Storage Area Network

from Wikipedia, the free encyclopedia

In the field of data processing, a storage area network ( SAN ) or storage network is a network for connecting hard disk subsystems (disk array) and tape libraries to server systems. Storage area networks are designed for serial, continuous, high-speed transfers of large amounts of data. Today, for high-availability, high-performance installations, they are based on the implementation of the Fiber Channel standard, and for SMEs also on IP for cost reasons .

Definition of a SAN

A SAN is an extension of Direct Attached Storage (DAS), whereby disk storage capacity is assigned to a server via a network, but can also be assigned to a server dynamically, i.e. at any time, within operational limits / requirements. The latter functionality is mainly implemented in SAN-compatible disk arrays. While DAS forms a point-to-point connection between a server and a directly connected data storage device, a SAN enables several servers to be connected to several storage systems via a network, even over large distances.

In a SAN, as with DAS, data access is block-based , the access to the storage device and any file system located on it are therefore managed by the accessing computer. With NAS, on the other hand, entire files (for example: "File /home/user/readme.txt") or sections of files are accessed via SMB / CIFS or NFS at a remote storage location. In this case, the associated file system is mainly managed in the server . In SANs usually one is SCSI - communication protocol used, the on Fiber Channel or iSCSI as the transport protocol touches. Implementations based on ATA over Ethernet (AoE) are seldom and practically only represented in the SOHO area . There are also the first SAN solutions based on other communication models such as RSIO or InfiniBand as well as within the framework of a convergence of IP and storage networking.

A SAN is structured in the same way as a Local Area Network (LAN): There are hubs , switches and routers.

Development and principle of operation

The SAN was developed to counter the management problem of dedicated hard disks in server systems or network-attached storage systems (NAS), as these systems can only use the storage capacity efficiently and flexibly to a limited extent.

Another problem with NAS systems is that they also burden the existing LAN with access to the data. Furthermore, the Ethernet , via which the NAS systems are connected to the servers or clients , with its small frame sizes and its relatively large protocol overhead, is not designed for fast access to mass storage devices.

Today, SANs are mostly formed using fiber optic cables; the system used for this is called Fiber Channel . A simple SAN consists of a fiber channel switch , one or more hard disk subsystems and the servers, which are connected to the fiber channel switch via so-called host bus adapters , or HBA for short.

Today you work with transfer rates of up to 16 Gbit / s. Since they use a special protocol adapted to the requirements of mass storage usage, transfer rates of theoretically 1.6 GB / s are possible. In addition, there is the concept of multi-pathing , which is consistently pursued in the SAN.

Schematic representation of a SAN

This example SAN consists of two switches, each of which forms its own fabric . The servers are connected to each fabric with a host bus adapter , as is the disk array.

Disaster tolerance

Another great benefit of SANs is their disaster tolerance. This includes hard disk subsystems that have internal components with integrated RAID functionality and that can handle the various RAID levels (usually RAID 1, 10, 5, 50, 6 or 60) automatically and independently of the server. Furthermore, all important systems can be designed multiple ( redundant ). In a typical small storage area network, it is conceivable that an identically constructed disk subsystem is located at two locations on the company premises that are as far apart as possible, each of these disk subsystems being connected to one of two switches that are also separately installed. The servers, unless they are also distributed systems, have at least two host bus adapters . Each of the two host bus adapters on a computer is connected to one of the two switches. The data of both disk subsystems are mirrored, depending on the installation, either by the server or automatically by a functionality integrated directly into higher-quality disk subsystems ( RAID 1 ).

In the example above, failure of individual lines, a switch or even a disk subsystem would be conceivable without impairing the overall system performance.

In larger, high-availability systems, the server systems are also kept redundant. Furthermore, to protect against a complete failure, for example due to natural hazards or damage caused by external influences, it makes sense to move the redundant components to another, more distant location. Dedicated lines in the range from 1 Gbit / s to 9 Gbit / s are then used for permanent synchronization of the data on the disk subsystems . The basis for SAN networking in the WAN area is usually fiber optic networks, which also guarantee galvanic isolation . In order to achieve better fiber utilization, the glass fibers are often connected with Wavelength Division Multiplex systems (WDM). Using appropriate protocols such as Fiber Channel or Gigabit Ethernet, data can also be safely transported over very long distances. However, the latency , which plays a major role in transmissions over longer distances, must be taken into account .

Problems

Even today, after the SAN has established itself in many areas, full compatibility between components from all manufacturers is not necessarily guaranteed. When new components are used, they must therefore be checked for compatibility with the existing components, which is costly and time-consuming. In the past, many users have therefore committed to the components of a single manufacturer in their storage area networks in order to avoid these complications. The Storage Networking Industry Association (SNIA) is an initiative of various manufacturers that aims to counteract such problems through uniform standards. In the meantime, however, various manufacturers guarantee the required functionality even in a multi-vendor configuration that has been extensively tested by them.

history

Technologically, a SAN is a further development of the ESCON technology introduced by IBM , resp. the FICON technology based on it.

See also

literature

  • Ulf Troppens, Rainer Erkens u. a .: storage networks. 2nd Edition. dpunkt, Heidelberg 2007. ISBN 978-3-89864-393-1
  • Björn Robbe: SAN - Storage Area Network. Hanser Fachbuchverlag, Munich 2004. ISBN 3-446-22597-8
  • Roland Döllinger, Reinhard Legler, Duc Thanh Bui: Practical Guide to Storage Solutions. dpunkt, Heidelberg 2010. ISBN 978-3-89864-588-1

Web links

Individual evidence

  1. ↑ RSIO technology information (PDF; 362 kB)
  2. ↑ InfiniBand Storage (PDF; 210 kB)