Any computer with a SCSI or Fibre-Channel adapter can function with a Sunrise RAID array. The RAID array is system-independent, so it appears to the host computer (or computers) as one large-capacity hard disk drive. No special drivers or utilities are required - just set up the RAID array in exactly the same way as you would any hard disk attached to your computer.

It is not even necessary to have a server to plug the RAID array into because a 'thin server' can be built into the RAID system, allowing the RAID to be connected directly to the Ethernet network. The thin server can be configured from any computer using a web browser and accessed by any Windows ™ or NFS client.

RAID (Redundant Array of Inexpensive Disks) systems are faster and more resilient than single large hard disk drives. A RAID system stripes or mirrors your data across multiple hard disk drives. Therefore, when you access your data, all the disks are working at the same time, giving a major increase in speed. The RAID hardware controller also contains a large memory cache (typically 32MB or 64MB of RAM) which dramatically increases performance. A RAID system uses the capacity of one disk to retain parity information for the other disks, so that if any disk fails, the contents of the failed disk can be automatically reconstructed from the data on the other disks. Even with a failed disk, the data in the RAID array can still be fully accessed.

RAID offers different ways of achieving greater control over levels of reliabiliy, capacity, performance and cost. There are a number of different methods of configuring a system, commonly known as RAID levels. RAID levels 0, 1, 3, and 5 are by far the most common. A Sunrise solution can utilise any one or a combination of these RAID levels, ensuring that even the most individual requirements can be met.

Each RAID level has different characteristics, benefits and disadvantages. The level number does not directly reflect the level of performance, capacity or redundancy, and are purely used for identification.

The appropriate RAID level for any installation depends entirely upon the individual application and business needs.

Until recently, all RAID systems used SCSI disks. However, it is possible to use just about any type of disk in a RAID array. IDE disks with DMA can now run at up to 100 MB/sec, but because they cannot be daisy-chained in the same way as SCSI and Fibre-channel drives, they require a special RAID controller.

Economy RAID systems such as the E800 series offer good to very high performance whilst using inexpensive drives. However, they do need one channel per disk, so are limited to a maximum of 8-12 disks in one RAID array. The RAID controller uses an Ultra-3 SCSI host channel because IDE has a very limited cable length and lower bandwidth. This means that although IDE disks are used, the RAID appears to the host as a normal SCSI device.

There are several ways to do this. One way is to use multiple SCSI host channels on the RAID controller to connect to several host computers. One RAID set can be partitioned into several disks, and any of these disks can be mapped to a number of different channels, LUNs, or even to different SCSI IDs.

If you are going to allow multiple computers to access the same section of the disk, care should be taken to ensure that the same file cannot be accessed by more than one computer simultaneously. If different computers try to write tot he same file simultaneously the data on the disk could become corrupted (although this in itself will not cause the RAID array to fail).

This very much depends on how you intend to connect your server(s) and how much redundancy is required. Fibre-channel systems are currently more expensive than SCSI, although costs are slowly coming down. However, there are more advantages to Fibre-channel than just bandwidth (200 Mb/sec for dual loop fibre). Fibre-channel can be used to connect up to 128 devices into a storage area network (SAN) and hubs can be used so that one device cannot break the loop. Dual loops can be connected to reduce single points of failure, and fibre-optic links can be used to cover distances of over 10 miles (with mono-mode fibre). SCSI is both reliable and cost-efficient, but cannot offer the cable lengths and scalability of fibre-channel.

SCSI (Small Computer System Interface, pronounced 'skuzzy') was originally developed as you may have already guessed for small computers. However, today you will find it in everything from a small notebook to a large server and it is used as standard on most servers.

So what are the different standards, and which are the best?

SCSI has been continually developing over many years and has seen many changes in specification. It would take a long time for me to write a 'full' history of SCSI so a summary of the SCSI standards is shown below.

Narrow SCSI: 8-bit bus and data transfer rate of 10 MBytes/sec.

Fast & Wide SCSI: This is where SCSI started to become 'fast', according to the people who named it. It was 'wide' because it had grown from an 8-bit bus to a 16-bit bus and was therefore a wider data path than before. Fast & Wide SCSI can transfer data at speeds up to 20 MBytes/sec.

Ultra SCSI: Fast & Wide was an excellent improvement on older standards, but as it had to remain compatible with slower technologies it had many limiting factors. Ultra SCSI increased data transfer rates up to 40 MBytes/sec.

SE or HVD: Both the Fast & Wide and Ultra SCSI standards were capable of being supplied in two formats - SE (Single Ended) or HVD (High Voltage Differential).

The greater the cable length, the greater the amount of noise it will pick up. When the noise reaches a certain level, errors reading the data across the bus will occur. One of the disadvantages of increasing the speed of SCSI was that there was a corresponding increase in susceptibility to noise.

The SE (Single Ended) standard is common, but limits the cable length to around 1.5 metres on Ultra SCSI. This can obviously be extremely limiting for external storage applications.

HVD (High Voltage Differential) converted the data into + & - pairs much like you find on many network configurations. This reduced the susceptibility to noise and thus functions over significantly longer cable lengths than the SE standard.

Ultra-2 (LVD): This is where the SCSI standard really started to move! We knew that one of the limiting factors of increasing the speed was interference from noise. We knew that HVD solved this issue but at the same time, high voltage meant greater power consumption, leading to greater heat, etc. The new LVD (Low Voltage Differential) standard solved these technical issues.

Ultra-2 SCSI or LVD can transfer at up to 80MBytes/sec over cables of up to 12 metres length.

Ultra-3 (LVD)/Ultra-160: The current industry standard, and twice as fast as Ultra-2 SCSI. Data transfer rates of up to 160 MBytes/sec can be achieved.

Ultra-4 (LVD)/Ultra-320: Transfer rates of up to 320 MBytes/sec.

Fibre-channel is a highly reliable Gigabit interconnect technology which allows concurrent communications among workstations, mainframes, servers, data storage systems and other peripherals using SCSI and IP protocols.

It provides interconnect systems for multiple topologies that can scale to a total system bandwidth of the order of a Terabit per second. Fibre-channel delivers a new level of reliability and throughput.

Switches, hubs, storage systems, storage devices and host adapter cards are among the products that are required to implement a total system solution.

The answer to this question is explained in the white-papers and technical documents within the High Availability Software area of this website.

If the system, in this case a storage system, is capable of continuing to function after a component has failed, then that component is considered to be a 'redundant' component.

Sunrise Computer Maintenance Ltd enclosures are supplied with a minimum of dual or triple power supplies and the system power requirements are 'split' between them. If one should fail, then the remaining power supply units will simply provide more power until the faulty unit is replaced. Audible and network alarms will be activated to alert a technician, but at the same time the RAID system will continue to function.

Redundant components such as power supplies, fans, RAID hardware controllers, etc, are often referred to as 'hot pluggable' or 'hot swappable', which fundamentally ensures a clean component replacement and zero RAID system downtime while the component is replaced.

Let's look at a real-world example:
Vodafone utilise a Sunrise Computer Maintenance Ltd RAID array on their 24-hour mobile phone network.

The RAID array is attached to two hosts, in this case OEM Sun Enterprise 450 based systems. The RAID level 5 protected capacity is split 50/50 and each half assigned to one host.

Under normal operation, each host simply uses its own half of the RAID array. However, if one host should fail, then the 'good' host simply picks up the failed host's half, therefore allowing users to continue to access their data.

NAS is a disk array storage system that is attached directly to a network rather than to the network-server (i.e host-attached); functions as a server in a client/server relationship; has a processor and processes protocols such as SMB and NFS.

Sunrise Computer Maintenance Ltd offer a complete range of Network-Attached Storage solutions.