Replication is consuming all bandwidth
As Nicoli pointed out, replication takes all the available bandwidth since it does not know what is available until you tell it. You can throttle the bandwidth in one of two methods the one I am going to recommend is the one using the agent.cfg located in /disk/etc/puredisk. If you set the bandwidth to a KBps setting that you would like to use things should be well.
The issue is NBU uses KiloBytes and networking uses KiloBits so some math is involved. In your case you have a 30Mbps link (3840KBps [(30/8)*1024]) and you want to use only 20Mbps for replication. you would calculate the 20Mbps change to KBps by dividing the 20 by 8 to get MBps (2.5MBps) and then multiply that by 1024 to get KBps (2560KBps). Edit the agent.cfg and locate the bandwidth=0 location and change that to 2560. The very next time the replication process checks the agent.cfg for that setting the bandwidth will change for that job eventually all jobs conform.
The reason I recommend this method is because you will always use the 20MBps regardless of how many jobs you have running. If you use the second method you will need to calculate how many potential jobs you would be running and then set the OPDUP_BANDWIDTH setting to the 20Mbps divided by the potential amount of jobs and that is what will be used PER job regardless of the number running. If you set OPDUP_BANDWIDTH to 124 KBps and only one job was running then that job would use only 128KBps. With the agent.cfg model you would use all 20Mbps if you had only one job
A more in depth understanding can be seen in the Deduplication administrator's guide http://www.symantec.com/business/support/index?page=content&id=DOC6466 look around page 86.