Is this a solution for me?

[danielefiore]

Hi guys,

Just discovering this new product which seems to be powerful but i still need to understand if it fits my needs.
I would like to get rid of DFS technology to replicate huge amout of data (contained in VMs which are file servers) and replicate the VMs themselves.

Do u see that possible with this product ?

i mean replicate VMs from a branch office to a central location and start the VMs in the central location when the VMs in branch offices are down.

Please i need your feedback i did not find whan i wanted to know in the white papers.

Many thanks in advance,

Daniele

[Gostev]

Hi Daniele, sure - this is exactly what our replication functionality is designed to do. Thanks!

[danielefiore]

Dear Gostev,

Many thanks for your answer but then my questions would be

- using this product over a 1MB WAN link may i'm able to replicate VM with 8 TB of VMDKs ?
- if the VM in brach office is down would the switch to the replicated VM be fast ? we need high availability
- when the VM in brach office comes back how can i replicate the data back to have the VM in brach office back again ?

Many thanks in advance.

Daniele

[Alexey D.]

Hello Daniele,

1) For such a slow connection one can do replica seed (take a look at this post): initial replication is done locally, and only increments are sent through the wire. Also, you may want to use WAN acceleration tool, such as HyperIP, which is free: HyperIP 2Mbs subscriptions are free to Veeam Customers.

2), 3) Take a look at this bunch of topics, replica failover and failback are covered:
Re: DR Site Replication
Re: Replication and failback
Re: How to replicate back from DR to Main Corp Site

[Gostev]

using this product over a 1MB WAN link may i'm able to replicate VM with 8 TB of VMDKs ?

The answer is maybe - depends on the amount of daily changes in virtual disks of those VMs. Generally, these numbers are very far from realistic for most typical workloads. This comes down to pure math - how much data you need to move with incremental runs over the pipe you have.

If these disks have mostly static content (read-only archive), then yes this should be possible.

However, even if 1% of virtual disk blocks change daily, this translates into having to move 80GB of data over 1Mb WAN each day. As you may guess, this is simply impossible even theoretically.

[danielefiore]

Hi Gustev,

If i understood properly : we are goona have VMs as file servers with data that are update every day.
So are you telling me that if the data are updated frequently it is unrealistic to replicate data on remote locations.
Let's talk about a pure backup-restore solution from branch to remote central location : is still that impossible using this product over a 1MB link ?

[Gostev]

It does not matter if it is backup or replication - you need WAN link with enough throughtput to physically be able to move the actual amount of changed data over (to be precise, approximately 30-40% of sheer amount of source VM disk changes on average - because we do dedupe and compress before sending data to target over WAN).

I can say that for most typical VM workload (for close to 100% virtualized shop, where VMs run just about everything), daily incremental backup of 8TB of VM data on 8 hours backup window would definitely require 1Gb LAN connection to backup target storage. And you can do the math from here, but it should be clear that having 1000 times smaller pipe just won't cut it.

[st0623]

Daniele, I'm a Technologist with the HyperIP team at NetEx. We can talk, if you like, about accelerating the bandwidth you have and understanding how much actual application throughput you can get with HyperIP and Veeam Replication. Contact me at steve.thompson@netex.com.