file system consolidation, distribution, and optimization, which he summarizes using a complaint from one of his beta test customers: “I have had to change my desktop three times this year to account for changes in NFS mountpoints.”

The ARX Resource Switch also does some voodoo similar to Rainfinity’s GridSwitch. It adds intelligence to aid in migrating storage based on a variety of variables from one storage platform to another. Howe describes a rich set of functionality, including the ability to count accesses made to files and to use this data to create and implement policies for data migration—one of the lynchpins of true ILM.

The key difference between GridSwitch and ARX revealed at our briefing is that GridSwitch is brought on line (or in band) on an as-needed basis to facilitate behind-the-scenes data migration. ARX, on the other hand, is always in band and connected to a regular LAN switch, where it operates to evaluate storage accesses and to execute policies that will position data where it makes sense (in many cases, near the community of clients that make the most use of it).

In effect, ARX is managing the network resources that support file access and data movement. This is no big deal until your shop has a lot of files and a lot of file servers or NAS devices. The financial industry, manufacturing, audio/visual post-production houses, and other big file-system users are natural candidates for the technology, Howe says.

This vertical market focus puts ARX Resource Switches into direct competition with Silicon Graphics’ and Panasas’ high-end NAS on the one hand, and with Network Appliance/NuView on the other. However, Howe is quick to point out that his company is not selling storage, only intelligent data movement.

“We spent some time looking for a way to stripe file directories across platforms, but we quickly learned [about the idiosyncrasies of arrays manufactured by different vendors that prohibited this approach],” Howe says.

Instead, Acopia has decided to take a page out of IBM’s dictionary and to treat arrays as user-defined storage pools: “We can connect virtually any kind of storage to our switch to hold metadata, but the back-end storage pool is any kind of NFS or CIFS mount.”

According to Howe, plans are afoot to support network-based block storage protocols like iSCSI and also direct memory-to-memory mapping schemes such as RDMA over IP.

If Cisco doesn’t buy these guys, we could think of a few others who might be interested—at least in teaming up with them to support their larger customers’ file storage requirements. Do you have your ears on, CNT?

As always, we are interested in your thoughts. E-mail us at [email protected].