B&R 10 Build 11105 tasks lock and bad index performance on centralized vault with dedupe
Hey friends :)
I am using ABR10 with DeDup on 2 Servers and 2 Workstation for well over a year by now, actually I think it almost two years by now.
After talking to the Acronis Helpdesk in Germany and refitting my Backup Machine my Indexing is done in no time, speedy, reliable
and most important, ALL of my recovery jobs have worked so far since we use Acronis in general.
I am well aware that Indexing for Dedup does cause some trouble if your servers layout is not optimal and/or if you follow the wrong initial
setup procedure, like ridin' it in :)
What you need :
Multi-Core CPU - the more the better , Core i7, Xeon, Q8600 etc..
LOTS of RAM - since RAM is so cheap, dont go below 8GB, I use 12GB
SAS-RAID: Using a Dual Core Adaptec SAS Raid Card 16-Port with Raid-1 for OS Raid-1 for SQL ( NOTHING else on that drive )
and Raid-6 for the Storage Node(s), ALL Drives are SATA with Big Caches, the key is the Controller and not the RPM of the drive.
If you can go SAS 15k for the SQL it will be even faster. ( I might do that since I have 6 x 146GB 15k lying around )
We had lotsa speed problems ( index took days and days ) when the SQL was not in a seperate Raid-1 Hardware Raid.
OS is SRV2008R2, Asus WS-Board, Adaptec Raid
One thing to mention, I am NOT using SQL 2005 Express, I opted to install the newer SQL2008R2 Express since it better utilizes the HW, Cores and RAM.
I have around 250 Backups in my vault, my dedup saves me TONS of Terabytes that I never needed to purchase or validate.
Follow the advice and Dedup will work very speedy ( the first one my take 2-6h - yes agreed ) and save you TONS of space.
The more servers you backup the better dedup plays its roll.
* De-Install the whole thing -- REBOOT, install SQL2008R2, patch it, reboot -- then install ABR10 with SQL2008R2 as SQL instances and do ONE backup of ONE template machine, index it and THEN you can backup any other machine in any sequence or all together.
It worked for me and I was also about to kick Dedup to the Moon when i first got it. Now, man I LOVE IT ! It`s worth spending the time gettin' it to work
Lars
what I want to stress out here again since I am just redoing the setup on a 2 x notebook ( hardcore gamer laptops ) with a 3-TB USB30 external drive as the node and the SQL on the 2nd internal sata SSH drive, sql 2008R2, 16GB in my laptop, Core-i7 740QM, 7-Ult-x64.
...and its soooo slow since my laptop's SSH drive sux big time. That Seagate Momentus XT is BS, especially with high I/O.
I am really thinking of getting a Vertex2 of payable size and retest it with the SQL on OCZ Vertex2 EX 240GB.
When I look at my performance meter I can clearly see the bottleneck being the SQL Database I/O count and wait-time, plus
on my laptop it must share the drive with other data that is being monitored, shared, etc... and thus SQL doesnt have exclusive I/O
on it.
Since this was for testing only I can just say "what the heck, I actually knew before that indexing would be the killer app"
So, dont share the SQL harddrives with anything else, just SQL
if possible dont use USB20/USB30 for your node/sql drives, that protocol aint made for high & random I/O pal !
as I said, I knew before, just wanted to let you know that indexing is a very high load I/O process that will not tolerate
any mis-configurations in your data-path layout.