Comprehensive data protection for all workloads
Post Reply
brandonfro
Novice
Posts: 4
Liked: never
Joined: May 26, 2011 8:33 pm
Full Name: Brandon Froehlich
Contact:

v6 Off-site Replication

Post by brandonfro » Oct 12, 2011 3:57 pm

I'm excited for the new version of Veeam, but I have a question. Will I gain any performance benefit by installing Veeam at the target site and "pulling" replicas from the source? Or is it comparable to have Veeam at the source and "push" replicas to the target?

Thanks,
Brandon

Vitaliy S.
Product Manager
Posts: 22772
Liked: 1524 times
Joined: Mar 30, 2009 9:13 am
Full Name: Vitaliy Safarov
Contact:

Re: v6 Off-site Replication

Post by Vitaliy S. » Oct 12, 2011 4:24 pm

Hi Brandon,

With version 6 Veeam Backup server placement (production or DR site) no longer matters, so notions such as push/pull replication, or ESX/ESXi target differences are no more. You just need to dedicate at least one proxy server per site.

Thanks!

brandonfro
Novice
Posts: 4
Liked: never
Joined: May 26, 2011 8:33 pm
Full Name: Brandon Froehlich
Contact:

Re: v6 Off-site Replication

Post by brandonfro » Oct 12, 2011 4:30 pm

Great! Also with v6, if a disaster scenario actually happens, I understand that the vSphere Client can choose which restore point to load, but if I wanted to utilize the failback options, I would have to use Veeam. My question with that is, can I use a new installation of Veeam to do that, or does it have to have some SQL data?

foggy
Veeam Software
Posts: 18030
Liked: 1533 times
Joined: Jul 11, 2011 10:22 am
Full Name: Alexander Fogelson
Contact:

Re: v6 Off-site Replication

Post by foggy » Oct 12, 2011 4:54 pm

The same SQL DB is required. You can install a new instance of Veeam B&R and point it to the same SQL Server.

brandonfro
Novice
Posts: 4
Liked: never
Joined: May 26, 2011 8:33 pm
Full Name: Brandon Froehlich
Contact:

Re: v6 Off-site Replication

Post by brandonfro » Oct 12, 2011 5:03 pm

Would that have a negative impact on performance since the source and target are connected via a 10 Mb IPSec VPN tunnel?

Vitaliy S.
Product Manager
Posts: 22772
Liked: 1524 times
Joined: Mar 30, 2009 9:13 am
Full Name: Vitaliy Safarov
Contact:

Re: v6 Off-site Replication

Post by Vitaliy S. » Oct 12, 2011 7:35 pm

Remote SQL Server doesn't have any impact on job performance rates.

Nikolay
Influencer
Posts: 18
Liked: never
Joined: Dec 01, 2012 9:14 am
Full Name: Nikolay Evdokimov
Contact:

[MERGED] Push vs Pull Replication

Post by Nikolay » Dec 04, 2012 7:52 am

Hello,

We have a classic DR scheme: 1 production site and 1 active DR site (with local DCs and Exchange 2010 servers).
Currently we have Veeam B&R 6.5 installed in both sites (separate SQL database for each installation).
So, I'm curious, is there any significant difference or any advantages when replicating virtual server from Prod site to DR site using B&R installation in DR site (in other words I'm pulling virtual machine from Prod site) versus replicating virtual server from Prod site to DR site using B&R installation in Prod site (pushing virtual machine to DR site)?

After I made some tests with replication job (from Prod and DR site B&R installations) I've got the following results:
1. For both jobs replication proccessing rate is roughly the same. With test virtual machine size of 65GB at the end of initial replication job it shows Transferred 9.9GB (6.5x). Does it mean that compression for that machine is as high as 6.5 times?

2. Am I correctly understand that with Push replication (when using B&R local server in relation to ESXi server holding Prod virtual machine which replicates) all the processor workload of replication job is done by this local B&R server and virtual machine disk blocks are transferred at LAN speed between ESXi server and B&R server where these data blocks are being compressed before transferring via WAN link to DR ESXi server?
And with Pull replication (when using B&R server at DR site which is remote in relation to ESXi server holding Prod virtual machine which replicates) all of the virtual machine hard drive's data blocks are transferred without compression to DR B&R server across WAN connection and only after that this DR B&R server compresses data and transfers it to local DR ESXi server?

Thanks,
Nikolay

Vitaliy S.
Product Manager
Posts: 22772
Liked: 1524 times
Joined: Mar 30, 2009 9:13 am
Full Name: Vitaliy Safarov
Contact:

Re: v6 Off-site Replication

Post by Vitaliy S. » Dec 04, 2012 8:21 am

Hello Nikolay,

There is no difference in push/pull replication jobs in terms of performance rates, however if you happen to lose your primary site with the backup server, it would be better to have a standby server deployed on the DR location to perform quick recovery actions.
Nikolay wrote:Does it mean that compression for that machine is as high as 6.5 times?
Not exactly, please keep in mind that your source VM disks might have zero blocks, white space that has been skipped by the job, so only blocks with data has been compressed and transferred to the target.
Nikolay wrote:Am I correctly understand that with Push replication (when using B&R local server in relation to ESXi server holding Prod virtual machine which replicates) all the processor workload of replication job is done by this local B&R server and virtual machine disk blocks are transferred at LAN speed between ESXi server and B&R server where these data blocks are being compressed before transferring via WAN link to DR ESXi server?
Yes, each backup server has a default proxy server which does all the job. The VM traffic does not always go through the LAN segment. If you have configured your proxy servers to work in HotAdd or direct SAN mode then VMs blocks are retreived by the source proxy through either ESX(i) I/O stack or directly from the storage device.
Nikolay wrote:And with Pull replication (when using B&R server at DR site which is remote in relation to ESXi server holding Prod virtual machine which replicates) all of the virtual machine hard drive's data blocks are transferred without compression to DR B&R server across WAN connection and only after that this DR B&R server compresses data and transfers it to local DR ESXi server?
This design is not very efficient, as for WAN replication it's recommended to use two backup proxy servers - one on the source site to retrieve VM blocks, compress them and then send this data to the backup proxy one on the remote site to decompress VM blocks and place them on the DR ESX(i) host .

Thanks!

Nikolay
Influencer
Posts: 18
Liked: never
Joined: Dec 01, 2012 9:14 am
Full Name: Nikolay Evdokimov
Contact:

Re: v6 Off-site Replication

Post by Nikolay » Dec 04, 2012 10:37 am

Thank you for you reply, Vitaliy!

But I still do not fully understand why there us no difference between push/pull scenarios.

In "push" scenario the data will be firstly compressed by the local B&R proxy server and after that will be transferred to destination DR ESXi host where the data will be decompressed and added to the replica hard drive. So, in this scenario the data transfers via WAN link in compressed way.
In "pull" scenario the data will be firstly transferred from Prod ESXi server to DR B&R server, after that it will be compressed there and only after that it will be decompressed (by proxy or by ESXi host?) and added to the DR replica virtual machine. So, in this scenario the data transfers via WAN link decompressed.
Please correct me where I'm wrong.
This design is not very efficient, as for WAN replication it's recommended to use two backup proxy servers - one on the source site to retrieve VM blocks, compress them and then send this data to the backup proxy one on the remote site to decompress VM blocks and place them on the DR ESX(i) host .
This sounds very reasonable for me. Could you please give me some instruction on how to use DR B&R proxy as target proxy for Prod B&R server.
Not exactly, please keep in mind that your source VM disks might have zero blocks, white space that has been skipped by the job, so only blocks with data has been compressed and transferred to the target.
OK, very good, so in any way I can say that transferred amount of data via WAN link is no more that this transferred value? And these zero block are determined by vmware API, not by veeam proxy?

Thank you,
Nikolay

veremin
Product Manager
Posts: 16692
Liked: 1394 times
Joined: Oct 26, 2012 3:28 pm
Full Name: Vladimir Eremin
Contact:

Re: v6 Off-site Replication

Post by veremin » Dec 04, 2012 12:08 pm

This sounds very reasonable for me. Could you please give me some instruction on how to use DR B&R proxy as target proxy for Prod B&R server.
Even though I’m not Vitaliy, I believe I’m able to clarify some moments related to replication.

As Vitaliy’ve already mentioned, for such scenario it’s recommended that you allocate one proxy in your source site and one proxy server in target site. Thus, firstly, all of the necessary data will be received by source proxy, where it’ll be compressed and passed to target proxy via WAN. Secondly, target proxy will decompress it and add to the DR replica virtual machine. Thereby, in case of you having one proxy in Prod site and one in DR site, there is no difference between ”push” and “pull” scenarios.

Furthermore, there are no special requirements - the role of proxy can be assigned to each machine that can be reached by Veaam B&R Server. So, if there is no connection problem between Production and DR site, you’ll be able to do it easily through Backup Proxy Wizard (GUI: Backup Infrastructure - > Backup Proxies -> right-click -> Add Backup Proxy).

Thanks.

Vitaliy S.
Product Manager
Posts: 22772
Liked: 1524 times
Joined: Mar 30, 2009 9:13 am
Full Name: Vitaliy Safarov
Contact:

Re: v6 Off-site Replication

Post by Vitaliy S. » Dec 04, 2012 1:16 pm

Nikolay wrote:But I still do not fully understand why there us no difference between push/pull scenarios.
In both situations you will have two proxy servers, one on the source site, another one on the target one, so dataflow will be the same.

Post Reply

Who is online

Users browsing this forum: apopkov, J.Giroux and 61 guests