A dual system (デュアルシステム) is a setup in which two systems carry out the same processing at the same time and compare the results to confirm they are correct.
Input
Both systems run the same processing
Compare the two results
Output once they match
Both systems handle live processing, so if the comparison turns up a mismatch, one of them has gone wrong. Running two full sets of equipment at all times is costly, but it suits work that demands high reliability.
AnalogyIt is like an accounts office where two clerks each total the same voucher and check every time that their answers agree.
How to remember"Dual" means "two". Picture a swordsman with a sword in each hand, both blades in action.
Exam points
ExamWhen one side of a dual system (デュアルシステム) fails, that side is cut off and the other carries on with the processing alone.
A duplex system (デュプレックスシステム) provides an active system (現用系, the primary) that handles live processing and a standby system (待機系, the secondary) kept ready in case it fails, and switches over to the standby system when a failure occurs.
Active system handles live work
Active system fails
Switch over to standby
Standby system takes over live work
The standby system takes on live processing, and work carries on.
4 / 4
In normal operation only the active system does live processing, and the standby system can take on other jobs while it waits. When the active system stops, processing moves to the standby system so work can continue, though service may pause briefly during the switchover.
AnalogyIt is like a bus company whose relief driver does office work most days and only takes the wheel when the regular driver suddenly calls in sick.
How to rememberIn English, a "duplex" is also a house split into two homes that share a wall. Picture the unit next door, always ready to move into.
Exam points
ExamIn a duplex system (デュプレックスシステム), only the active system (現用系) handles live processing, and the spare capacity of the standby system (待機系) can go to other work.
Mix-upA dual system (デュアルシステム) runs live processing on both sets and compares the results. A duplex system leaves live work to one set and keeps the other waiting to take over.
A cluster (クラスタ) links several computers over a network so they work together and appear to users as a single system.
Users can work without knowing which machine is handling their request. If one machine fails, the others take over its work, so the system as a whole keeps running. Adding machines also makes it easy to scale up.
AnalogyIt is like a shop where several clerks share one checkout line. When one goes on break the line keeps moving, and when more customers arrive the shop just adds clerks.
How to rememberA "cluster" can be a bunch of grapes. Many separate grapes look like one bunch, just as many machines look like one system.
Exam points
ExamA cluster (クラスタ) is used not only to withstand failures but also to share a heavy load of requests across machines and raise processing capacity.
Replication (レプリケーション) is a function, mainly provided by a DBMS (the software that manages a database), that keeps a copy of a database on another server and applies each update made at the source to the copy, keeping the two in sync.
Data is updated on the source server
The update is sent to the replica
The replica applies it and the contents match
The replica makes the same change, and the two databases now hold the same contents.
3 / 3
Because the replica holds the same contents as the source, processing can continue on the replica if the source server fails, which raises availability, the degree to which a system stays usable without stopping. Sending read requests to the replica also lightens the load on the source server and improves performance.
AnalogyIt is like a head store whose every entry in the stock ledger is copied into the branch's ledger too. Even when the head store is closed, the branch can answer inquiries from its own ledger.
How to rememberA "replica" is a copy, the same word used for museum reproductions. Think of a function that keeps making copies you cannot tell from the original.
Exam points
ExamReplication (レプリケーション) is a DBMS function that keeps copying and synchronizing data between servers, raising availability (可用性) and helping performance as well.
Mix-upA backup (バックアップ) saves the contents as of a certain moment so they can be restored to that point later. Replication applies every update to the replica and keeps the contents matching.
RAID combines several disks so they can be handled as one, and uses the way data is placed across them to gain speed or fault tolerance.
Level
How data is written
When one disk fails
Usable capacity (n disks)
RAID0 (striping)
Split and written to several disks in parallel
Data is lost
n disks' worth
RAID1 (mirroring)
The same data is written to several disks
The remaining disks keep running
One disk's worth
RAID5
Data and a value computed for recovery (parity) are spread across three or more disks
Can be rebuilt by calculation from the remaining data and parity
n − 1 disks' worth
The ways of placing data are identified by numbers called RAID levels (RAID レベル), and each level strikes a different balance between speed, resistance to failure and usable capacity. The parity (パリティ) in RAID5 is recovery information calculated from the data, so if one disk's worth is lost, the original contents can be recalculated from it and the remaining data.
AnalogyIf you keep a separate slip with just the total of three people's scores, losing one person's slip is no problem: subtract the other two scores from the total to get it back.
How to rememberThe R in RAID stands for Redundant. Only level 0 has zero redundancy, which matches its number.
Exam points
ExamRAID1 (mirroring, ミラーリング) writes the same data to every disk, so it keeps running on the remaining disks if one fails. The trade-off is that usable capacity is only one disk's worth.
ExamRAID5 can recalculate lost contents from the remaining data and parity (パリティ) when one disk fails. Since one disk's worth goes to parity, usable capacity equals the number of disks minus one.
Mix-upStriping (ストライピング, RAID0) splits data and writes it in parallel for speed, while mirroring (RAID1) writes the same data more than once to guard against failure.
NAS is a storage device, connected directly to a LAN, that specializes in sharing files.
Because it has its own built-in functions for managing and serving files, several terminals on the network can read and write the same files without a separate server computer. Many models also build their internal disks as RAID to guard against failure.
AnalogyIt is like a shared bookcase in the middle of the office. Anyone can come over from their own desk to take out or put back the same documents.
How to rememberNAS stands for Network Attached Storage. Read it literally, "storage attached to the network", and the picture is right there.
Exam points
ExamNAS is a file-sharing storage device connected directly to the network, letting several terminals use the same files over it.
Mix-upRAID is a technique for binding several disks into one, while NAS is a device that connects to the network so files can be shared.