Prorest Performance

Delo123

New Member
I run following task to restore our DB each night for our Datawarehouse

It is apprx 70Gb and takes 1.40Hours from our SAN

call "c:\Program Files\Progress\bin\prorest" training \\dc01-grnpwr\backupMOS\mos.bck

I noticed the restore routine only uses one physical CPU Core, is it somehow possible to use multiple cores? or multiple threads??

I deeply remember there was some kind of switch/parameter...?

Somebody now how to speed up?

Need to do 4 restores each night, and running out of time since the db keeps growing...

Thanks!
 
Each instance of prorest is aggressively single-threaded.

But you could run one for each database in parallel.
 
But I need to restore the same Database multiple times.

No other options?
Some way around?

I also don`t understand why Progress locks the Backup file, it doesn't have to write, right?
 
Locking the backup file is a Windows bug. It doesn't get locked on UNIX. You might try reporting it and complaining to tech support.

I suppose you might try making extra copies of the backup using OS tools and then restoring from those. If you have mirrored drives you could un-mirror them and double your throughput, rename one set of restores and then re-sync.

Or maybe using a something similar to a UNIX link might fool the file lock?

Are you deleting the target db prior to restoring? Or are you overwriting it? (Overwriting would be faster since it skips the "prostcrt create" phase.)

Just curious but why are you restoring 4 copies of the same db onto the same server?
 
Hi Tom,

Actually I am restoring to 4 Different Servers (Workstations actually)
After the Restore we do Dbtool (due to field length) and then mount each database and let SAS pull the Data out. The SAS Task used to take 12 Hours on one Workstation, so now we splitted it into 4.
Then again now they all do a restore one after another which takes a lot of time also...
Also, when something goes wrong on the first workstation, the other 3 will not do their tasks also...
At the moment i am using schtasks.exe to remotly start schedules tasks on the next machines so i don`t have extra waiting time.
So. Step 1. 23.00 we do a online Backup, this is finished around 00.50 (Oh, sometimes it can take appr. 1 Hour extra but usually no users are logged in, this wonders me) and since the time differs, in the same batch when the backup is finished i use schtasks to start the restore on the first worstation, then the 2nd etc....
Prostrct is taking care of earlier each day, so the databases are empty and prepared...

It all isn't pretty, but I haven't found a better process....
 
What is your Progress version?
We restore 35 GB database in about 10-15 minutes...
Are you sure you haven't got a disk issue (raid5 on SAN can be problematic to with such actions).
Do you use typeII storage area's, this can aprove perfromance of extracting data with SQL enormously. Especially if you pull of entire tables from the database. 12 hours seems a lot to me for pulling the data from the database. Do you use complex queries to extract the data? Do you run update statistics (table,index, column) on a regular basis on that database?

Casper.
 
Progress 9.1D (Sadly enough unable to update at this Time, because the Main Application is compiled in 9 and do not have the Source Code).
Data is pumped over LAN to the Workstation (1gb lan). I will try to hook a workstation up with a HBA and see if this improves performance.
Do you use complex queries to extract the data? Do you run update statistics (table,index, column) on a regular basis on that database?
Yes, the Queries are quite complex. SAS needs this time, and has everything to do with CPU Power and MEM. Currently running 2 Nehalems 3.2Ghz with 64GB RAM running XP 64Bit. Nothing much more i can do there...
The Gain i am looking for is in the restore part.

What I was thinking of:
Sunday:
Do a Full online Backup (23.00)
Monday:
Restore this Backup Monday. (17.00)
After working hours do a After Image Backup.(23.00)
Roll Forward the restored database with this AI(23.15)
Do a full online backup.(23.30??)
Monday night do the SAS tasks.
Tuesday:
Restore the Full Backup from Monday 23.30 (17.00)
Ai
etc.....
This way i hope i can win the time to do full backup and full restore when instead the machine could do something usefull!

Ps. We are not using ai at the moment, but i suppose backup and roll forward will be quicker then a full backup/restore right??

Hope it makes sense......
 
Back
Top