Low transfer speeds on Dockstar, but why?

# ramdisk to mount stuff to
tmpfs          /mnt/ram1/      tmpfs   defaults         0       0

async, noatime

/etc/rc.local
blockdev --setra 16384 /dev/sdX  # set the readahead value to 16MiB
ifconfig eth0 txqueuelen 5000  # buffer 5000 outgoing packets in memory
echo deadline > /sys/block/sdX/queue/scheduler # use alternative I/O-scheduler 

/etc/sysctl.conf

# % allocation of memory for I/O-Requests, before data is written immediatly to disk
vm.dirty_ratio=8
# % allocation of memory for I/O-Requests, before data is written in the background to disk
vm.dirty_background_ratio=5
# use memory for I/O-Cache first, rather then applications, when possible
vm.vfs_cache_pressure=50
# swap out unused applications, instead of flushing I/O-Cache
vm.swappiness=100
# buffer up to 5000 received packets in memory, to avoid packetloss
net.core.netdev_max_backlog=5000
# disable TCP-Extensions, to save space in TCP-Header
net.ipv4.tcp_sack=0
net.ipv4.tcp_timestamps=0
# reuse sockets faster
net.ipv4.tcp_fin_timeout=10
net.ipv4.tcp_tw_recycle=1