WAN optimization will remove the WAN bottleneck, but that means that suddenly other bottlenecks can appear:
The amount of traffic on the WAN side of the Steelhead appliance will decrease, but on the other side the amount of traffic on the LAN side will increase.
Without WAN optimization, the maximum amount of traffic on the LAN side is the same as the maximum amount of traffic on the WAN side. With WAN optimization, the maximum amount of traffic on the LAN is suddenly not limited to the maximum amount of traffic on the WAN, it will be a multitude of it.
For example, if the WAN link is 10 Mbps and the optimized factor is 15x, then the amount of traffic on the LAN side will be 150 Mbps. If the speed of the Ethernet segment on the LAN side is 100 Mbps, then the LAN side is suddenly the bottleneck.
Figure 1.6. LAN/WAN bandwidth comparisons
Without WAN optimization B p s | y e e |+x+x+x+x+x+x+x+x+x+x+x+x+x+x+ + is incoming WAN t r c | x is outgoing LAN e o | s n | d +----------------------------- time -> With WAN optimization B p s |xxxxxxxxxxxxxxxxxxxxxxxxxxxxx y e e | + is incoming WAN t r c | x is outgoing LAN e o |+++++++++++++++++++++++++++++ s n | d +----------------------------- time ->
The traffic pattern will change: There will be higher bursts of traffic on the client-side LAN, but they will be shorter.
Figure 1.7. Traffic pattern on the client-side LAN side.
Without WAN optimization With WAN optimization LAN traffic out LAN traffic out | | ___ | ||...| | _________ ||...| ||.........| ||...| 0 +----------- 0 +------------- time -> time ->
The CPU utilization on routers and switches behind the Steelhead appliances will change: The amount of packets and data transferred will stay the same over time, but the number of packets per time-period will increase.
The CPU utilization on routers and switches between the Steelhead appliances will change: The number of packets transferred will decrease and the payload will be smaller. Due to the possibility to send more data, the number of packets will increase.
On file servers, the I/O pattern to hard disks will change because the speed with which files are transferred over the network is faster.
Figure 1.8. Disk load on the server.
Without WAN optimization With WAN optimization Disk load Disk load | | ___ | ||...| | _________ ||...| ||.........| ||...| 0 +----------- 0 +------------- time -> time ->
Latency optimization will not only cause the data to be retrieved faster from the servers, it will also increase the amount of traffic in cases where transaction prediction failed.
Figure 1.9. Traffic pattern on the server-side LAN side.
Without WAN optimization With WAN optimization LAN traffic in LAN traffic in | | ______ | ||......| | _________ ||......| ||.........| ||......| 0 +----------- 0 +------------- time -> time ->