General Repl quest...plus few specific...

[ryan_labarre]

Good morning all. First time poster.
We are a partner using veeam for a while now, and I have a client that is in a tough situation
I am working on their replication from one site to another over 1.5m t1... i know thats pretty slow.

Background - first replication i did with physicals sitting next to each other, then took the remote server to its site.
Seemed like the replications worked ok for a while.
Someone then added that remote esx host in to vcenter, changing vm signatures, and I was forced to rebuild the replications (after long support call with veeam).
So, I did new replications, to disk first, copied them, etc... That seemed to work, although replicating the 2nd time, unberable, 20-30 hours,
Yesterday I took a test machine, moved its pagefile to a new datastore, and removed it from the replication. First time it ran, it took 7 hours (acceptable over that pipe). I wanted to see what would happen immediately after that, where the machine had not changed a bit, with no pagefile on system disk.
It took another 3 hours for a 2nd replication. What gives? NO changes whatsoever had been done to this small 2003 OS.
Clearly, some of the bigger machines like exchange and DB server will have much bigger changes each night. So... that said... is there any way possible over this pipe to get 5 vms replicated?
exchange
db
citrix
citrix
citrix wi

thanks

[Vitaliy S.]

Hello Ryan, and welcome to Veeam community! Could you please clarify what version of ESX hosts your client is currently using. Thank you!

[Gostev]

Ryan, 1.5Mb link is definitely very slow for image-level replication of multiple VMs. Exchange alone will generate at least 10-20GB of disk changes per day due to transaction logs, same with your DB. If increasing the link significantly is not an option, you can try improve current replication times by leveraging WAN accelerators... I would expect 5-10 times improvement based on current metrics other customers provide - you can search the forum more more info and actual customer experiences with this. Thanks!

[ryan_labarre]

esx3.5. customer has a scsi tape device back end mounted prohibiting upgrade to vsphere.
I think even 5-10x wont be enough honestly. I need to check logs, but i thought this was working ok at one time.
and wan accelerator isnt likely an option either. This client could be described most easily by referncing Cameron from Ferris Buellers Day Off

[ryan_labarre]

One thing i didnt get answered.
Why would a replication take 3 hours, even over a slow pipe, for a VM that hasnt changed a BIT....????
I replicated it, when it was successfull, i started another one right away. This is on an un used web interface server with pagefile on a difference DS, etc.

EDIT : would someone please give me a general price point for a WAN accelerator.. THey wont buy it, so general price is fine... I just want to present options.
Thanks

[Gostev]

Why would a replication take 3 hours, even over a slow pipe, for a VM that hasnt changed a BIT....????

In case of ESX 3.5, Veeam Backup needs to actually read full VM to understand that no block has changed - each replication cycle, this is what takes time. This is not the case with ESX4, where ESX itself tracks disk changes by itself with feature known as "ESX4 Changed Block Tracking". For most of our customers, this functionality has been one of the main drivers for vSphere upgrade.

[Gostev]

Unfortunately I don't know the price points, but you can try to research on this yourself. Some WAN acceleration solutions that were mentioned/recommended before on there forums are these (in no particular order):

HyperIP (software appliance)
Cisco WAAS (hardware?)
Riverbed (hardware?)