Particularly businesses with mission critical or highly sensitive data benefit from the use of a Storage Area Network as they help to secure the bits and bytes distributed within a network in a sensible and effective way.
Storage Area Networks, SAN for short, are high-performance networks that primarily ensure communication between storage devices and computer systems. Such a network has a switched structure in which a wide variety of storage systems can be connected, from tape libraries and JBODs to entire storage arrays. In addition, a SAN consists of further components such as host bus adapters, switches, routers, multiplexers and much more. Within a SAN, data can be stored in various ways, directly in primary storage (hard disks or solid state drives), via various storage devices on tape or as mirrored data in a backup data center. Almost all storage processes can be automated. In contrast to a network attached storage system, the SAN network supports SCSI protocols, IP and the Fibre Channel Virtual Interface (FC-VI). Unlike NAS devices, SAN provides block-level access to data.
Heterogeneous or homogeneous
Storage networks can be heterogeneous or homogeneous. Despite all the advocacy of open networks, a homogeneous network is still the most effective and easiest to manage. A homogeneous network has components from only one manufacturer. This ensures a smooth interaction of the devices, and the software supplied is precisely adapted to this storage network. However, if the administrator inserts a component from another manufacturer into his SAN, this new component cannot simply be managed at the same time. In most cases, an additional tool must be purchased to ensure management. This severely limits the flexibility of homogeneous networks, as they do not allow for arbitrary extensions.
In contrast, heterogeneous storage networks are composed of products from a wide variety of providers. The administrator sets up the SAN independently of manufacturers. But not every hardware is compatible with that of another manufacturer. Interoperability is the focus for heterogeneous SANs. In addition, these networks place increased demands on the management software, as it must recognize and manage all individual devices. As the trend is more and more towards heterogeneous SANs, many manufacturers have joined forces to create manufacturer standards for such an “Open SAN” and to ensure the interoperability of their products. However, many promises of the manufacturers fall by the wayside, especially in the area of interoperability, as the lack of standards means that not all SAN components can recognize and work with each other. Software providers such as Fujitsu-Softek, Tivoli or Veritas already offer products that manage the storage network relatively device-independently. Nevertheless, the IT administrator should find out exactly which systems are supported by the respective software. Hardware manufacturers such as Brocade, HP Enterprise, DellEMC or Pure also try to allow an open storage network through cooperation, joint developments and mutual certification.
Advantage storage network
Especially companies with business critical or highly sensitive data benefit from the use of a Storage Area Network. For example, SANs use different backup levels to ensure highly available data that can be accessed quickly and at any time. Companies that provide web services or video streaming, but also insurance companies and e-commerce businesses depend on high availability to provide their customers with the desired service at all times. Another advantage of a storage network, especially in economically difficult times, is the consolidation of storage systems. Many companies work with numerous storage islands, which are difficult to manage as the flood of data increases. A SAN can not only centralize these islands, but also avoid unnecessary redundancies. Instead of connecting several hosts to different systems, each server or even client can be allocated one or more volumes of storage – within a network. This gives administrators a quick overview of all their resources. Policies can also be used to set an “overflow” marker so that more storage is automatically added as needed.
A SAN also facilitates backup processes. With FC-over-IP, backup data can be transported directly into the storage network, thus relieving the data traffic of the production network, the Ethernet network. Faster backup processes relieve the CPU of the host and thus ensure that the freed capacities can be used for another application, for example. In addition, a SAN promises better bandwidths and faster data exchange, since the block-level protocol is currently one of the fastest network protocols. With to 128 GBit/s via Fibre Channel and 400 Gbit/s via Ethernet a SAN can easily outperform local Storage, which typically is linked between 6GBit/s on SATA and up to 64GBit/s on NVMe Drives.
The biggest advantage, however, is certainly the disaster tolerance. To enable fast restores and business continuity, SANs can be set up over long distances or data can be mirrored to a second SAN quickly and easily. The SAN connection for this is made via MAN/WAN, whereby wavelength multiplexers and channel extension systems, for example from CNT, are used. An often used scenario is the mirroring of data via MAN. Here, data from a SAN located in the company is sent to a remote location. Often the infrastructure of the second data center also forms a mirror of the local SAN. Automation processes enable the user to access the second network directly in the event of a system failure without losing data or having to accept downtime and a costly restore. Especially the financial sector uses this kind of data backup for their business continuity.
Before a company implements a SAN, it is important to carefully examine what requirements the storage network must meet and whether it is the right solution for this purpose. In addition to reliability, fault tolerance, throughput, scalability and feasibility, the costs must be considered in any case. If the primary question is what such a network costs, the real question is what a loss of data or downtime costs. Depending on the industry, financial losses can range into the millions, and in the financial sector they can be many times more. Depending on the size of the SAN to be implemented, the investment costs certainly seem lower compared to the loss figures.
Furthermore, it must be examined whether a SAN really represents a sensible solution to a problem. If you are a large company with more than 100 employees who also exchange different data with different backup requirements, a SAN is certainly the right solution. Smaller companies, which may just want to relieve their server and share files, can use a NAS filer from Network Appliance.
Since every company wants to grow, the network to be set up must of course be modular and its components must be able to be expanded independently – and above all without interrupting productive operation. This applies to both current and future technologies. A SAN can also fulfil these conditions.
In order to successfully implement a SAN, many manufacturers also offer the corresponding services as Professional Services. First they evaluate the existing infrastructure and data volume and determine which problem the storage network should solve. From this, they then derive the available technical options for high availability, consolidation, fail-safe and much more. Then they determine which hardware and software meets these requirements. As a third step, the architecture is defined and tested to determine whether it can cope with the company’s requirements. Once the architecture is in place, the configuration is tested for functionality, interoperability and troubleshooting. In the fifth and final step, the data is migrated from the existing hardware to the SAN and the network is put into operation.
Design and configuration
Due to today’s common switch design, all components of a SAN can be arranged according to a fixed pattern regardless of the size of the company. From the simplest structure (disk – computer) to common network components (disk – switch – computer) to complex enterprise networks (for example tape – disk backup – switch – disk – director – switch – computer), many configurations are conceivable. If the expandability of the network was taken into account in the original design, adding components later is usually unproblematic.
In general, the topologies point-to-point, arbitrated loop and switched fabric can be distinguished in a Fibre Channel architecture, whereby the latter is mostly used in the SAN. In a point-to-point architecture, the components are directly connected to a switch or a bridge to fully utilize the Fibre Channel bandwidth. Arbitrated-Loop, on the other hand, connects up to 127 devices in a ring formation without the need for a switch. Data is sent to each device in turn and the bandwidth is shared. Only storage devices that do not require high bandwidths are suitable for such an arrangement. A loop configuration can be used well as an automatic backup system. This involves backing up data over multiple drives that are rarely used.
Switched fabric is the predominant form of SANs. This type of network allows easy scaling, as switches can be used to connect thousands of devices. Switches also allow more flexible use of bandwidth by using different paths between nodes. Within the network there are different switches that control the data transfer: switches that integrate arbitrated loops and thus reduce data islands, switches that connect point-to-point topologies and switches that connect to other switches.
Depending on the amount of data and storage conditions, the SAN is designed and configured. In addition to servers, JBODs, tape storage devices or RAID systems, the components also include host bus adapters, cables and GBICs, hubs, switches and possibly multiplexers as well as DWDM devices for connecting to MAN/WAN connections. A smaller SAN, for example, could send information from three hosts via an FC switch simultaneously to a JBOD and a RAID system storage device, from which all changed data is sent at night to a tape library as a backup. In larger configurations, clusters are connected to different types of storage such as storage arrays, JBODs, and RAID disks via host bus adapters and multiple switches. To guarantee high availability, the data is automatically mirrored to a backup location. Usually the backup data center is located several kilometers away, whereby the distance is bridged by multiplexers or channel extension systems.
The SAN configuration depends strongly on the value of the data to be backed up. Business critical data is kept in a RAID configuration, while negligible information is immediately archived on tape or deleted according to a time schedule.
Already during the design and planning phase, care must be taken to ensure that the infrastructure to be set up can be managed as logically and efficiently as possible without great effort. The entire data strand must be considered here, i.e. from the hard disk via the switch and HBA to the database and its applications. Here, too, the devil is in the detail, as often only the hard disk and the switch are considered. Ideally, a single control application should be able to centrally manage all the above-mentioned components from as many manufacturers as possible.
A key benefit of SANs is that they make it easier to manage their storage resources. Virtualization in a storage area network stands for an abstraction layer that allows hardware-independent storage access. The installed storage systems appear to the administrator by means of special software – often a volume manager – as a uniform storage pool or as a virtual hard disk, whose storage capacity the administrator allocates, copies and moves to any server from a central console. This concentration alone allows 80 percent of the storage capacity to be used instead of 50 percent, as in the case of directly connected storage systems. This storage virtualization allows the user to adjust the storage capacity required during operation by “drag-and-drop”. In addition, new hard disks and additional storage systems can be integrated into the storage pool without interrupting operation. Simpler variants of storage virtualization work either on a server, which has several storage systems available, or on the storage system itself, to which several servers are connected. Virtualization simplifies the previously complex procedures for changing the individually available storage capacity, for data mirroring, data duplication and data backup.
By means of virtualization, the administrator distributes the storage space to individual hosts or allocates storage for clusters. The automation rules for backup, data transfer, data clones or mirroring and snapshots can also be set using virtualization software. SANs simplify many problems of data protection and guarantee highest data availability. However, it takes some effort until such a complex infrastructure is in use. Every company must consider whether a storage network really solves all problems or only adds new ones. The decisive factor for a successful implementation is always exact planning, taking into account future requirements and the expected data volumes.