The Intelligent Storage Network: Sun's latest pursuit

What are Sun's plans in the distributed storage market? Do they look promising?

By Rick Cook
SunWorld
May  1998
[Next story]
[Table of Contents]
[Search]
Sun's Site
Abstract
One of Sun's biggest business opportunities these days lies in the heterogeneous storage market. The company's Intelligent Storage Network (ISN) initiative is built around Java and storage technology it has acquired from Encore Computer Corp. What are the roles of the major components in Sun's distributed storage strategy? And what are analysts saying about Sun's chances for success in this emerging market? (2,500 words)


Mail this
article to
a friend
Sun's decision to move into the heterogeneous storage market in a big way reflects both perceived opportunities and the changing nature of the computer business.

Part of this, of course, is simply an acknowledgment of the growing importance of Windows NT in the enterprise. Windows NT represents one of the fastest growing segments of the computer market, and while Sun may not like it, the company can't very well ignore the opportunity to profit from it. One way to profit is to sell storage systems that will attach to NT as well as Solaris.

Sun's move makes sense from both a business and a strategic perspective. According to Sun, the heterogeneous storage market is currently worth about $28 billion a year. It is expected to grow to about $35 billion by the year 2000, and Sun intends to have a major piece of it. Sun is already the leading storage vendor in the Unix market. With its new initiative, the company intends to double the size of its storage business by 2001 with more than one third of the business coming from heterogeneous platforms.

The strategic argument is even more stark. The move to heterogeneous storage is vital if Sun intends to be a major player in enterprise computing. "In my opinion, they have to," says David Vellante, a senior vice president at IDC, a market research firm headquartered in Framingham, MA. "You're crazy if you want to be an enterprise server player and you don't have a strong storage business."

But beyond business opportunities and corporate strategy, there's also a bit of sheer serendipity at work here. Sun's decision reflects a sea change in the nature of the heterogeneous storage market, one that plays directly to Sun's strengths in networking, technology, and sophisticated management capability.

The classic storage model
Traditionally enterprise storage architecture in the non-mainframe world is discrete. Each chunk of storage capacity, whether a disk array or a single device, is attached to a server which is in turn attached to the local area network. Requests for information are routed to the proper server that sends the information back over the LAN.

Like most architectures, this one breaks down if you push it too far. The big problem in enterprise storage right now is that more and more companies are pushing past its limits. Storage is growing explosively, thanks to a combination of hunger for ever-increasing amounts of data and the continually falling prices of storage that make holding that data economical. Past a certain point and a certain growth rate, classic architecture becomes progressively harder to manage, harder to expand, less reliable, and more of a drain on the network.

One of the key jobs of managing this architecture is knowing which server has what information. As you add servers, management obviously gets harder because each server is connected to one chunk of storage. Adding significant amounts of storage means adding new servers, otherwise the server becomes a bottleneck. That usually means repartitioning the data among the servers to make the best use of the new capacity. The more servers, the more points of failure, especially since the servers usually can't take up each other's load.

Another problem is that the LAN clogs up with data flowing to and from storage servers. The files that the servers are moving tend to have significantly different characteristics than the traffic generated by other nodes on the network. (For example they're larger.) That makes it harder for LAN managers to tune the network for optimum performance.

The SAN model for storage
Enter the storage area network (SAN). It ties enterprise storage together with its own high-speed network and intelligent arrays, usually connected to the enterprise network (the LAN) through one or more servers. The SAN itself can use routers, gateways, and hubs like a LAN. The difference is that everything is optimized for storage, and the protocols used are storage protocols, such as SCSI, rather than communications protocols like Ethernet.

SANs offer considerable advantages in the modern storage environment. With a SAN, adding new storage doesn't mean adding new servers. Storage can be widely distributed and added as needed. All the enterprise storage is on a sub-network that can be managed like any other network. While the bandwidth requirements on the SAN are quite high (as close to bus speed as you can get), the entire LAN doesn't have to be upgraded to those speeds.

This isn't exactly a new idea. Such networks have been commonplace in the mainframe world for years. DEC's VMS network is based on them, for example. However, the concept is only beginning to spread to non-mainframe systems.

In non-mainframe systems, a SAN has another characteristic that is even more important to Sun. SANs are by their nature operating-system neutral. The enterprise storage becomes a separate subsystem with its own network and only minimal, well-defined links back to the enterprise OS. This makes it easier to sell and support heterogeneous storage systems, and it takes the focus off the operating system the enterprise has chosen.

The early adopters of SANs include multimedia companies such as post-production houses. Editing broadcast-quality video with associated sound typically takes bandwidth of about 21 megabytes per second per video stream (two high-quality video streams and several high-quality audio channels). With growing storage demands and the popularity of storage-intensive technologies such as data warehousing, SANs are now moving into the general corporate world. Already the manufacturers of storage networking products have formed the Storage Industry Network Association (SNIA) to help develop the market.

Advertisements

Sun's opportunity
If you've noticed that this brave new world of storage networking sounds like Sun's mantra, The network is the computer, you're not exactly alone. "Our extension of [the network is the computer] is the network is the storage," says Jack Androvich, director of strategy and business operations at Sun's enterprise servers and storage group.

And here is where the serendipity comes in. Sun is already a powerhouse in networking, high-speed network technologies such as Fibre Channel, scalable systems, and related areas. The move to storage networks lets Sun capitalize on all that experience to take a commanding position in an emerging market.

"What we're looking to do is to apply the network technology Sun is so famous for to our storage products," says Adam Hawley, senior product manager for high-end disk arrays in Sun's enterprise servers and storage group. "We want to enable our customers to start small and grow large without throwing anything away. Just snap it together and scale up their performance, scale up their capacity, and ultimately allow them to share that data across their organization and have available all the advantages that come with clusters and network-type architectures."

In short, the enterprise storage parade has shifted direction, and Sun has found itself at the head of the pack. Clearly Sun doesn't intend to let the rest of the parade pass it by.

"I think it's the strategy of leadership," says Robert Gray, IDC research manager for storage systems, of Sun's moves into heterogeneous storage. "I think Sun is setting a direction which, if you look carefully, you'll see the rest of the industry following."

Sun's approach: The Intelligent Storage Network
Sun's take on storage networks is the Intelligent Storage Network. The goal is to combine reliability, availability, performance, and manageability with the virtues of distributed computing and networked storage.

Sun conceptualizes distributed storage as a cloud connected to the LAN. The user or the LAN doesn't have to know exactly what's in the cloud, just that it connects to the storage. The cloud is made up of many different types of storage devices, all tied together by Fibre Channel switches and an intelligent storage server. For example, the cloud could include storage arrays, for workgroups and departments, tape libraries, hierarchical storage management facilities, and optical disks.

Sun is building ISN around several technologies. Some of these are industry standards, and some are specific to Sun.

First is Fibre Channel, the fiber optic networking technology that will probably replace SCSI as the standard storage connection over the next several years. This isn't exactly original with Sun. Almost all the SAN architectures are built around Fibre Channel. Sun, however, has the advantage of experience with high-speed Fibre Channel architectures, including the switched Fibre Channel Fabric.

"Sun is now in their second generation of Fibre Channel products," points out Gray. "They've probably shipped more [Fibre Channel storage] than the rest of the industry together at this point."

Fibre Channel has the bandwidth and distance capabilities to handle distributed storage arrays -- something SCSI generally does not.

Some of the other pieces in Sun's ISN plans definitely are original, however. One of the most significant of these is the DataShare technology that Sun acquired when it bought Encore Computer Corp.'s storage business. The DataShare technology automatically translates information into whatever format is needed, allowing a single copy of the data to serve both Unix and NT systems. Competing systems require full or partial replication of the data or on-the-fly translation. The first needs a lot of extra storage, and the second involves significant performance degradation.

Management is the Achilles heel of modern storage, especially distributed storage. A great deal of Sun's success will turn on the quality of its management tools for storage networks. Sun's INS architecture uses intelligent servers running full (Solaris) operating systems to control the storage functions.

To help manage storage in a heterogeneous environment, INS makes heavy use of Java. Sun's Java-based storage management products use a browser interface to make management functions independent of both processors and operating systems.

To support its plan Sun has announced a new line of storage products under the name StorEdge that emphasizes a modular architecture for scalability and easy change. Sun says that storage processors, disk arrays and controllers, tape libraries, and management software will all snap together and work with future upgrades.

One of the major jobs for Sun is making all this work as neatly in the enterprise as it does on paper. While some of the components of the INS, such as the storage processors and Fibre Channel equipment, are well developed, other components are less so.

"There's definitely an implementation challenge," says IDC's Gray. "Some of this is uncharted territory."

Sun's chances
Having a strategy and executing it are two different things, as the remains of dozens of once-powerful companies illustrate. Can Sun make its new heterogeneous storage strategy work?

Most observers think Sun has a real opportunity. IDC's Vellante, for one, rates Sun's chances in the heterogeneous storage market as good "if they have the stomach for it."

Commitment is the key, Vellante says, and he points out that such commitment isn't automatic -- especially for a computer company. "The problem is that a number of companies have gone after this business and have either not understood what it took or didn't stick with it," he says. Among those who didn't make it, Vellante cites Hewlett-Packard and Digital Equipment Corp., both of whom made strong moves into enterprise storage and then pulled back.

One of the reasons computer companies have had problems in the heterogeneous storage business is the fundamental conflict between selling storage that hooks up to anyone's computer and selling storage to help sell your computer. "A lot of these companies have gotten caught because the objectives of the storage business are not in synchronization with the server side of the business," Vellante says. "Sun doing a partnering deal with Hewlett-Packard [in storage] is not seen as a good thing [by the server business.]"

This is one of the reasons that companies like EMC, which has no server business, are major competitors in heterogeneous storage. Their full corporate attention is focused on servers, not divided between storage and computers.

Channel conflict is another problem for computer companies in the heterogeneous storage market. Selling heterogeneous enterprise storage requires a direct sales force, but the company has to keep from stepping on its VARs and systems integrators. This was a particular problem for DEC, and it may be for Sun as well.

Sun's challenge
Most observers agree that Sun understands what it is getting itself into and what it will take to succeed. Sun executives themselves talk as if they do.

"Our challenge is `how can we make it easier for customers to choose Sun, and how do we make it easier for them to manage their assets?'" says Sun's Hawley. "It's a dynamic environment and in the next decade most data center managers are going to have to be able to deal with three operating systems at a time."

Sun's technology is right for a major move into heterogeneous storage. The question, as always, is in the execution. The next two years or so will test Sun's ability to execute as well as its fundamental commitment to an important new market. Given that, heterogeneous storage could be an important part of Sun's business in the next century.

"If they're committed to it, are willing to make the investment in the channel and technology, and they want to do it for the long term, then yeah, there's gold in them thar hills," Vellante says.

Resources

About the author
Rick Cook divides his time between writing about the Web, computers and high technology, and novels. His most recent stories for SunWorld are "Linux lines up for the enterprise" (January 1998), "IBM bets on Java" (December 1997), and "Sun turns its rays on big iron" (November 1997).

What did you think of this article?
-Very worth reading
-Worth reading
-Not worth reading
 -Too long
-Just right
-Too short
 -Too technical
-Just right
-Not technical enough
 
 
 
    

SunWorld
[Table of Contents]
Sun's Site 		[Search]
Feedback 		[Next story]
Sun's Site

[(c) Copyright Web Publishing Inc., and IDG Communication company]

If you have technical problems with this magazine, contact webmaster@sunworld.com

URL: http://www.sunworld.com/swol-05-1998/swol-05-isn.html
Last modified: