| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 462.1 | | EEMELI::MOSER | Orienteers do it in the bush... | Mon Apr 14 1997 10:18 | 6 |
| >On V6.0 the size of the converted file is 40% greater than that on 6.2
Looks like a good improvement to me, maybe V6.2 is doing a better job
than V6.0
/cmos
|
| 462.2 | | EPS::VANDENHEUVEL | Hein | Mon Apr 14 1997 12:19 | 20 |
| > On V6.0 the size of the converted file is 40% greater than that on 6.2.
> That is, on 6.2 CONVERT produced a smaller file from same source.
> Is this the expected behavior ?
Likely to be a side-effect. My guess would be a different CLUSTER_SIZE
on the target disk. You need to provide more details on the convert
command used to understand what is happening. Specifically you need to
explain whether using the 'real' convert code through the default
CONV/FAST or plain RMS when CONV/NOFAST is chosen. Are the FDL specs
used exactly the same? (they _are_ using an FDL I hope!). Is the
allocation/extend specified in the FDL adequate or is the output
growing during the convert? Are COMPRESSION and BUCKETSIZE explicitly
requested in the FDL or defaulted? There really were as many records?
If you still can not figure it out, then you may need to post the
results from ANAL/RMS/STAT for the files both AFTER a 6.0 convert
and after a 6.2 convert.
2�,
Hein.
|
| 462.3 | CLUSTERSIZE is different on target disks | TAV02::ZVI_P | Here we are | Tue Apr 15 1997 03:52 | 13 |
|
It seems CLUSTERSIZE issue.
The disk that produces the smaller file has smaller CLUSTERSIZE - 3
the other disk has CLUSTERSIZE of 9.
All other FDL issues are the same. He did not touch them.
He uses it to play with the INDEX.
Thanks
Zvi
|
| 462.4 | | EPS::VANDENHEUVEL | Hein | Tue Apr 15 1997 16:26 | 13 |
|
> It seems CLUSTERSIZE issue.
> The disk that produces the smaller file has smaller CLUSTERSIZE - 3
How many blocks smaller is 40 % smaller? a few blocks can easily
be explained. ten-of-thousands of blocks difference would suggest
that the allocation and extent in the FDL is whoefully inadequate
as RMS aligns buckets on cluster bounderies when extending.
hth,
Hein.
|