The intelligent switch, which is currently creating a lot of buzz in the field of storage networking, dates back to the 1980s in the field of telephony. It made its first appearance in voice networks as a means of automating certain aspects of network management, such as provisioning and call handling, to lower administrative costs.
Similarly, when data networks became ubiquitous in the mid-1990s, the intelligent switch once again appeared. In this incarnation, the main purpose of the intelligent switch was to reduce the complexity of network management by automating certain functions. But, the intelligence in the switches was, initially, rudimentary at best. However, over time, the intelligent switch evolved into the super smart routing switches that are available today from vendors such as Extreme Networks, Foundry Networks, and Riverstone Networks.
As storage networks become more complex, the concept of the intelligent switch is again being pressed into service. And, as was the case with local area networks (LANs), vendors are touting automation of management functions as the answer to the problem of the escalating costs of storage area network (SAN) management.
In its reincarnation, the intelligent switch is taking the form of a sophisticated routing switch, which closely parallels many of the LAN routing switches. A recent study authored by Tom Curlin, a principal analyst with the Royal Bank of Canada Capital Markets defines these new routing switches as "devices that combine multiprotocol switching architecture with port-level I/O routing intelligence."
Vendors are introducing intelligent switches to permit heterogeneous SANs - SANs comprised of storage arrays from different vendors - to be managed easily and affordably. While vendors are initially providing embedded virtualization, multiprotocol capability and port-level I/O intelligence, the ultimate goal is to automate SAN storage management through rules-based policies.
Blazing the trail
Thus far, only one SAN routing switch has appeared in the marketplace - the PSX-1000, which was introduced in March 2002 by Pirus Networks of Acton, MA. The PSX-1000 comprises a six-slot carrier-class chassis with redundant power supplies, fans and optional modules. It consists of a networking component, a system component and a storage component layer.
The networking component layer module, which Pirus calls a LAN Resource Card, houses four network processors, providing TCP acceleration, Ethernet switching and security in the LAN environment. In addition, this layer contains eight Gigabit Ethernet ports, which allow the Pirus switch to attach to eight IP-based LANs or SANs.
The system component layer consists of two modules: the 40GB crossbar switch and a Management Interface Card. The crossbar switch, along with associated software, provides the multiprotocol capability in the PSX-1000. The Management Interface Card contains a dedicated microprocessor for system management, and an out-of-band management port.
The storage component layer contains a Storage Resource Card and a Storage I/O Card. The Storage Resource Card consists of four processors that provide file services, iSCSI protocol support and security functions in the SAN domain and 1GB of RAM for data caching. The Storage I/O Card contains four Fibre Channel ports and four gigabit interface converters (GBICs) that support those ports.
All of the hardware functionality is controlled by the PirusView software suite of management software that mainly resides in the Management Information Card (some is distributed to other modules in the system). Specifically, the PirusView software suite supports virtualized block and file services, logical file servers, file system point-in-time copy and browser-based management. The block and file services permit administrators to establish a common pool of storage that can be accessed from the same platform by network-attached storage (NAS) and SAN applications and users. The virtualized file services let users establish up to 12 logical servers, allowing them to consolidate servers in their storage environment. The block services supports storage pooling so that storage administrators can allocate storage logically in a heterogeneous environment.
Perhaps the PSX's most interesting feature is the Secure Virtual Storage Domain (SVSD). The SVSD feature permits administrators to partition the storage infrastructure into multiple domains, much in the same way as administrators in the LAN world can set up virtual LANs (VLANs). Using this feature, storage administrators can set up storage capacity and storage services to match the needs of particular applications and departments within the enterprise. For example, an administrator could up the storage capacity for a high priority application. Similarly, the administrator could limit access to storage capacity and services to some departments, while granting access to others, according to predefined policies.
Other switch vendors have entered the fray as well. Brocade Communication Systems Inc. says it will build a second-generation intelligent storage switch, using its SilkWorm 12000 switch as the foundation. In this next-generation switch, Brocade plans on providing switched-based virtualization by adding what it calls "wirespeed mapping capabilities," consisting of ASICs and network processors to the SilkWorm 12000 to accelerate the mapping portion of third-party virtualization software. Additionally, the company plans to add improved security and ASIC-based quality of service (QoS).
In the security area, Brocade will build on its SecureFabric OS, according to Jay Kidd, vice president of marketing. Brocade's QoS initiative gives users the ability to monitor use of the SAN and enforce management policies that prioritize applications. Brocade will also introduce multiprotocol capability in its intelligent switch.
McData, like Brocade, has been highly successful as a vendor of Fibre Channel switches that have been used in the SAN marketplace. According to Brandon Hoff, McData's senior director of product management, the heart of its intelligent switching strategy revolves around fabric virtualization, which Hoff defines as "application-centric storage infrastructure management through self-managing and self-provisioning systems."
Hoff says fabric virtualization will be delivered in three phases: visualization, virtualization and automation. The company recently delivered part of the visualization intelligence in its SAN Navigator management tool. This intelligence consists of discovery capability and policy management of capacity planning tasks. In the virtualization phase, it will provide third-party virtualization to make use of intelligent components of the switch, such as network processors and ASICs. The final phase - automation - will take the form of automated storage management such as automated provisioning of capacity (based on QoS), automated backup and automated hierarchical storage management - all based on a policy management engine. Multiprotocol capability - iSCSI, InfiniBand and so forth - will be added to the McData switches as they come into the mainstream.
One of the major beneficiaries of the SAN intelligent switch may well be Veritas, the largest ISV in the storage management software field. Veritas equates the "intelligence" in these new devices with software - not hardware - according to Rob Soderbery, senior director of strategic relations. "We believe that the intelligence is going to remain in the software. Our strategy is to enable partners ... to take our storage software and embed it in their new architectures," he says. So far, at least six hardware vendors will embed Veritas software in their hardware devices.
This move is quite significant - it frees users from vendor lock-in when those users want to purchase certain software management applications. In the past, some of the best-of-breed storage management software has been proprietary, running only on certain vendors' hardware. If a user wanted to run one of these software packages, they would also have to buy the same vendor's hardware - becoming locked into that vendor. Specifically, Veritas will provide a special software package consisting of its volume manager, file system, and NAS and SAN appliance software. The switch vendors, including Pirus Networks, Rhapsody Networks, and Troika Networks, will produce a variety of solutions, according to Soderbery. He says these products will offer virtualization capabilities and the ability to handle both file and block-storage traffic.
Dave Stevens, the vice president of business development for Rhapsody Networks, one of the participants in the Veritas-powered program, is much more involved than just being an OEM for Veritas software. "The Veritas-powered program is taking Veritas source code, and modifying and embedding it into a hardware platform," says Stevens. He adds that his company will produce unique, high performance products for the SAN infrastructure with the embedded Veritas code.
Before being accepted in the SAN marketplace, the intelligent storage switch, no matter what form it ultimately takes, will have to pass muster with the storage array vendors. Practically speaking, this acceptance must come from EMC and Hitachi Data Systems. According to James Rothnie, senior vice president and chief technical officer at EMC, the intelligent storage network, particularly one that features multiprotocol capability, has a definite place in EMC's vision of the future SAN.
Presently, EMC has a lot of intelligence in the storage array controller, but the advent of the intelligent storage switch may mean that the company will put some of the intelligence currently residing in the array controller on the intelligent switch, says Rothnie. "It's possible that some aspects of what we do now in our management area will find a home in these switches; for example, there are functions involved in redirecting I/O in some circumstances that might be effectively done in the [intelligent] switch." However, Rothnie is quick to add that "most of what we do in storage management is not best done in the switch; it's best done either on the server or on the storage system itself."
Rothnie also says that SAN switch vendors can learn a lot from IP networking vendors. In particular, he contends that the history of IP networking proves that "the most successful products have been those that have focused on very high-speed capability because that [capability] allows the switches to maximize the ports that they can handle." Rothnie maintains that taking this same approach with SAN intelligent switches will "minimize costs for users, while delivering good performance."
Like EMC, Hitachi Data Systems is a powerful force in the storage networking market. It has come on strong in the last year through OEM agreements with Sun Microsystems Inc. and Hewlett-Packard and through the introduction of new storage array technology. Also, like EMC, Hitachi has been providing intelligence in its storage array controller for some time. But, according to Hu Yoshida, vice president and chief technology officer for Hitachi, the company welcomes the intelligent storage switch to the world of storage networking.
Yoshida sees one major advantage to intelligent storage. "You have to move up higher in the management chain if you want to manage heterogeneous storage, such as being able to copy between different types of storage arrays. That's where I see the main value of the intelligent switch," he says.
The storage intelligent switch will have a significant impact on storage networking. Combined with virtualization, it seems that it will finally bring about the ability to manage storage in heterogeneous environments. Embedded intelligence will give users an opportunity to avoid vendor lock-in. And the apparent buy-in to the concept by some of the influential storage array vendors will hasten their acceptance of the intelligent switch.
Furthermore, over the long haul, these switches will probably lower storage management costs by automating many of the storage management tasks within SAN and NAS environments. But, keep in mind that the intelligent storage switch is still in the developmental stage and a tremendous amount of work needs to be done before the vision becomes reality. Most of the announcements currently being made pertain to vendors' visions of what they think the network will look like in two or three years. It will be 2004 or 2005 before vendors will be able to deliver substantive implementations of features such as QoS, policy management of block and file data and advanced security. In the short term, though, users can look forward to switch-based virtualization, more applications executing on the switch, and improved security.