By Dave Simpson
Although alternative disk drive interfaces such as SATA and SAS, and alternative SAN protocols such as iSCSI, tend to get more ink these days, Fibre Channel is holding its own in the disk drive market and still rules the SAN space.
For example, Fibre Channel is expected to account for almost one-fourth (24%) of all enterprise-level disk drive shipments this year (compared to 47% for SCSI, 23% for SATA, 5% for SAS, and 1% for Parallel ATA). And Fibre Channel still accounts for almost 100% of the SAN market, with iSCSI barely putting a dent in Fibre Channel’s dominance.
Nevertheless, the Fibre Channel community has formidable competition, and the vendors and standards bodies in charge of Fibre Channel advances aren’t resting on their laurels.
Perhaps the most significant development in the Fibre Channel market is the latest throughput rate-4Gbps, which is twice the speed of the previous standard. Even though most Fibre Channel users don’t really need the speed, many are upgrading anyway if only because 4Gbps Fibre Channel products cost about the same as 2Gbps products. Early deployments of 4Gbps Fibre Channel have been primarily in applications such as video, imaging, backup and recovery, and campus-area replication.
As with previous generations, 4Gbps Fibre Channel is backward-compatible.
In a QuickVote poll of InfoStor’s readers, 32% have already upgraded some of their Fibre Channel SAN infrastructure to 4Gbps, and 4% have upgraded their entire SAN infrastructure (see figure). Another 28% plan to migrate to 4Gbps later this year, while less than one-fourth (23%) have no plans to upgrade. (The remaining 13% of the survey respondents do not have a Fibre Channel SAN.)
There are a number of other Fibre Channel standards, most of which are still in the development stage, that could provide benefits to certain types of users. Those standards include FC-SP, FC-SCM, FC-BaseT, FC-SATA, FC-IFR, NPIV, and FAIS.
The Fibre Channel Security Protocol (FC-SP) provides authentication mechanisms to ensure that only approved devices can log into a SAN fabric. The standard was completed earlier this summer, and a number of SAN components have already implemented part of the FC-SP specification, such as authentication via the DH-CHAP protocol. (DH-CHAP is a standard, based on a secure key-exchange authentication protocol, for the authentication of devices connecting to a Fibre Channel switch. DH-CHAP supports both switch-to-switch and host-to-switch authentication.)
In addition to authentication, FC-SP provides for encryption (that can scale beyond 10Gbps) and fabric policies.
“FC-SP addresses the fact that security becomes much more important as SANs grow and are connected over the Internet and WANs,” says Claudio DeSanti, a technical leader in Cisco’s storage technology group and vice chairman of the ANSI T11 committee, which governs Fibre Channel standards.
The Fibre Channel Simplified Configurations and Management (FC-SCM) standard will be of particular interest to small to medium-sized businesses (SMBs) and departments of larger enterprises because it is designed to simplify implementation of SANs in small environments.
“FC-SCM will allow personnel untrained in storage to connect and configure a Fibre Channel SAN…where either performance or scalability are or will be needed,” says Thomas Hammond-Doel, chairman of the Fibre Channel Industry Association (www.fcia.org) and director of technical marketing at Emulex. Hammond-Doel adds that the FC-SCM will include a definition for a bridge to link FC-SCM-based SANs to larger enterprise SANs.
Technically speaking, FC-SCM is a profile, as opposed to a standard. The profile is a simplified feature set that defines, for example, default setting for certified devices.
FC-SCM is currently in the FCIA requirements phase, and compliant products are not expected for another 18 months or so.
The FC-BaseT standard will enable Fibre Channel over twisted-pair Ethernet copper cabling (at up to 100 meters), which will reduce cabling costs. Like FC-SCM, FC-BaseT is designed to move Fibre Channel into smaller environments. Although the technical specifications are complete, products that adhere to the standard aren’t expected for another 12 to 18 months.
The FC-SATA specification provides for Serial ATA (SATA) tunneling over Fibre Channel, which in turn, will enable tiered storage via the ability to combine high-performance Fibre Channel drives and infrastructure with low-cost, high-capacity SATA drives without the need for bridges. In a sense, FC-SATA is similar to Serial Attached SCSI (SAS), which supports both SAS and SATA drives.
The FC-SATA specification was spearheaded by Emulex, but a number of other vendors are currently involved with the spec, including Agilent, Brocade, Finisar, Fujitsu, Hitachi, McData, Seagate, SiliconStor, Western Digital, and Xyratex. Compliant products are expected in 12 to 18 months, according to Hammond-Doel. (For more information on FC-SATA, see “Emulex pushes FC-SATA spec,” InfoStor, January 2006, p. 15.)
Fibre Channel Inter-Fabric Routing (IFR) has already been partially implemented by several switch vendors, and the technical specifications are completed, although fully compliant products are not expected within the next year. The FC-IFR standard provides for heterogeneous fabric routing (including physical and virtual fabrics) and improves scalability and interoperability.
Products based on N_Port ID Virtualization (NPIV) have already been proven in the mainframe world in the context of FICON. NPIV essentially makes the port ID autonomous from the server and improves the sharing of physical host bus adapters (HBAs) in server virtualization environments. Essentially, NPIV allows a single Fibre Channel port to appear as multiple, distinct ports providing separate port identification and security zoning within the fabric for each operating system image as if each image had its own unique physical port. For example, NPIV allows a Fibre Channel HBA to perform multiple “logins” to a fabric using a single physical port. NPIV makes it easier for administrators to manage virtual servers on a Fibre Channel SAN.
NPIV is currently supported at the switch level, with fully compliant HBAs due within the next year.
The Fabric Application Interface Specification (FAIS) standardizes the interface to the fabric (and improves interoperability) for fabric-based applications. FAIS defines an API (not a protocol) between control path and data path functions in storage management applications and intelligent SAN platforms. Splitting control-path functions from data-path functions allows management applications to delegate processing of all data-path functions to the SAN platforms (such as intelligent switches) while maintaining control functions. FAIS will make it much easier for software developers to port their applications to a variety of different SAN platforms.
FAIS will be implemented by the switch vendors (and already has been in varying degrees) and vendors of other types of fabric-based switching/routing devices. Fully compliant switches are expected in about 18 months, according to FCIA officials.
For detailed information of all the standards mentioned in this article, visit www.t11.org.
A quick review of Fibre Channel management
By James Long
The ANSI T11 Fibre Channel Generic Services (FC-GS) specification defines several services that augment the functionality of FC SANs. Among these is an in-band management service. The FC Management Service is actually composed of many supporting services and includes
■ A Fabric Configuration Server;
■ Performance Server;
■ Fabric Zone Server;
■ Security Policy Server;
■ Unzoned Name Server; and
■ Host Bus Adapter (HBA) Management Server.
The FC Management Service is distributed and typically runs on switches in the fabric. Each switch maintains an identical copy of the management information for the entire fabric. This enables a single point of access for management of a SAN. In other words, a management application (such as a fabric manager) can connect to a single switch to access all management information for all switches in the fabric.
The Fabric Configuration Server supports discovery of the fabric topology and operating characteristics. When a fabric manager connects to an FC switch, connectivity information for all devices in the fabric is visualized using data extracted from the Fabric Configuration Server.
The Performance Server collects and aggregates performance statistics for the entire fabric. Management applications can query any FC switch to access performance statistics for the entire fabric. The Performance Server also supports threshold monitoring.
The Fabric Zone Server restricts communication between devices attached to a SAN. Although FC zones are security-focused, the Fabric Zone Server is integral to the management service. This is because the Fabric Zone Server is the mechanism by which FC zones are configured and managed.
The Security Policy Server facilitates the definition and management of security policies related to the Fibre Channel Security Protocols (FC-SP) specification.
The Unzoned Name Server provides full access to the FC Name Server (FCNS) database for management applications. Initiators and targets are subject to zone policies that restrict access to the information in the FCNS database. However, management applications require unrestricted access to the FCNS database.
The Host Bus Adapter Management Server enables HBAs to register certain data (called HBA management information) with the FC fabric. This enables management applications to query the fabric for HBA management information instead of querying the end nodes directly. The mechanisms for HBA management information registration and retrieval are defined by the Fabric Device Management Interface (FDMI).
A standard Application Programming Interface (API) for HBAs complements the FC Management Service. The HBA API is defined in the ANSI T11 SM-HBA specification. It supports direct management of HBAs. Vendor-supplied agents implement proprietary APIs. A standard API enables management application vendors to write software without the overhead of HBA-specific API calls. Thus, the HBA API enables management applications to discover, monitor, and manage heterogeneous HBAs consistently across disparate operating systems and server platforms. Note that the same information can be gathered from the FC fabric via the Performance Server, the HBA Management Server, and the FCNS.
The FC-GS specification series also defines a notification service called the Event Service. Fabric-attached devices can register with the Event Server to receive notification of events. Currently, a limited number of events are supported. Like the FC Management Service, the Event Service is distributed.
The FC-LS specification defines various Extended Link Service (ELS) commands that can be issued by a fabric device to ascertain state information about devices, connections, Exchanges, and sequences. Examples include Read Exchange Status Block (RES), Read Sequence Status Block (RSS), Read Connection Status (RCS), Read Link Error Status Block (RLS), and Read Port Status Block (RPS). These commands are issued directly from one device to another.