Storage switches continue to increase in both speed and features. The biggest change in the market, however, is that the storage switch vendors are increasingly adopting iSCSI and other protocols to supplement their Fibre Channel (FC) offerings.
“The original proposals were to put SCSI on Ethernet, but Ethernet was lacking certain capabilities that SCSI required,” said Robert Passmore, an analyst at Gartner in Stamford, Conn. “The SCSI protocol expected commands to be delivered, while the Ethernet protocol allows things to be lost at will.”
SCSI also expected reasonable flow control, so if a lot of SCSI nodes were asking for something, they were prioritized. Ethernet lacked these types of controls and, since the network vendors were unwilling to add them to the Ethernet protocol, this led to the creation of the FC protocol and the first FC SANs in 1997.
“Ethernet traditionally focused on lower cost, mass adoption, interoperability and economies of scale with performance based on bandwidth,” said Greg Schulz, founder and senior analyst of The StorageIO Group in Stillwater, Minn. “FC focused on low latency deterministic or predictable performance for storage specific applications.”
Work continued, however, on running SCSI over Ethernet, culminating in the iSCSI protocol in 1994 and the first iSCSI products shipping the following year.
iSCSI enabled companies to create SANs using standard Ethernet switches and low-cost network interface cards rather than host bus adapters. Where initial purchase cost is the most important consideration, iSCSI is the best option; however, higher performance applications require FC.
A third protocol now on the market is FC over Ethernet (FCoE). With organizations increasingly moving to replicated storage systems for disaster recovery, FCoE overcomes the distance limits of FC without the heavy overhead of TCP/IP.
According to Passmore, Brocade Communications Systems and Cisco Systems, both of San Jose, Calif., dominate the storage switch market. Together, they account for more than 90 percent of sales. QLogic Corp of Aliso Viejo, Calif., with 1 percent to 2 percent of the market, sells embedded switch technology as well as its own FC switches.
Brocade calls its latest storage strategy the Data Center Fabric (DCF). Announced at its user meeting on October 23, the DCF is an application-oriented architecture that supports multiprotocol connectivity and policy-based automation across the data center. Services include continuous data protection and disaster recovery; file and block data migration across heterogeneous environments; server and storage virtualization; and encryption for data in flight and at rest.
The company plans to release a series of products by the end of 1Q08 to implement these strategies. In mid-October, Brocade upgraded its 48000 director-class switch from 4Gb to 8Gb. Coming up is a director-class switch, called the DCX, which supports Ethernet, FC, FCoE and iSCSI, and includes software for management and data encryption.
If Brocade is rolling out a strategy, you can count on Cisco to release an alternative strategy. In this case, it’s Data Center 3.0.
“Cisco provides a holistic approach to data storage solutions and to the overall data center, with a broad range of products and solutions that have been developed and tested together to optimize data center resources,” said Deepak Munjal, Cisco’s marketing manager for data center solutions. “The Cisco vision for Data Center 3.0 entails the real-time, dynamic orchestration of infrastructure services from shared pools.”
Cisco’s portfolio includes both Ethernet and FC switches. New products this past year included the MDS 9134 Fabric Switch and the MDS 9222i Multiservice Modular Switch. The MDS 9134 is a 4Gb switch designed for use in small SANs or on the edge of a large SAN. The MDS 9222i, an 18-port, three-rack unit box, provides multiprotocol connectivity in open and mainframe environments. It supports encryption across heterogeneous tape libraries, virtual tape libraries and disk arrays; has redundant FC over IP topologies; and provides both control path and data path isolation between virtual SANs running on the same physical fabric.
“With a compact form factor, modularity and advanced capabilities normally available only on director-class switches, the MDS 9222i is an ideal solution for departmental and remote branch-office SANs,” said Munjal.
QLogic has been making HBAs for about a decade and providing chips to other switch vendors.
“Only in the last year have they decided to take their own switches into the market,” said Passmore. “These are mostly still being tested in the labs of the storage vendors, but a couple of these have come through.”
In August, QLogic became the first company to offer 8Gb FC switches and HBAs to storage vendors for compatibility testing. It provides technology for HP’s Blade System c3000 enclosure, and it also has its own SANbox 9000 director class switch and SANbox 5600Q stackable switch.
Making a Strategic Switch
No matter how attractive the features of a particular switch, it is not an isolated purchase, and it must be considered in terms of the overall storage, network and data center infrastructure. The switches should therefore be purchased from a SAN vendor that has certified a product works with the organization’s software and architecture.
“HP’s focus is in delivering complete end-to-end solutions that solve customer problems,” said Edgardo A. Lopez, product marketing manager for HP StorageWorks in Marlborough, Mass. “Switches, directors and multiprotocol gateways are important components in this larger solution, but the real value for the customer is in a complete, reliable infrastructure solution that delivers on the promised value.”
This year HP introduced a new set of embedded 4GB FC switches for its Blade System c-Class using features from Brocade and Cisco.
“Both the Brocade Access Gateway Mode and Cisco N-Port Virtualizer extend the capabilities of the embedded switches with N-Port Virtualization support, thus enabling them to act as HBA aggregators rather than traditional switches,” said Lopez. “In this mode, they become I/O devices that continue to provide value through reduced management (set and forget), greater scalability (they do not count as a domain in the fabric), lower cost through reduction of cables and SFPs, and greater interoperability options, since they do not connect as an E-port to the fabric.”
This article was first published on ServerWatch.com.