Avoiding Data Loss

Losing Data Costs Money

If a company stores data, then that data has value. If it did not have value, then the company would not store it.

Some company data is extremely valuable. Think about the Sales Ledger. This is the list of who owes us money. If the Sales Ledger is lost it will be extremely difficult to collect this money. The Sales Ledger is valuable data.

Ways in Which Data Might Be Lost

1. Fire, started accidentally or deliberately.

2. Files being accidentally or deliberately deleted or corrupted.

3. The file server is stolen.

4. Power fails during a file write operation.

5. A power surge (too many volts) effects the operation of the computer.

6. A component in the computer (for example its hard drive) fails.

7. A natural disaster occurs, such as a flood or an earthquake.

All of these can lead to data being lost.

Various ways exist to reduce the damage caused by these problems.

Tape Backup With Off Site Storage of Tapes

What kind of tape system?

As with most things, you get what you pay for. More expensive systems have bigger capacities, work faster and are more reliable. On commonly used technology is DAT - digital audio tape.

What backup software?

‘Free’ backup software is often included in the NOS. Most administrators do not use this, preferring to buy better quality programs from other companies. Some systems can do unattended backups during the weekend or evening.

Which files to back up?

Some administrators back up everything - data and programs. They hope that this will make it easy to get the network running after a failure.

Others back up only user generated data files. They plan to reinstall programs from the manufacturer’s CD after any failure.

The role of the "archive required" attribute?

Files have attributes such as ‘Read Only’ and ‘Hidden’. They also have an attribute called ‘archive required’. This attribute is used to indicate that a file needs to be backed up. Each attribute is represented by one bit of a special byte that is stored along with the file.

When a new file is created its archive bit is set to ‘1’ to indicate that it needs backing up. When the file is backed up, the archive bit is set to ‘0’. Next time the file is modified it is reset to ‘1’ to indicate that it needs backing up again.

3 kinds of backup.

Full backup: All files are copied to tape and all archive bits are

        set to  0.

 

Incremental backup. Backs up only files with their attribute bit

         set   to 1. The attribute bit of these files is then changed to 0.

 

Differential backup: Backs up only files with their attribute bit

         set to 1. The attribute bit of the backed up files is left as 1.

Typically an administrator might do a full backup once a week, then either an incremental or differential backup each day. A different tape is used each day.

Ideally, backups should be done when the network is not busy.

Storage of backup tapes

At least some of the tapes should be stored off site. On site tapes will not help if the office burns.

Backup logs:

In an organization with many servers and very important data, there will be many backup tapes. Some organized system of cataloging the tapes will be needed, along with a record (log) of what was backed up when.

Where should the tape drive be put?

It can be connected to any computer on the network. Direct connection to the server is best because the data to be backed up will then not have to travel across the network.

(Some backup software also lets the hard drives of workstations be backed up to the server’s tape drive.)

UPS - Uninterruptable Power Supplies

A UPS is a box containing a big battery that goes between the server and the wall power socket.

It does several jobs:

1. It filters the power coming from the wall socket. If a ‘spike’ of dangerously high voltage comes, the UPS stops the spike damaging the computer.

2. If the voltage from the wall becomes too low to operate the computer reliably (or the power fails completely) the UPS:

a) Powers the PC using its internal battery

b) Sends a message to the server causing the it to:

i. send warning messages to the Administrator and all users

ii. run normally for a few minutes,

iii. close all files then shut down.

RAID - Redundant Array of Inexpensive Disks

RAID is a standard for having data duplicated on multiple hard drives within a server. If one drive fails, there is a copy of its data on another drive.

There are 5 ‘levels’ of RAID. Windows NT Server supports levels 0, 1 and 5.

                        Level 0 - Disk Striping

Imagine saving a 12 KB file to a server with six hard drives. With a striping system 2 KB with be written to disk #1, 2 KB to disk #2, 2 KB disk #3 and so on. The file forms a stripe across the disk set.

This makes the system run a little faster, but does not give protection against data loss.

Level 1 - Disk Mirroring

For every hard drive in the server, there is also a matching second drive. All of the data written to one drive is copied to its mirror drive.

Everything is duplicated. If one drive fails, we can access our data from the mirror drive.

Level 5 - Striping With Parity

In this system a file is spread across a strip - as in level 0. The difference is that for each file an extra block of data is also written - called parity information. This data is calculated from the bytes of the data file using a mathematical formula. It has the special property that if any piece of data from the file is lost it can be regenerated from the parity information. The size of parity information is typically about 15% of the size of the associated file.

The parity information is not kept just on one disk, but for different files it is kept on different disks. If any disk fails, the data on it can be recreated from the parity information for each of the files it held.

 

Note: RAID by itself is not enough. Even with RAID, tape backups must still be made.

 

Back to top

| list of lessons | Course Plan | Instructor's Info | Links | UAE | Guestbook |View Guestbook | Books | Search|HCT|