Network bottleneck showing as >95% + Slow Processing

[BisBestow]

We have just purchases a Synology RS815RP+, fitted it with 4 x WD Red 6tb drives and configured it in RAID 5. It has 4 gigabit link aggregation connected to a gigabit switch. The backup server is also connected to the same switch with 2 port link aggregation. The server is running Windows Server 2012 R2 with 8gb RAM and a Quad Core. The backup server connects to our cluster with all the production servers on via a 10gb link on the switch. We also have a virtual proxy server installed on the cluster.

We are running a reverse-incremental backup job however it crawls along at 7mb/s with the bottleneck showing as 95% or higher, yet the throughput is minimal on the server and the NAS shows around 10% CPU utilisation. I can't find a reason as to why it's like this. Does anyone have any ideas?
First VM: 14/05/2015 12:19:00 :: Busy: Source 13% > Proxy 6% > Network 96% > Target 2%
Second VM: 14/05/2015 12:24:22 :: Busy: Source 6% > Proxy 5% > Network 95% > Target 0%

Edit: Forgot to say that the Repository is added as a Linux Server, rather than using SMB.

[Shestakov]

Hello,
What transport mode are you using?
It can be recognized by [nbd], [hotadd] or [Direct SAN] tag in the jobs` actions. Thanks.

[BisBestow]

It shows as [hotadd] in the job

[foggy]

I would look for data transfer issues between the virtual proxy and the Linux server.

[emachabert]

Are you talking about 7MB/s or 7mb/s ?
Considering you are using reversed incremental, wich involves a random I/O pattern, with a 4 SATA spindles RAID5, you can't expect more than 7-10 MB/s for incremental runs.
How fast is an active full running ? An active full will be fully sequential and thus will show offbetter numbers.

[dellock6]

I think Eric is spot-on, regardless having added the NAS as a linux machine or as an SMB share, I suspect the limit is the small raid (only 4 spindles) used with reversed incremental.
If the main reason to use reversed incremental is to save disk space by storing only one full backup, you should look at forever forward incremental mode: same space efficiency and a lower IO profile on the storage.