Information Technology Reference
In-Depth Information
of the CPU, this algorithm would result in a waste of time, which cannot be
affordable by any ubiquitous system.
The ADoRe algorithm overcomes the drawbacks of the RxW. Let us
suppose that the queue is usually composed of
n
elements and
K
is a
natural number constant. The ADoRe sorts
n
requests in descending order
according to the RxW value and is responsible to select the first
K
requests
that have this highest value. The queue is thus reduced to
n
−
K
elements. A
subsequent implementation of the algorithm will reduce the queue to
n
−
K
elements, if other requests have not overlapped in the meantime. ADoRe
essentially requires a less number of runs of the scheduling algorithm than
those required by the RxW. This means a lower load for the CPU and faster
response times to requests (in fact, in this case, data related to sets of
K
requests are searched on a disk, as a consequence the search optimization
on disk is significantly improved, especially when
K
is high). So if
K
= 1,
the ADoRe algorithm is the same as the RxW algorithm. We observe that
a peculiarity of ADoRe is represented by the utilization coefficient of
the disk (among other things, if the number of elements in the queue is
less than
K
, the algorithm sorts and sends requests to the disk by holding it
still busy).
2.2
The Flush algorithm
The good performances of the ADoRe algorithm are due to high disk
utilization and optimization of the scheduling of the server waiting queue.
The key idea for the development of FLUSH is to maximize the
utilization of the disk. Although FLUSH always uses the C-LOOK as disk
scheduling algorithm, it manages differently the user requests to the server
node. Each time the disk ends the service for a single request, all the others
currently on the queue of the server node flow towards the disk system and
are included in its queue, thus becoming subject to the processing of the C-
LOOK algorithm. Then, the FLUSH essentially does not employ scheduling
algorithms for the input queue of the server. The final result is that long
queues are formed in the input of the disk (the so-called scan lists): the C-
LOOK algorithm can optimize the data search on disk; as a matter of fact, it
now has an overview of all data, which it must search on disk, and not just a
partial view of them.
It analyzes the disk from the inside and had learned what are all the data
to be taken: it is like as if at one time it had the chance to collect all
data required, without returning close to the disk axis, with further radial
movements (the reduction of the radial movement is the predominant factor
in the optimization of data search speed on the disk) [7].
