Types of Storage
There are three types of storage you can provide to your customers: Direct Attached Storage (DAS), Network Attached Storage (NAS) and Storage Area Networking (SAN). What you choose will affect the cost, the solution and the levels of support and services you can offer.
This is the most common type of storage, ranging from disks inside servers to USB, eSATA and SAS drives and drive arrays. The most common is internal storage with a server. The drawback is that it is limited to a single server and you often end up with a lot of unused or partially used storage as it is in the wrong place or the server performance is too poor to allow it to be used effectively.
The current range of USB storage solutions, while very cheap, are not really suitable for server related tasks as anything other than as a backup device. They should only be used for laptop or desktops.
eSATA offers a much higher speed than USB; comparable with internal SATA devices. Many eSATA drive arrays will hold up to four drives, though larger systems are available; they can be configured as RAID 5 storage and provide cheap expansion. Their main use is in environments where there is no real space to add more internal drives or for desktop users with very specific application requirements.
SAS drive arrays are promoted by storage vendors large and small. They are directly attached to either a small switch or, more commonly, to a single server. Many of the larger vendors have designed their SAS arrays so that they can be chained together using a single server as the access point. Enterprise solutions in this space are getting cheaper; the HP MSA 50, which can hold ten small form factor drives, is currently selling for around £1,300 new and £750 second-hand.
Network attached storage was originally designed as a poor man’s SAN. NAS boxes have their own bespoke, embedded operating system (usually Linux based) and increasingly come with additional software to do backup from multiple servers and desktop computers. 4TB rack-mount solutions cost between £1,600 and £1,900.
The danger is that many of these solutions have a limited upgrade path; for example, you may not be able to change out the drives for larger capacity drives due to the way the internal operating system and software work. They also lack redundant power supplies and often have just a single 1GB Ethernet port. This means that they can quickly become a bottleneck on the network.
At the top end of the NAS market you get multi-core, multi-processor servers with as many as 24 drive bays, multiple network ports and the ability to add or change the storage as you like. These come with redundant power supplies, multiple network ports with the option of adding more, and support for iSCSI and even Fiber Channel. This support for storage protocols often makes them crossover devices between NAS and SAN. Starting price is often in the £10,500 region and as you add software features, the price will increase.
There are two main issues with using multiple NAS boxes. Firstly, like servers, they need to be managed individually. Secondly, many of the vendors require you to buy replacement drives directly from them and they add a significant premium over the cost of commodity drives. However, you run the risk of invalidating your warranty if you use cheaper drives.
Storage Area Networking takes all of your storage and creates a virtual management layer over the top. Costs can be extremely high as you need to buy a storage fabric and then integrate your hardware into it.
Over the last five years, however, software virtualisation products capable of taking your existing hardware and creating a SAN out of it have arrived. Datacore and HP Lefthand are two of the leading vendors in this space. It is not necessarily cheap to get started but with features such as deduplication, single instance management and thin provisioning, the savings mount up over time. A 32TB SANmelody licence will cost just under £21,000 but this does allow you to reuse existing hardware.
The current street price of desktop 1TB SATA hard disk drives is under £50 (ex-VAT) compared to the cost of an enterprise 300GB SATA drive at £140 (ex-VAT). This makes it very tempting to install cheaper drives when upgrading, building or restoring hardware. But what makes the difference between these drives?
In a standard desktop drive the spinning speed is no more than 7,200 rpm while enterprise drives spin at either 10,000 rpm or 15,000 rpm. This difference in speed is significant when it comes to the quality of the components used for moving parts, including the spindle, the read/write heads and the actuators.
Desktop drives have a single chip responsible for handling the command queuing (reads and writes). That queue controls how data is written and read from a drive and a heavily accessed drive will have to manage a lot of head movement. In an enterprise drive, there are two chips, one for read and one for write. These manage separate queues for reading and writing allowing them to be more efficient as they can read and or write more data in a single pass over the disk.
The ability of a drive to read and write data is also assisted by the buffer size: RAM on the drive motherboard. The larger this buffer, the more can be held in memory. But here is a risk. When data is held in the buffer, it can be lost when there is a power outage. As a result, most drive array controllers turn off buffering of data on the drive and use memory on the controller card which is battery backed up in case of disaster.
The hardware differences mean the difference between consumer and enterprise drives is quality, performance and longevity. Choose cheaper drives over reliable storage at your peril and cost.