Data Redundancy
The most vital and most expensive thing to a company is their data. My family has a family business. One of the aspects of the business is processing and developing pictures for photographers. Photographers these days use digital. These images are saved on a file server, so multiple people on the network can have access to these images. If the hard drive of the file server fails, and if there is no data redundancy, the company could very well go out of business.
Data is no joke. Fortunately, there are ways to provide data redundancy as well as increased performance. Some guys saw the importance of this, and created RAID.
RAID stands for Redundant Array of Inexpensive Disks/Drives. There are three main RAIDs that are implemented. Those are RAID 0, RAID 1, and RAID 5.
RAID 0
RAID 0 is disk striping. This provides no data redundancy. Let me repeat. This provides no data redundancy. This only provides increased performance. Basically a file is broken down into smaller chunks of data, and those chunks are saved to multiple drives. So lets say there is a RAID 0 array of two hard drives. A 40 MB file is saved. About 20 MB will go to one hard drive, while another 20 MB will go to the other hard drive. Because the data is broke down into chunks, the hard drives save and load less data. This speeds up the saving and loading process.
RAID 1
RAID 1 is disk mirroring. This provides data redundancy. Basically data is saved on one drive, and a back-up is automatically saved to another hard drive. There is also duplexing. Duplexing is where two RAID controllers are installed in the server. Each hard drive is connected to each controller. This provides increased system performance, because each controller writes the data once, as opposed to one controller having to write the data twice.
RAID 5
This is the RAID that is implemented in the optimal server. RAID 5 is disk striping with parity. At least three hard drives are needed for this RAID level. Basically data is striped along three hard drives, but there is an extra parity bit that is added to the data. This parity information is saved to each hard drive. If one hard drive fails, the other hard drives are able to reconstruct the data using the parity bit. When the failed hard drive is replaced, data will automatically be transferred to the new drive with the parity information. This not only provides data redundancy, but it also provides increased system performance because the data is striped along at least three hard drives.
Back-up Power Supply
If the file server in my family’s company were to go down, work could not be done, and orders would be late. Customers wouldn’t like this, and would probably go to another lab in the future. This would in turn hurt the company in sales.
Fortunately, there is a way to get past this. You can install a back-up power supply to server machines. If one power supply fails, the other power supply will automatically kick in. Then, you can replace the bad power supply. And you can even do this while the server is running!
Uninterruptible Power Supply (UPS).
An UPS provides clean back-up power as well as surge suppression. If the power were to go out, you get about a few minutes of power to your system. Then you are able to shut down the system cleanly. It also provides surge suppression. If a power surge were to hit your building, the surge would go in a capacitor in the UPS. This in turn will prevent the surge of electricity to fry the fuse in your power supplies. There are two types of UPSs. There are Stand-by Uninterruptible Power Supplies, and UPSs. In SUPS, power goes from the outlet to the SUPS, and then power from the batteries in the SUPS goes to the computer. This provides line conditioning. That means you get clean, steady power. Power naturally sags and spikes from your power company. Line conditioning gives steady power. UPS do not support line conditioning. SUPS would be implemented on my optimal server.
Back-up Network Interface Card
If the NIC were to go bad, data could not be transferred. And well, work couldn’t be done, and sales would also probably decrease for the company. It is important to install two NICs into your server. If one goes bad, plug the network cable into the other NIC, and then replace the bad NIC. (NIC is also known as an Ethernet card.)
Make sure to get the proper NIC for your network. If your network can run at 100Mbs, you don’t want to get a 10Mbs NIC. This would cause the server to transfer data at 10 times slower than the rest of the network! If anything, get a NIC that can go faster than your network speed. That means if you upgrade your network, you don’t have to replace your NIC.
The more expensive NICs support full duplex. This allows the NICs to send and receive at the same time. These NICs are put into what I call the optimal server.
Continue to Page 2 |