RE: Failsafe System BackupSorry for technospeak guys, and bad typing
FYI
SLA = Service level agreement,
eg we garuntee system available 95% of the time
In oracle customers databases are stored on a logical partion, effectively a logical disk. Logical disks are different from real disks often (and hopefully) becuase most servers these days use RAID arrays.
RAID = Redundant array of inexpensive disks.
For example, if I have 4 drives, ABCD each 100GB there are many ways of configuring drives in an array.
RAID 0, No Redundancy, only one copy of data, total capacity 400GB
RAID 1, Mirroring, Total capacity 200GB, each physical drive would be mirrored to one other, A=B and C=D.
RAID 5, Capacity is (n-1) drive capacity, 300GB
If you think of data as 1's and 0's, then add up the bits on three drives and record parity on 4th, eg, 000=0, 001=1, 011=0, 111=1
Thus if any one drive fails, by adding up the bits from the other three, you can calculate the missing data.
RAID10, Two Raid5 Mirrored on separate controllers, gives contigency against controller failure...
A 200GB array can be partiioned into 1x200GB, 2x100GB or 4x50GB... logical drives. The raid controller presents these 'logical drives to the server operating system, (Windows, Unix, Novell...) To them it looks like a familiar physical hard drive you have in your PC. The raid controller simply lies to the server. If a drive fails, it may contain data on several logical disk partitions, but becuase there is contingency, there is no data loss, the server doesn't even know it has happened, usually two or more drives have to fail to cause data loss. Once the failed drive is replaced, it is automatically formatted and populated with data, agin without the server knowing.
Clustering:
Imagine you have a box of drives in a RAID configuration.
You plug two or more servers into these drives.
servers also talk to each other and share out the work to be done, if one server fails, it work is picked up and completed by remaining servers with no more than a second or two hiccup.
Oracle being plain clever.
In the same way drives are mirrored in an array, Oracle can mirror records in a database, such that if one database corrupts or fails, servers can continue to operate using data in the mirrored server. I don;t know too much about oracle but I beleive these mirrored copies can be on separate sites, and both can be posted to simultaneously, both updating each other. Thus if you lose a site, the operation can still continue.
As I said, betond backups, I I don't know how of this fault tolerant stuff we have, it tends to add to costs and money has been tight. I would expect this stuff implemented once e see income rolling in and money and technical resource situation easing up.
Idle