3 Replies Latest reply: May 9, 2018 6:23 AM by Benoit BUAUD RSS

WAN throughput too high

Benoit BUAUD


With a Steelhead Riverbed 2050L (9.1.4#62) in an in-path topology, I don't understand why there is a such difference between Riverbed WAN throughput statistics and bandwidth sized on router.

Router have a link at 100Mb/s. On work hours, output traffic is going up to 4Mbps while it reaches 95Mpbs in peak throughput at the same time.  There is no mistake about the router and the Riverbed concerned.

In spite of I've looked for articles on support (cf. below), I don't have a consistent explanation

If someone could give me an consistent explanation, I would apreciate


Benoit BUAUD.


extract from support


[…]Viewing WAN Throughput Reports

The WAN Throughput report summarizes the WAN throughput for the time period specified. In standard in-path and virtual in-path deployments, the throughput is an aggregation of all data the system transmits out of all WAN interfaces. In a server-side out-of-path configuration, the report summarizes all data the system transmits out of the primary interface.


The WAN Throughput report doesn’t include any traffic that is hardware bypassed, either by an in-path interface in hardware bypass, or the portion of traffic that is bypassed by hardware-assist rules on supported Fiber 10 Gigabit-Ethernet in-path cards.

The WAN Throughput report includes a WAN link throughput graph that provides these statistics describing data activity for the time period you specify.



https://supportkb.riverbed.com/support/index?page=content&id=S14315https://supportkb.riverbed.com/support/index?page=content&id=S14315 extract from KB S14315 ( https://supportkb.riverbed.com/support/index?page=content&id=S14315

« […] Why are there throughput spikes in the WAN and LAN graphs that are higher than the physical link capacity?

Throughput statistics are extremely sensitive to variation in timing, and spikes can result from delays in statistics collection for a given time interval. For example when the sport process is under high load, the statistics collection event might be delayed by 0.1 seconds, which would lead to a recorded throughput value ~110% of the physical link throughput for that second.[…]