1.3 Text Overview
The first edition of this work attempted to provide a comprehensive overview of SAN technology for both technical and nontechnical readers. Because SANs were relatively new arrivals on the storage scene, it was possible to accomplish this goal in a fairly short work by using a dense writing style that one reader has called "information compression." The current work necessarily includes additional discussion of new SAN technologies that have emerged over the past three years, but it attempts to preserve an information compression ratio that delivers more knowledge in fewer pages. Because few people in today's information technology space have the time to read a work cover to cover, it is hoped that this edition of Designing SANs will be a useful reference for readers who need ready access to specific information on various SAN topics.
The chapter progression begins with lower-level technical overviews of storage and networking primarily related to the underlying SAN plumbing that supports higher-level applications. This discussion includes both abstract standards and the products developed from them. SAN infrastructure components that were previously based on Fibre Channel alone now include IP-based solutions, so additional space in the current work is devoted to heterogeneous SAN design options. Following the infrastructure discussion, middle-layer SAN issues, such as storage-specific applications, management, and virtualization, are reviewed. With these additional concepts in place, it is then possible to discuss concrete implementations of SANs in customer environments. The text concludes with a general overview of outstanding issues yet to be resolved in the areas of standardization, interoperability, and management, along with speculation on the future direction of storage area networking as the adoption of SANs expands throughout the market.
Chapter 2 provides an overview of both storage and networking concepts and explains how the fusion of these technologies has created new means to solve data storage issues. As in any dialectical synthesis, the resulting product has new attributes quite distinct from its constituent parts. Storage has thus placed new demands on networking, and networking has fundamentally altered storage relationships. As a fusion technology, storage networking requires a basic understanding of networking and storagea need addressed by this chapteras well as new concepts addressed by later chapters.
Storage area networking has deep roots in Fibre Channel technology. Fibre Channel was the first serial transport to successfully solve gigabit transport, data encoding, optics, transport protocol, and topology issues required for high-performance shared storage applications. Chapter 3, Fibre Channel Internals, reviews the lower-level operation of the Fibre Channel physical and protocol layers, including framing, class of service, flow control, and naming and addressing conventions.
Chapter 4 provides the foundation for understanding Fibre Channel topologies that have been employed for SANs. Today, most Fibre Channel SANs are based on switched fabrics. Chapter 4 therefore focuses on fabrics and fabric switch features and devotes less space to point-to-point and arbitrated loop configurations. Arbitrated loop, however, is still commonly used for JBODs ("just a bunch of disks") and as back-end storage for some NAS products, and so this chapter also discusses loop principles.
Chapter 5 examines the Fibre Channel products that have been created to accommodate server and storage attachment to a SAN. This spectrum of products includes transceivers and cabling, host bus adapters (HBAs), Fibre Channel-attached storage and tape, loop hubs, and fabric switches. Because extension of native Fibre Channel over wide areas using tunneling or dark fiber is dependent on the same Fibre Channel behavior demonstrated in a local environment, native Fibre Channel extension and IP tunneling products are also grouped into this chapter.
Over the past two years, storage over IP has entered the SAN market to complement or sometimes compete with Fibre Channel solutions. Chapter 6 provides a brief overview of IP SAN technology. Native IP storage protocols include Internet Fibre Channel Protocol (iFCP) and Internet SCSI (iSCSI). As with Fibre Channel, IP storage must provide discovery and naming mechanisms, and these are reviewed in the sections on Service Locator Protocol (SLP) and Internet Storage Name Service (iSNS). Because IP SANs directly converge mainstream IP networking with storage, additional IP services, such as Quality of Service (QoS) and IP Security that can be applied to storage, are discussed. IP is also an enabling technology for metro and wide area SANs, and Chapter 6 concludes with a review of buffering and speed-of-light latency issues that a storage architect must consider when designing SANs over distance.
Chapter 7 examines various classes of IP SAN products that have been or will be introduced into the market. Much of an IP SAN infrastructure can be borrowed from mainstream IP networking. Gigabit Ethernet switches and IP routers, for example, are not specific to SANs but can be incorporated into SAN design without modification. Similarly, TCP offload engines (TOEs) have been developed to solve host processing issues historically associated with TCP but also have become enablers for high-performance IP SANs. IP storage gateways can perform Fibre Channel-to-iSCSI or Fibre Channel-to-iFCP conversion, or both. As in the discussion of Fibre Channel products, the focus here is on product functionality and not specific vendor implementation. Although it is tempting to review and compare individual vendor products, the market is moving too quickly for a book to capture capabilities of specific vendor products.
After establishing a foundation for understanding SAN plumbing options in the preceding chapters, Chapter 8 begins a higher-level discussion of SAN software applications. This chapter reviews storage-specific applications, including server clustering, backup, data replication, distributed file systems, and file sharing. Some of these solutions are software only; others can be embedded in storage hardware products.
Chapter 9 examines problem isolation in SANs. Problems can and do occur eventually and can occur at any layer in a SAN configuration. This chapter looks at some simple problem isolation techniques as well as hardware and software tools that facilitate identifying and isolating faults in a SAN.
Management of SANs is a broad topic because it includes management of the SAN transport as well as data placement and resource management. Chapter 10 discusses the major categories of SAN management and management protocols that can be used. This chapter also includes an overview of the Common Information Model (CIM) initiative. Although CIM has been under construction for some time, it has recently received strong industry support in the hopes that it will enable a coherent and unified solution for managing both data transport and data placement from a single platform.
Chapter 11 takes the reader into the Twilight Zone technology of storage virtualization. Like a quantum particle in a transient state, storage virtualization can both exist and not exist in the same moment. Elementary storage virtualization such as RAID certainly exists and has provided significant value for storage applications for some time. Advanced, application-aware storage virtualization has yet to appear and will require significant engineering effort to bring into being. The vendor and marketing hype aside, storage virtualization holds great promise for SANs and will enable customers to implement shared storage solutions throughout their networks.
Chapter 12 covers application studies of SAN solutions. From storage-intensive operations, such as full-motion video editing and prepress graphic creation, to more ubiquitous customer requirements for streamlined tape backup, SANs have successfully addressed a wide range of business needs. In discussing storage consolidation, server clustering, content distribution, disaster recovery, and other applications, the material in Chapter 12 attempts to offer examples that accommodate Fibre Channel, storage over IP, or both in a heterogeneous solution. Where possible, the SAN infrastructure is presented in a technology-neutral way so that the SAN architect can thereafter overlay the customer transport of choice.
Chapter 13 reviews some of the main areas of concern for SANs, including standardization, interoperability, and management. These are mutually dependent and overlapping issues that vendors and customers together must address. Standardization of a technology, for example, has little value for customers if it does not result in interoperability of vendor products. Similarly, even the most sophisticated management platform would be ineffectual if its cycles were consumed reporting errors due to standards violations or interoperability conflicts.
Chapter 14 is my opportunity to engage in wild speculation about the future of SAN technology. No one has a crystal ball in such a fast-paced technology, and the storage industry itself is being shaped by the push of emerging technologies and the pull of customer demands. Although InfiniBand may help transform the high-performance server market, storage virtualization and IP storage networking are now the main forces driving SANs to ubiquitous adoption.
This edition of Designing SANs also includes supplementary material in the appendixes to provide background and reference information on storage networking. In addition to resource and vendor listings, material is provided on the standardization process behind SAN development, including the work of NCITS/ANSI T10 and T11 groups and the Internet Engineering Task Force (IETF). The structure and activity of the SNIA are also included, as is the work of the SNIA Technical Council on the Shared Storage Model and the SNIA Education Committee's SNIA Dictionary. Some of my previously published articles on storage networking are grouped in Appendix G, SAN Essays.