SHDSL max data rate

kent.plummer · ‎10-19-2007

Hi,

I have two 2304kbps sync'ed SHDSL Cisco 878's in our lab direct connected to a DSLAM.

The maximum data rate that I can reach on the ATM0 interface of each 878 device is as per below 2041kbps, beyond this I get output drops. I am generating 1000B udp packets at 247pps which by my calcs equals:

1967kbps application layer (udp payload only)

2055kbps PPP layer (inc PPP header and up)

2304kbps (ATM PPPoA)

OtNTest5#sh int summ

*: interface is up

IHQ: pkts in input hold queue IQD: pkts dropped from input queue

OHQ: pkts in output hold queue OQD: pkts dropped from output queue

RXBS: rx rate (bits/sec) RXPS: rx rate (pkts/sec)

TXBS: tx rate (bits/sec) TXPS: tx rate (pkts/sec)

TRTL: throttle count

Interface IHQ IQD OHQ OQD RXBS RXPS TXBS TXPS TRTL

------------------------------------------------------------------------

* FastEthernet0 0 0 0 0 2020000 246 0 0 0

FastEthernet1 0 0 0 0 0 0 0 0 0

FastEthernet2 0 0 0 0 0 0 0 0 0

FastEthernet3 0 0 0 0 0 0 0 0 0

BRI0 0 0 0 0 0 0 0 0 0

BRI0:1 0 0 0 0 0 0 0 0 0

BRI0:2 0 0 0 0 0 0 0 0 0

* Vlan1 0 0 0 0 2058000 247 0 0 0

* ATM0 0 0 0 108 1000 2 2041000 248 0

* Dialer0 0 0 0 0 0 0 1476000 177 0

* Virtual-Access1 0 0 0 0 0 0 0 0 0

* Virtual-Access2 0 0 0 0 1000 2 1990000 249 0

OtNTest5#

I expected to see an output rate on the interface up to the port sync speed of 2304kbps.

On the LNS terminating the PPPoA sessions the virtual-access interface's report bandwidth inc PPP header and up. But I'm not sure what the 878's are reporting on the ATM interfaces. It does not align with any of my calculations.

Has anyone measured this is sufficient detail to really understand these metrics?

Cheers

Kent.

paolo bevilacqua · ‎10-20-2007

Hi,

You are calculating the ATM / PPP overhead incorrectly.

1000 bytes ip + 8 llc/snap + 8 bytes PPP overhead + 8 aal5 pdu trailer = 1024 bytes.

1024 divided 48 (payload in a cell) makes 21 comma something, rounded up each packet is taking 22 cells.

22 cells * 53 (bytes in a cell) x 247 pps x 8 (bits in a byte) = 2,304,016 bps, that is 5434 cps.

Do show "atm pvc" and if the indicated cell rate exceeds the numbers above, it's normal that there are drops.

Show interface doesn't take into account certain overhead especially for ATM, and should never be used for an *exact* indication of bitrate on an interface.

hope this helps, please rate post if it does!

kent.plummer · ‎10-22-2007

Thanks for the response.

My cacls are essentially the same as yours. However the 878's do not report a cell rate. All I have to go on is the Interface rate. So I need to be able to convert from Interface reported bit rate to actual line rate inc all ATM headers and padding. I'm also not using LLC - we use AAL5MUX encapulation hence 8 bytes less header.

So working from the reported ATM interface bit rate of 2044,000 bps...

bits/sec 2,044,000 - as reported by show int ATM0

Bytes/sec 255,500

PPP packet size 1040 Bytes (1000udp payload +8 +12 +20)

Cell padding 8 Bytes

Total 255,508 Bytes/sec

Total 2,044,064 bps

2,248,470.4bps inc five-byte ATM cell header (multiply by 1.1 for 5Byte per 48Byte cell)

Close but still not the 2304kbps I'm sync'ed at?

This is a somewhat useful doc http://www.cisco.com/en/US/tech/tk648/tk362/technologies_tech_note09186a0080093c9a.shtml

Cheers

Kent.

paolo bevilacqua · ‎10-22-2007

Hello,

The math comes unchanged, you have not added the AAL5 trailer that is always present even in AAL5MUX:

PPP packet 1040 bytes + 8 AAL5 trailer = 1048 Bytes.

1048 divided 48 = 21.83 rounded up = 22 cells per packet.

As stated above 247 pps of these packets makes

2,304,016 bps that exceeds the sync rate, hence the drops.

kent.plummer · ‎10-22-2007

Thanks again.

What I really want to calculate is the show int atm 0 reported 5min rate I would expect from a sync rate of 2304kbps. From what I read with AAL5MUX PPPoA this should be just the PPP throughput which I calculate to be 2055kbps. In my testing I get 2044kbps as a max reported interface rate (30sec average) which is close but not exact.

paolo bevilacqua · ‎10-23-2007

Hi,

There was another mistake in the previous calculations. The PPP overhead is 6 bytes, not 8 (1 addr, 1 ctrl, 2 fcs).

That doesn't change the cell math:

PPP packet 1,038 bytes + 8 AAL5 trailer = 1,046 Bytes.

1,046 divided 48 = 21.79 rounded up = 22 cells per packet. Cell padding is 10 bytes.

But it changes the interface utilization a bit: 1,046 * 247 * 8 = 2,051,088

that is a closer figure to what you are observing.

Anyway, as mentioned before, you should not ever use the interface reported usage for anything beside approximate interface utilization.

Thanks for the nice rating and good luck!