It requires
that an update operation on a logical object should be completed successfully
within a quorum. Without enough members to form a quorum, operations
will not be performed thus avoiding conflicting results.
The reliability and performance characteristics of the quorum method can
be controlled by appropriately adjusting the R andW. In the general situation,
comparative to the file replication scheme in [1], each physical copy at a single
server can be regarded as being assigned the same votes, say 1. Thus, W and
R determine the number of nodes composing each quorum. There is a tradeo?®
between overhead consumption in write operations and read operations. The
characteristics of writes improve as W decreases, and similarly for reads by
decreasing R. The choice of R and W will depend on an application??™s readto-
write ratio, the cost of reading and writing, and the desired reliability and
performance.
In [3], the number of nodes within a quorum has possibly been reduced
to approximately ???n. That is, n servers in the system can be partitioned
conceptually into ???n rows and ???n columns. Each column and row serves as
a quorum; all of the quorums form a quorum system. There is always a server
located in the intersection of each pair of row and column. The overhead of the
system is controlled at O(???n). Compared with the O(n) servers involved for
accessing data in traditional read-any/write-all or read-all/write-one replica
schemes, to O(n/2 ) in those having majority criterion schemes, and to O(???n),
the quorum method has significantly reduced tra?±c in the system which can
be translated into significant energy savings.
Pages:
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385