In an effort to improve business productivity, corporations are implementing web and Internet applications such as customer relationship management (CRM), enterprise resource planning (ERP), and e-mail. This move has resulted in the accumulation of large amounts of corporate data, and these voluminous stores of data are critical to a company’s operation.
Large amounts of data require large amounts of storage. Storage area networks (SANs) have emerged as the premiere technology for advanced storage requirements. SANs address an IT manager’s needs by providing a scalable, manageable, and efficient deployment of mission-critical data.
Aside from offering advanced technology for storage needs, SANs also reduce costs when deploying highly scalable storage versus traditional direct-attached storage (DAS). Hence, SAN reduces cost and offers new levels of performance and scalability previously unavailable. It’s a popular technology that is still being defined and explored.
SAN addresses the following issues currently experienced with DAS:
Difficulty of managing large, distributed islands of storage from multiple locations
Complexity of scheduled backups for multiple systems
Difficulty of preparing for unscheduled outages
Inability to share storage among multiple systems
Sheer expense of distributed disk farms
SAN addresses these issues by doing the following:
Reducing management costs through centralized control for monitoring, backup, replication, and provisioning.
Reducing subsystem costs through any-to-any connectivity between storage and servers. This setup allows networks to match servers with underutilized storage subsystems.
Reducing backup costs through the centralization and consolidation of backup functionality.
Offering highly available disk services by providing redundant, multiple paths between servers and storage devices. This provision allows for automated failover across all storage in an easily scalable manner.
Offering highly scalable and location-independent disaster recovery. You can replicate entire data centers to multiple locations, allowing quick and efficient switchover if the primary data center becomes unavailable. DAS networks are unable to provide this level of disaster recovery.
Traditional JBOD (just a bunch of disks) DAS networks are file-system– and platform-dependant. The disks are associated with a single set of servers, and only the attached host can access them. Examples include small computer systems interface (SCSI), fiber channel, and enterprise system connection (ESCON). Redundant Array of Inexpensive Disks (RAID) addresses fault tolerance.
Because SAN removes the device, operating-system, and location dependencies of traditional DAS, new capabilities emerge. Storage expansion no longer has an impact on servers and vice versa. Bandwidth is available on demand, and load-balancing can occur across multiple active paths.
The two access methods available for SANs are fiber channel and Internet Protocol (IP).
Dedicated fiber-channel networks attach servers to storage devices. The fiber-channel network passes data blocks around (blocks are used by disk access), which the servers access through host-bus adapters. The servers then attach to LANs to provide the information to the rest of the network. SAN components include host-bus adapters, storage systems (RAID, JBOD, tape, and optical disk), hubs, switches, and SAN-management software.
Fiber channel is an American National Standards Institute (ANSI) standard that combines both channel and network technologies. SCSI and IP are the primary upper-layer protocols available on fiber channel. Fiber channel operates at gigabit speeds.
You can use several topologies for fiber-channel deployment:
Point to point—. Dedicated connections between a server and storage device. This method is suitable when storage devices are dedicated to a single file server.
Arbitrated loop—. Storage and file servers connect to each other in a closed loop with up to 126 nodes. Maximum bandwidth is 100 MB shared between all the nodes. Therefore, the number of attached nodes directly affects the performance of the loop.
Switch fabric—. As with Ethernet, performance and reliability vastly improve when moving from shared media (such as an arbitrated loop) to switched fabrics. A fiber-channel switch offers up to 100 MB to each switched port. You can trunk switches together to attach up to a theoretical 16 million nodes.
The other access method to fiber channel is IP.
Because SANs connect to IP networks, the storage can be accessed across LANs, WANs, metropolitan-area networks (MANs), and the Internet. You can manage the devices using Simple Network Management Protocol (SNMP) and operate in a secure mode with the use of Internet Protocol Security (IPSec), firewalls, and virtual local-area networks (VLANs). Storage devices can connect to Ethernet devices.
Whereas fiber channel separates disk-block–based traffic from the rest of the network, IP-based SAN tunnels block traffic through the IP network using the Internet Protocol SCSI (ISCSI) protocol. ISCSI uses Transmission Control Protocol (TCP) as its transport.
Fiber channel over IP (FCIP) is another technology that allows fiber-channel networks to connect to each other over LANs, MANs, and WANs. FCIP uses TCP to tunnel fiber-channel block traffic across the IP network.
At-A-Glance—Storage Area Networks
Most companies use online storage to house business-related information. As more information from business functions—such as sales, inventory, payroll, engineering, marketing, and human resources—is stored online, the efficient acceptance, management, and retrieval of this information becomes more critical to the success of the business. By storing data in common devices, companies can also leverage economies of scale, reducing the total cost of data storage.
</division><division> <title>What Are the Problems to Solve?</title>For many years, information was stored in server-centric architectures. As businesses grew and encountered greater storage needs, they added servers when and where they needed them.
However, this direct-attached storage (DAS) approach is limited in its ability to scale because it is expensive to manage and it uses resources inefficiently. Each server that you add (in each department, for example) is associated with a specific CPU. Additionally, standard activities such as backing up data or adding capacity are difficult and inefficient.
</division><division> <title>What Is Storage Networking?</title>
Overcoming the drawbacks of server-centric storage is a network-centric model called storage networking. Storage networking is the software and hardware that enable you to consolidate, share, access, replicate, and manage storage over a shared network infrastructure.
To understand how storage networking works, you must understand the different storage methods and technologies.
</division><division> <title>Traditional Server-Centric Storage Methods</title> <division> <title>Redundant Array of Inexpensive Disks</title>Redundant Array of Inexpensive Disks (RAID) is a fault-tolerant grouping of two or more disks that a server views as a single disk volume. It is a self-contained, manageable unit of storage.
</division> <division> <title>Just a Bunch of Disks</title>Aside from being one of our favorite acronyms, just a bunch of disks (JBOD) is a simple and efficient method for raw storage. Each drive is independently attached to an input/output (I/O) channel.
This method has limited scalability because it requires separate servers to manage multiple volumes of disks. Drives share common power supplies and physical chassis.
</division> </division><division> <title>Storage Networking Technologies</title> <division> <title>Small Computer Systems Interface</title>Small Computer Systems Interface (SCSI, pronounced “skuzzy”) is a parallel-bus interface port used to connect peripheral devices such as RAID, tape devices, and servers. SCSI is a low-cost method of directly connecting devices but is limited in scalability and distance.
</division> <division> <title>Fiber Channel</title>
A fiber channel is both a physical connection and a Layer 2 protocol used by storage area networks (SANs). Fiber channel is the most common method of transporting SCSI commands and data between servers.
Fiber-channel capabilities have kept up well with companies’ ever-increasing need for bandwidth. Using fiber channel, application servers can access data on a SAN without impacting the company IP network.
</division> <division> <title>Internet Small Computer Systems Interface (ISCSI)</title>ISCSI (pronounced “i-scuzzy”) is a method of encapsulating SCSI data and command frames into IP packets, enabling universal access to storage devices and SANs over standard TCP/IP networks.
</division> </division><division> <title>Retrieval Methods</title>The two main methods for retrieving stored data are block retrieval and file retrieval. A block refers to the largest amount of data that a single operation can access. Blocks are the most elemental units of storage. SANs access blocks directly.
A file consists of several blocks. In network attached storage (NAS) systems, the server organizes blocks into files.
</division>At-A-Glance—Network Attached Storage Versus Storage Area Networks
NAS uses specialized file servers to connect storage devices to a network. NAS is well suited for collecting, storing, retrieving, and sharing data over IP networks. NAS also supports multiple operating systems, file system, and protocols. NAS is optimized for file-based access to shared storage over an IP network.
</division> <division> <title>Storage Area Network</title>
A SAN is an independent network designed specifically for connecting storage devices. SANs are optimized for the efficient collection, storage, and retrieval of raw block data. Most SANs use fiber-channel interconnections and require a media converter to connect to an IP network.
</division> </division>
