IBM 1130 User Manual page 599

External Sorting Methods
When a file cannot be contained within core storage,
additional external passes and intermediate storage
devices, such as disks, are required to sort the
file. The internal sort, then, is only one phase of
a generalized multiphase (or multipass) sort that
may have three or four phases. In such a multi-
phase sort, the internal sort phase is concerned
with the creation of suitable sequences from the
main file, while the external sort, or merge phases,
are devoted to the reduction of those sequences to
one continuous sequence.
Practically all the internal sorting techniques
described earlier can also be used -- with varying
success -- for external sorting by changing the
terms of reference appropriately. Thus, the inter-
nal storage area is replaced by several input and
output areas on disk.
It has been suggested earlier that sorting tech-
niques could be categorized according to the degree
of and relative need for data accessibility. Thus
far, sorting techniques suitable for one extreme of
data accessibility have been described. The inter-
nal sorts were seen to be best suited to high-speed,
direct (random) access storage media, such as
magnetic core storage.
In
these media, any record
or string of records can be accessed immediately,
without the need for passing over other, unwanted
Section
Subsections
Page
75
30
I
20 01
records.
Despite their name, direct (random) access file
storage media (such as disks) provide a degree of
data accessibility less than core storage. The time
to access a record in these devices is not com-
pletely independent of the location of the previously
accessed record (as in core storage), but neither
does it depend on the entire sequence of records
stored before it (as in magnetic tape). The time to
access the next record depends on the number of
cylinders the access mechanism must be moved
from the previous record. (Each one or two cyl-
inder move on the 2310 disk drive requires 15
milliseconds.) However, since internal core
storage is generally insufficient to hold an entire
file, auxiliary storage devices such as disks are
usually necessary.
With disks some attention must be given to the
relative advantages of key or tag sorting and sort-
ing of complete records. It has been found in inter-
nal sorting that key or tag sorting (involving either
record numbers only or short control records) is
conSiderably faster than sorting of complete re-
cords. However, because of the substantial seek
time, this is no longer true for disks, when the
orginal records must be retrieved at the end of the
sort.
The follOWing paragraphs explore some of the
considerations pertinent to disk sorting.