Slow transfer to cloud provider

[khughes]

We're transitioning to backup to a cloud based service provider but it seems like our backup jobs are going slow even for WAN speeds. We have a 100Mbps fiber line which I know isn't the fastest but its also not slow. Our first replication job was a 2.6TB virtual machine that took about 50 hours to complete (avg 11MB/s). The load metrics were: Source 55% > Proxy 33% > Network 46% > Target 59%. I know by those numbers the Target was the bottleneck listed but I'm trying to get the remaining VM's seeded over this long weekend but I'm getting really slow speeds on the current job (8MB/s) and at that rate it will take forever.

There is only one replication job running at a time. Looking at the physical server performing the backup, there isn't any load showing on the NICs which I would've expected. Looking at other posts a typical question was if the use multiple streams was checked, and it is with the default value of 5. As far as I know none of our equipment has QoS settings enabled.

This is our first time trying to replicate out of our internal data center so if I'm overlooking something obvious I apologize but I'm just trying to get the speeds up for the remaining seed jobs as well as down the road doing the incremental backup jobs so any help would be appreciated.

Thanks

[khughes]

One small update, it looks like it's using our secondary veeam proxy server to process the backup job as that's where I'm seeing the network load and the processing load. Should that matter which Proxy is being used if they are both on the same network and both are directly attached to the SAN?

[foggy]

Hi Kyle, the speed looks expected for the 100Mbps link (which is 12.5MB/s). Haven't you thought of seeding the initial sync by physically sending the copy offsite and then mapping the job to it? Further incremental runs should copy changes only so will be faster.

[oliverL]

May i ask what have driven you to choose cloud-backup since it takes so long and what you will do when you need a full-backup back from the cloud?

[khughes]

Thanks for the validation that things look as expected. As for sending the physical copy of the files as the initial seed, it was a thought but we wanted to see how the link would work first and these were the first couple jobs. I wanted to be sure we weren't doing anything stupid on our end that may have been causing the slow network speeds. I was messing around with the proxy selection last night and was forcing it to use the server that was performing the replication and not sure it helped but things did speed up slightly.

11/24/2016 6:24:11 AM :: Load: Source 32% > Proxy 32% > Network 74% > Target 21% (processing rate of 25MB/s)
11/24/2016 7:35:40 AM :: Load: Source 13% > Proxy 27% > Network 92% > Target 13% (processing rate of 21MB/s)

I figured the initial seed would be the most painful part and the incrementals would be a quick job. At this rate we should be able to get through the remaining VM's over the weekend.

oliverL - to your question, this is strictly for a DRaaS event. We have always put backups on LTO tapes, but trying to improve our DR stance. We will still continue to do local backups to our fiber channel SAN which as you can imagine goes quite a bit faster.

Thanks everyone for the thoughts/help.