on 05-19-2010 11:22 AM
Hello,
due to the fast development of processors and RAM I have to think about migrating our big SAP Systems from Solaris to Linux. The problem is the downtime. With heterogenous system copy I can do a copy of 1 TB database within 24 hours (using options like table splitting, 70 GB cache_size, 12 parallel processes and parallel exporting and importing).
A homogenous system copy needs a downtime of about 30 minutes (full backup, inkremental backup, log-files). So it would be fine if I could change plattform this way.
Does anyone know an "endian converter"?
I was about to write a development request to bc-db-sdb, but it is not possible anymore. Would be nice to use backup/restore to change endian type.
Does the development team of MaxDB think about a "Endian Converter"?
Best Regards
Andreas
> due to the fast development of processors and RAM I have to think about migrating our big SAP Systems from Solaris to Linux. The problem is the downtime. With heterogenous system copy I can do a copy of 1 TB database within 24 hours (using options like table splitting, 70 GB cache_size, 12 parallel processes and parallel exporting and importing).
This seems to be pretty slow to me. Did you try to use prefetching to read the data faster? See Note 1327874 - FAQ: SAP MaxDB Read Ahead/Prefetch
> Does anyone know an "endian converter"?
Haven't tried this yet but you can try to do a remote copy using rsh and dd - and let dd change the byte order
# man dd
<...>
conv=value[,value...]
<...>
swab Swaps every pair of input bytes. If the
current input record is an odd number of
bytes, the last byte in the input record
is ignored.
<...>
So you could try like (on the source system):
mknod /tmp/pipe1 p
mknod /tmp/pipe2 p
Then start a media restore from a device that was created.
On the target system do the same:
mknod /tmp/pipe1 p
mknod /tmp/pipe2 p
start a backup to those pipes
then start two shells doing
dd if=/tmp/pipe1 conf=swab bs=32768 | rsh <target system> dd obs=32768 of=/tmp/pipe1
As said, haven't tried this but this sounds interesting. If that doesn't work one could write a simply byte order changer in C.
Markus
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
User | Count |
---|---|
91 | |
10 | |
10 | |
9 | |
9 | |
7 | |
6 | |
5 | |
5 | |
4 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.