geekhack
geekhack Community => Off Topic => Topic started by: osi on Tue, 17 December 2013, 09:59:06
-
Guys and Gals,
If anyone has never used the server 2012 deduplication feature, I highly reccomend it. Of course, always suggest following best practices to achieve the greatest gains.
For example, deduplication does not apply to virtual hard drives among other things. For file shares, it's amazing.
[attachimg=1]
Anyone else utilizing this feature out there??
-
Windows as a server. Eww...
-
Windows as a server. Eww...
For domain environments, definitley. We have our fair share of nix boxes and workstations as well
-
Windows as a server. Eww...
For domain environments, definitley. We have our fair share of nix boxes and workstations as well
Ya, I am bias. :) I work with nix servers all the time and I hate it when I have to touch a Windows server. :P
How bad is the performance hit on Windows with dedup enabled? The main issue with dedup usually is that it takes a huge amount of CPU to dedup while the data is being written. It will usually redline your CPU and reduce the IOPs because it can't write as quickly? Are you using it mainly for backup data?
Edit: In regards to dedup. SolidFire is doing some really cool **** with dedup. I know its off topic a bit from your point, but I build cloud computing infrastructure and we have been playing with is a bunch recently. Pretty impressive approach they have...
-
On this particular server, it's just a backup destination dump so the CPU use isn't a big deal.
Also have dedup turned on and active for an accounting server. So far, there have been no issues with CPU time/utilization that would effect business. Windows actually handles it well and it hurt to say that. So no NOTICEABLE load.
Of course you could always schedule your deduping task for off peak hours if it is causing performance issues.
The only other deduping I have got to use were for 'Greenbytes' storage devies. These had a backend powered by BSD and used a ZFS filesystem. Definitley more enterprise class with snapshot support.
Will definitely take a peek at SolidFire...
-
Curious which file types you have in that folder, your de-dupe ratio is ridiculously high.
-
Curious which file types you have in that folder, your de-dupe ratio is ridiculously high.
Its backups, so two backups with be about 95% the same... Dedup on backups is probably the best use case for it. You get great dedupe ratios because most of the backup data is the same between backups.
-
On this particular server, it's just a backup destination dump so the CPU use isn't a big deal.
Also have dedup turned on and active for an accounting server. So far, there have been no issues with CPU time/utilization that would effect business. Windows actually handles it well and it hurt to say that. So no NOTICEABLE load.
Of course you could always schedule your deduping task for off peak hours if it is causing performance issues.
The only other deduping I have got to use were for 'Greenbytes' storage devies. These had a backend powered by BSD and used a ZFS filesystem. Definitley more enterprise class with snapshot support.
Will definitely take a peek at SolidFire...
Ya, I wouldn't try to dedup a webserver, but backup servers can see a real benefit from dedup. Especially because they usually have very little to no load, so the extra resource usage is not an issue at all.
SolidFire is not a system for backups really. They use all SSD drives and they are not cheap. What is very cool about them is that they do dedup natively as well as many other cool things like guaranteed IOPS per mounted volume which is a real problem with shared infra. They have a VERY different approach, but it is pretty cool. But not cheap... Haha...
-
blank
-
Wish I could store my media files with "compression" like that.
-
Well... if you put 47 copies of your favorite Rick Astley tracks in random subfolders and set that to dedup you certainly could. I can see how super amazing this is for enterprise, since moronic people seem to feel a need to copy spreadsheet X to their own user folder so they can have their own independent out of sync useless copy of it instead of using the group share copy they are supposed to :rolleyes: