-
- Enthusiast
- Posts: 96
- Liked: 25 times
- Joined: Aug 16, 2013 5:44 pm
- Full Name: Matt B
Moved Veeam to VM = 60% Better Performance
This is more of an informational post than a question or request. Just thought I would share...
At a remote site we have a single Dell R620 vSphere host being backed up to local storage on a Dell NX3200 NAS over GbE with Veeam installed directly on the NAS (Server 2008 R2 Storage). Backups have been running just fine for almost a year but after seeing better backup performance at a different remote site (with a Dell R610 host to a 4-bay Synology NAS over GbE), I knew something was wrong.
I created a VM on the host (2008 R2), installed Veeam, created a CIFS share on the NAS for the repository and then ran a few backups. The throughput is now ~60% higher and almost saturating the GbE connection, so it is running "correctly" in my mind. Also, the Source > Proxy > Network > Target loads seem to be in much better balance as well. So I wonder why the Source was such a bottleneck before when the proxy was the NAS?
Active Full - Before (NAS Proxy)
Active Full - After (VM Proxy)
-----------------------------------------------------------------------------------------------------------------------------------------------------
Incremental - Before (NAS proxy)
Incremental - After (VM Proxy)
At a remote site we have a single Dell R620 vSphere host being backed up to local storage on a Dell NX3200 NAS over GbE with Veeam installed directly on the NAS (Server 2008 R2 Storage). Backups have been running just fine for almost a year but after seeing better backup performance at a different remote site (with a Dell R610 host to a 4-bay Synology NAS over GbE), I knew something was wrong.
I created a VM on the host (2008 R2), installed Veeam, created a CIFS share on the NAS for the repository and then ran a few backups. The throughput is now ~60% higher and almost saturating the GbE connection, so it is running "correctly" in my mind. Also, the Source > Proxy > Network > Target loads seem to be in much better balance as well. So I wonder why the Source was such a bottleneck before when the proxy was the NAS?
Active Full - Before (NAS Proxy)
Active Full - After (VM Proxy)
-----------------------------------------------------------------------------------------------------------------------------------------------------
Incremental - Before (NAS proxy)
Incremental - After (VM Proxy)
-
- VP, Product Management
- Posts: 6035
- Liked: 2860 times
- Joined: Jun 05, 2009 12:57 pm
- Full Name: Tom Sightler
- Contact:
Re: Moved Veeam to VM = 60% Better Performance
Without more information I can only guess that this is due to the fact that installing Veeam in a VM allowed you to use Hotadd mode to retrieve the source data where as previously you were probably using NBD mode. This is almost certainly confirmed by the fact that the incremental, while showing a higher throughput number, actually took longer to backup less data as hotadd takes longer to setup and tear down. Unfortunately I can't tell for sure with the screenshots provided.
You could probably achieve even better performance by adding the NAS directly as a repository instead of writing to it via CIFS, although it might not be enough to matter since you're already saturating the 1GbE link.
You could probably achieve even better performance by adding the NAS directly as a repository instead of writing to it via CIFS, although it might not be enough to matter since you're already saturating the 1GbE link.
-
- Product Manager
- Posts: 20406
- Liked: 2298 times
- Joined: Oct 26, 2012 3:28 pm
- Full Name: Vladimir Eremin
- Contact:
Re: Moved Veeam to VM = 60% Better Performance
I side with Tom that the most obvious reason is that virtual backup server made it possible to use Hot ADD proxy mode. So, can you provide information about used proxy mode? It can be found in the corresponding job session statistics (look for special [san], [hot-add], [nbd] metrics). Thanks.
-
- Enthusiast
- Posts: 96
- Liked: 25 times
- Joined: Aug 16, 2013 5:44 pm
- Full Name: Matt B
Re: Moved Veeam to VM = 60% Better Performance
Correct. I verified that the backups were using hotadd mode on the VM and nbd mode on the NAS. Does the "Virtual Appliance" option in the proxy transport settings mean it will use hotadd? (basically are they synonymous?)tsightler wrote:I can only guess that this is due to the fact that installing Veeam in a VM allowed you to use Hotadd mode to retrieve the source data where as previously you were probably using NBD mode.
Like make the NAS an iSCSI target and mount it as a local drive for the Veeam repository? I will try that tonight.tsightler wrote:You could probably achieve even better performance by adding the NAS directly as a repository instead of writing to it via CIFS, although it might not be enough to matter since you're already saturating the 1GbE link.
On a side note - I know stuff like this isn't exciting enterprise-level stuff, but your continued support of us "little guys" (SMB) is very appreciated. As some of us advance into bigger companies with larger networks our experience with Veeam will not be forgotten.
-
- VP, Product Management
- Posts: 6035
- Liked: 2860 times
- Joined: Jun 05, 2009 12:57 pm
- Full Name: Tom Sightler
- Contact:
Re: Moved Veeam to VM = 60% Better Performance
Actually, I meant adding the NAS as a managed server to Veeam and then configuring it as a traditional repository. If the system was capable of running the full Veeam install, then certainly it can run just the repository role. This would allow the proxy to read data via hotadd, compress and dedupe it, and send it as a stream to the repository agent running on the NAS, which would then perform all of the file I/O locally instead of the proxy having to talk CIFS to handle all of the I/O over the network. For full backups or forward incremental I wouldn't expect much difference, but it should help a good bit for reverse incremental or synthetic full processing if you are using those options since all of the I/O will stay local to the NAS.sbbots wrote:Like make the NAS an iSCSI target and mount it as a local drive for the Veeam repository? I will try that tonight.
-
- Product Manager
- Posts: 20406
- Liked: 2298 times
- Joined: Oct 26, 2012 3:28 pm
- Full Name: Vladimir Eremin
- Contact:
Re: Moved Veeam to VM = 60% Better Performance
And answering your first question, yes, Virtual Appliance proxy mode means Hot Add. Thanks.
-
- Enthusiast
- Posts: 96
- Liked: 25 times
- Joined: Aug 16, 2013 5:44 pm
- Full Name: Matt B
Re: Moved Veeam to VM = 60% Better Performance
Done and boy what a difference in performance. Adding the NAS as a managed server easily outperformed any other setup. Thanks again for the help with my little experiment. It was a good lesson in the performance benefits of different setups.tsightler wrote:Actually, I meant adding the NAS as a managed server to Veeam and then configuring it as a traditional repository.
NAS as managed server
NAS as iSCSI target
NAS as CIFS share
NAS with Veeam running natively
-
- Chief Product Officer
- Posts: 31806
- Liked: 7300 times
- Joined: Jan 01, 2006 1:01 am
- Location: Baar, Switzerland
- Contact:
Re: Moved Veeam to VM = 60% Better Performance
Just a small note - you could have kept Veeam install on your NAS box, and just add new hot add proxy. This would allow you to achieve the same result with less steps. Thanks!
-
- VP, Product Management
- Posts: 6035
- Liked: 2860 times
- Joined: Jun 05, 2009 12:57 pm
- Full Name: Tom Sightler
- Contact:
Re: Moved Veeam to VM = 60% Better Performance
Glad to see the improvement, and Gostev's point is excellent, it really doesn't matter where Veeam itself is installed, it matters where the proxies and repositories are installed. When you install Veeam initially by default that server takes on all of these roles, but with the distributed architecture you can add/move those roles as appropriate for the environment. One of the best things about Veeam is that you can install it and "It just works!" but as you begin to understand the architecture and how it interacts with your environment, you can size and place the components appropriately and "It just works better!" Looks like you're pretty close to the optimal design for this specific environment.
-
- Enthusiast
- Posts: 96
- Liked: 25 times
- Joined: Aug 16, 2013 5:44 pm
- Full Name: Matt B
Re: Moved Veeam to VM = 60% Better Performance
Wow, don't I feel stupid.Gostev wrote:Just a small note - you could have kept Veeam install on your NAS box, and just add new hot add proxy. This would allow you to achieve the same result with less steps. Thanks!
Tell you what Mr. Gostev, maybe I will sign-up for your VeeamOn 2014 (since it takes place in my state), take the VMCE classes while I am there and get certified
-
- Enthusiast
- Posts: 96
- Liked: 25 times
- Joined: Aug 16, 2013 5:44 pm
- Full Name: Matt B
Re: Moved Veeam to VM = 60% Better Performance
I definitely see how the components are distributed now and I appreciate the Veeam 101 training these forums provide. I am still new to everything IT (was forced into the role last year) but I am learning. Dumping Arcserve and buying Veeam was the first (and best) decision I made. Thanks for the help.tsightler wrote:Glad to see the improvement, and Gostev's point is excellent, it really doesn't matter where Veeam itself is installed, it matters where the proxies and repositories are installed. When you install Veeam initially by default that server takes on all of these roles, but with the distributed architecture you can add/move those roles as appropriate for the environment. One of the best things about Veeam is that you can install it and "It just works!" but as you begin to understand the architecture and how it interacts with your environment, you can size and place the components appropriately and "It just works better!" Looks like you're pretty close to the optimal design for this specific environment.
Who is online
Users browsing this forum: AM1 and 66 guests