Disclaimer
The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.
Liability Disclaimer
In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.
Posting
The problem is, (I believe) many shapers don't account for L2 overhead. So, you need to shape slower than the nominal L2 bandwidth. The question then, is how much slower. The answer depends on the size of the your underlying packet as the percentage of L2 overhead increases as the packet size decreases.
If you looks at this Table 1 in http://www.cisco.com/web/about/security/intelligence/network_performance_metrics.html, you'll see the minimum size Ethernet packets actually use 84 bytes on the wire to send a payload of 46 bytes. So for those, your overhead is 38/84 = 45.2%. I.e. if your packets were all that small, even shaping at 25% slower would be "too fast".
Conversely, if you were sending maximum (standard) Ethernet packets, your payload is 1500 bytes with 1538 bytes on the wire. In that case, your L2 overhead is only 38/1538 = 2.47%.
Usually, though, all your packets are not minimum nor maximum size, and so you can often shape for average packet size, but when supporting VoIP, you want to err on the side of too small.
Personally, I've found 15% usually works okay, but although 25% seems excessive, it may not be. In fact, depending on your traffic, it may be insufficient.
The above is one of the reasons I'll often suggest running physical interfaces at the logical bandwidth (if possible). I.e. if your agreed bandwidth is 100 Mbps, run the GigEthernet interface at 100 Mbps, not gig. Then use physical congestion on your interface to cause any necessary "backpressure".
BTW, another issue with shapers is they average transmission rate over some time period, which can lead to transient congestion on a physical interface. The impact of this can be minimized by using a small Tc, generally 10 ms or less will work well. What Tc is your Telco using?
PS:
Nice Wiki article with image of parts of an Ethernet frame, http://en.wikipedia.org/wiki/Ethernet_frame
