Hi Commvault people,
I am looking for feedback on Backup Copy throughput.
We have a fairly new Commvault-AWS implementation.
We have a few different DSP’s, with clients spread across several of the DSP’s.
There is only one Media Agent, but we use Proxy’s in the other DSP’s.
For example DSP1 has the MA and it’s attached library.
DSP2 has a bunch of servers but only a VSA Proxy. Same for DSP 3 and 4 - clients and a VSA proxy but no Media Agent.
So what we basically have is a cross-DSP configuration.
The main media agent in DSP1 hosts the library.
Clients are snapped in all of the other DSP’s using the VSa proxy who then streams the conetnts to the MA and it’s associated library.
This is fully supportd and it seems to work well.
However, I have noticed that although the Snap works fairly quickly (15 mins for a few servers), the Backup Copy is slower. It will take a couple of hours to transfer a few hundred GB, averaging only 330ish GB/hr.
So my question:-
Is anyone else a cross-DSP backup solution and what rates of throughput are you seeing on your Backup Copy?
I am hoping to avoid dropping in any additional hardware (Media Agents), but I am concerned that as we add a lot more workloads into the DSP’s, we will run into problems with slow Backup Copies, which therefore introduces risk to the customer.
AWS - Cross-DPS Backup Copies (using IntelliSnap)
- Apprentice
- Vaulter
Hi MountainGoat,
Are you able to clarify what you mean by DSP? The only reference I can find for DSP in AWS refers to an advertising platform “demand-side platform”.
To more directly address your concerns, there will be a difference between snap and backup copy speeds because, a snap backup is largely an indexed AWS snapshot. A backup copy, uses the AWS snapshot to read and protect the EC2 instance as if it was a streaming backup.
I would also check if EBS Direct API is used, you can validate this by checking vsbkp.log on your access node for a backup copy job and looking for EBS and it should mention if EBS direct API was used.
You should also validate all of the recommendations here are met: https://documentation.commvault.com/11.40/software/software_hardware_and_other_requirements_for_protecting_amazon_ec2_resources.html
If all of the above is optimal, I’d raise a support incident to investigate because 330GB/hr definitely isn’t performing to expectations.
Cheers,
- Apprentice
Hi Jace,
Thanks for the reply.
The DSP’s in AWS are actually Digital Service Platforms, and not Demand Side Platforms.
Although it interesting to see Demand Side Platform plastered all over the internet (something not right with the Internet indexing algorithm perhaps!).
So basically, they are like separate environment within AWS, like one for TEST, one for DEV, one for Project sandpits.
In my specific example, DSP Prod hosts the MA, but we only have a proxy in DSP Test.
There is inter-DSP connectivity which works well, apart from my Backup Copy throughput.
If possible, I want to avoid needing an MA in DSP-TEST, which is why I’m trying to guage what other peoples Backup Copy throughput rates are.
I will check if the EBD Direct API has been implemented (I thought it had been, but will double check).
Regards.
- Vaulter
Hi MountainGoat,
There’s just too many acronyms these days.
A DSP shouldn’t impact performance unless it’s cross regional. I would also try other data transfer operations between DSPs and see if there’s any performance issues. Even a SMB/FTP file transfer between EC2 instances in different DSPs should provide some idea of whether a DSP could impact performance.
Cheers,
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.