If you have a source DTE with 4 1gb/s interfaces trunked into a duplex capable 10g switch then you should easily attain your desired throughput. Whether the end host system can receive 320/mbs throughput depends on it's configuration. It's noted that lacp is far from perfect without factoring in the latencies from NFS and/or VM translation. Also, your backend potential throughput is faster than a lot of local
disk buses.