Salta al contenuto principale

Data Deduplication - Performance Issues

Thread needs solution

Hi People,

Short story:
- I just started using ABR11 build 10.0.13545, data deduplication ENABLED
- First Backup run just fine, 5 boxes recently created (1 phys, 4 VMs) in about one hour. (Occupied space in vault was 70 GB for physical box, and around 80 MB for each VM).
- Now, the second backup is taking ages! Mainly because -I think- of this, new to me [Indexing Task] wich has been running for 17+ hours and and the time remaining reported is 1 day 7 hours :(

My concerns:
Q1
Can I stop the Indexing process, without compromising the validity of the first backup.
Q2
Does it make sense:
A) To cancel my currently executing full bakup tasks (the most advanced one says it´s gonna finish in two days)
B) Given that vaults ARE NOT editable, create another centralize vault with deduplication DISABLED
C) Use the new vault with another backup policy

Please Advice and Thanks in Advance

Arturo ##

PS - Addt´l Info:

The *centralized vault* has the following attributes:
Type: Managed
Path: Resides on an external disk [LaCie d2 Network 2, 2 TB, 1 Gigabit connection]
Deduplication: *ON*
Compression: On
Encryption: Not Encrypted

HW involved:
HP Proliant GL380 G7
16 GB RAM
4 300 GB HDs running under RAID 10

Virtualization SW:
Win 2008 R2 Hyper-V

0 Users found this helpful

Hi Arturo,

Version 11 or 10???

How do you backup your VMs within the guest or on the HyperV Host?
Is the DB and the vault on the same external Lacie disc? Have you tried the local RAID?

Hi Endurance,

THX for taking the time.
It is Version *10* build 10.0.13545

Now, concerning your question: VM bakup at guest or at Hyper-V host.
Answer:
NOt sure abut this one, hope this info helps.
a) I´d rather liked agent-less VMs. As far as I know this can be acomplished, somwhere in the initial installation linkink ABR10 and th Hyper-V. THis was I NOT done. At the installation for selecting component Hyper-V was greyed-out all the time in ABR10 Setup.
I ended nstalling the agent at EVERY VM
(BTW all VM are the same OS that Host box, Win Server 2008 R2 Enterprise Ed.)

b) I guess the DB [additional folder] you refer to, is the one that yore required to create whenever the vault is manabed and deduplication enabled.
Amswer: DB is on HD at the Hyper-V host, so it ends up on the RAID10 array.
Vault stores .META folder and .TIBs

*** UPDATE - What I did ***:
1) I deleted the original backup policie, it´s corresponding tasks were removed automatically also.
2) I kept the vault
2) Created new vault [no dedup] and new backup poliy
I executed the new tasks. It was not *AS QUICK* as the the first run, but transfer rates in the 30-40 Megs a second it´s okay
THe first run had transfer rates of 50+ MB a second. I guess no big heal here.
3) For the sake of starting from a fresh box, I decided to reset the physival box on Friday night. *Bad idea*, I was received with this at the console whe shutting down:
>>
OPERATIONS ARE IN PROGRESS, PLEASE WAIT
THE MACHINE WILL BE TURNED OFF AUTOMATICALLY AFTER THE OPERATIONS ARE COMPLETE.
<<
I guess this is consecuence of the [Indexing activity] I found running.

And the [OPERATIONS ARE IN PROGRESS] message prevents from actually login into the box.
So, Yo could say that I ended up with a server box that´s is taking between 36 to 48 hous to restart.

If by Monday is has not finished, I´ll go for preesing the power button for several seconds.

THX and Regards

Arturo M wrote:
>>
OPERATIONS ARE IN PROGRESS, PLEASE WAIT
THE MACHINE WILL BE TURNED OFF AUTOMATICALLY AFTER THE OPERATIONS ARE COMPLETE.
<<

One of the most anoying things with ABR.

Deduplication can be a nightmare - for me it works with full disc backup. Using it for filebases backups running via agents on VMs it works sometimes and sometimes not. I was not able yet to find the root cause of the slowness since all resources are idleing.

BR

Hi Arturo

ABR10’s Deduplication was a little slow but it’s been addressed in ABR11 which is running much faster!

One important step with ABR10 is to ensure you run your first backup then WAIT for the indexing task to finish BEFORE you run any other server backups. If you have time ran the next server’s back and let that index as well!

This will save you allot of time with indexing, if you backup all your servers at once it’s got a huge amount of data to work through and compare against and you will find you have to wait a long period of time to get the initial indexing completed.

ABR11 also has improved in the fact that uses a new Database that is 64bit, can access more than 2GB RAM and generally performs allot better. However I would still always recommend getting the first few backups done and completed before running other server backups.

I would also recommend having a read through the manual as it points out a good set of tips and back end workings so you can get a better understanding on what is happening.

Please note this is the ABR11 manual but will give you a good idea and I’m sure you will eventually move to ABR11 so might as well use this manual (principals are the same):
http://www.acronis.com/support/documentation/ABR11/index.html#3349.html

It’s also possible especially with ABR10 which is more disk intensive that the DB is using up most of your RAID unit’s I/O and causing performance issues, if you have dedicated HDD’s you can use for the DB with ABR10 you will also notice better performance. ABR11 is allot better and with the 64bit database and extra RAM usage as well it does not need as much disk resources which is great as RAM is so cheap now days.

Look forward to your reply.

THX Endurance for your comments.

The indexing task ended normally after 43 hours for proceesing.
Some numbers:
Occupied Space in Vault
Before After
Physical 70 GB 500 MB
VM 80 MB 60 MB

The saving was spectacular in the case of the Hyper-Vbox, but at I price in time processing way too much high to pay. So I´m discarding Data Deduplication for the time being. I don´t know, maybe in ABR11 I´ll take another look.

Have a nice day Endurance.

##

Datastor,

THX for answering my post.

I check the ABR documentation regularly, and I haven´t seen any comment regarding [ensure you run your first backup then WAIT for the indexing task to finish BEFORE you run any other server backups]. Anyway, this was a good learning experience by doing. Good luck for me, these boxes are not in production yet, and I started the second full backup, five boxes at once on a Friday afternoon.

According to your comments, it makes sense to me looking for the ABR11 upgrade in the future. I guess that SQL 2005 Express @ 32bit ain´t exactly the best companion for ABR10 on Win Server 2008 box.

Thank again for your Help
Kind Regards

##

Before you plan to upgrade wait until the next update (maybe better two updates) of ABR11 are out :(
ABR11 has currently a lot of issues so I cannot recommend an upgrade TODAY.