Connecting

There has always been some way to connect systems to storage products. Essentially, this means providing some sort of data transmission system between storage devices or subsystems and the system processor.

For the most part, storage has used either bus connecting technologies for DAS storage or network connecting technologies for SAN storage. The sections that follow briefly examine these two connecting technologies.

A Quick Look at Buses Used for Connecting DAS Storage

Bus technologies used with DAS storage are constructed from a collection of copper wires working in parallel, each transmitting a single bit of information at a time. Data transmitted over a bus is first disassembled from bytes into bits by the sender, transferred over the bus as binary signals (either a 0 or a 1), and then reassembled into bytes again by the receiver. All information is transferred in parallel, including application data, addresses, and communication protocols.

The process of disassembling data, transferring it over a parallel bus, and reassembling it at the receiving end is shown in Figure 2-1.

Figure 2-1. A Byte of Data Transmitted Across a Parallel Bus

NOTE

When we discuss storage data transfers, it seems logical to picture storage data "moving" from one place to another, but of course nothing physical actually travels anywhere. Serial network transmissions work by signal propagation, similar to an analog telephone call. You hear somebody talking, but it is a representation of what he said, not the actual acoustic phenomenon.

Parallel bus transmissions, like those in Figure 2-1, are different. A DAS storing controller creates signals by changing the voltage on the eight data lines. The receiving node measures the voltage across these lines in parallel as an actual byte of data.

The physical characteristics of a bus, such as length, are heavily constrained by the requirement for precise time synchronization in sending and receiving each bit accurately. As a result, buses have inverse relationships between transmission rates and lengththe faster a bus is, the shorter its maximum length can be.

Buses tend to be finite in scope with few degrees of freedomand therefore they have limited failure modes. Data on a DAS storage bus is guaranteed to be received in the same order it was sent. This is called guaranteed in-order delivery. The primary advantage of bus technology is the reliability of the connection. The primary disadvantage of buses is that they are rigid and inflexible. In addition to their physical limitations, they have limited addressing schemes that do not accommodate key concepts like bridging and routing. Bus environments also lack basic management concepts such as domains, naming, and authentication.

The Small Computer Systems Interface (SCSI) bus has been used for two decades to connect storage and other peripherals such as scanners and printers to computer systems. Another common bus has been the Advanced Technology Attachment (ATA) bus, which has been used over the last decade as a way to connect desktop and laptop storage to PC systems.

SCSI bus cables come in two varieties: internal and external. Internal SCSI cables are assembled as ribbon cables, and external SCSI cables are fairly thick and heavy multistrand cables within a common shield. ATA cables are available only as internal ribbon cables. Both SCSI and ATA cables have fairly large connectors containing many pins or connecting sockets.

NOTE

ATA is also often referred to as Integrated Device Electronics (IDE). The terms have been used interchangeably by many, but ATA is more accurate as it refers to the name used in standards documents.

Readers wondering about Serial ATA (SATA) as an exception to the discussion here will find it discussed in Chapter 7, "Device Interconnect Technologies for Storage Networks."

A Quick Look at Networks Used for Connecting SAN Storage

Unlike bus technologies, which have many parallel wires for transmitting binary signals, network technologies typically have four or fewer wires that serially transmit data as frequency modulated signals. Network cables can be made with fiber-optic technology, twisted-pair wiring, and coaxial cable as well as using wireless transmission. In practice, most storage network implementations use fiber-optic transmission technology due to its superior distance and bandwidth capabilities.

Networks also have much richer connection possibilities than buses. For starters, most modern networks have multifaceted topologies that enable highly scalable implementations. In addition, many types of technologies are used to transmit data between different physical networks, such as bridges, gateways, and routers, which increase a network's scalability. Perhaps most importantly, networking technologies usually have important management characteristics such as large address spaces, naming abstractions, and sophisticated measurement and alerting mechanisms.

For all their advantages, the fact that networks have so much more functional breadth than buses also means that they have many more variables and, therefore, potential problems. This is why products with advanced management functions are so important in storage networks.

As mentioned at the beginning of this chapter, one of the difficulties newcomers have in understanding storage networking is related to the assumptions they make based on their experience with TCP/IP and Ethernet data networks. In general, the design of data networking products assumes there might be delivery problems in the network and that data could be transmitted over different network paths. The bottom line is that in-order delivery of data is not guaranteed, as is the case with bus connectivity. The requirement for in-order delivery turns out to be extremely important for storage applications and is discussed more in Chapter 3, "Getting Down with Storage I/O," and Chapter 10, "Redundancy Over Distance with Remote Copy."

Fibre Channel networking technology has been the most common connecting technology used for SANs. The most recent connecting technology to be introduced is iSCSI, which is implemented on top of TCP/IP, meaning it can run on any physical network that supports TCP/IP. TCP/IP Ethernet networks are by far the most common for connecting systems to NAS storage appliances.

The Logical Side of Storage Connections

The connecting component of a storage network has logical aspects as well as physical. These logical processes incorporate such things as access algorithms, including fairness and congestion control, addressing and naming, network configuration and domain management, network operations management and other topics such as link aggregation and virtual networking. In essence, any technology used in a network for storage networking is part of the connecting component in a storage network.

This is a very rich topic with many interesting nuances. For the most part, this book focuses on storing and filing technologies in storage networks and does not address connectivity topics. Readers interested in exploring the connecting components of storage networks should seek other books that focus on such topics as storage protocols (FCP, iSCSI, and FC/IP), Fibre Channel, Ethernet, TCP/IP, and network switching technologies.