IEPM

Error free seconds between SLAC, FNAL CMU and CERN

Supported by DOE MICS Link to SLAC Home Page

This page illustrates the behavior of the fraction of Error free seconds (EFS) metric and its inverse the fraction of Errored seconds (ES) between SLAC, FNAL, CMU and CERN.

This metric is heavily used in phone company performance objectives and so it is important for Internet telephony to understand how well the Internet performs when measured according to this metric. The CCITT defines Error-free seconds as a measure of the quality of the signal being transmitted. It is a percentage representing the total amount of time over a 24-hour period that the signal contained bit errors and it is calculated using a test pattern defined in CCITT Recommendation 0.151. The Error free seconds & Errored seconds are defined as:

Error free seconds = EFS = (measured seconds with no errors / total seconds for measurement)
Errorred seconds = ES = (1 - EFS)
For more on how to measure error free seconds see: T1 for Computer Networks.

For our analysis of Internet performance the Error-free seconds is measured by looking at the Surveyor one way probes and looking for how long consecutive probes don't get through. The Surveyor probes are launched on average 2 times/second with a Poisson distribution and are one way probes.

To give a scale for the expectation for error free seconds it is noteworthy that AT&T's DDS service states 99.5% error free seconds for a DDS circuit, FTS2000 gives 99.6% error free seconds for an NSAP-1, American Telesis offers a 64kbps dedicated digital servive with 5 speeds up to 64kbps and 99.95 percent Error-Free Seconds (EFS) measured over 24 consecutive hours or 99.85 percent availability measured over 12 consecutive months, PacBell's Advanced Digital Network "automatically takes steps to correct any errors that occur, making your transmissions perform at 99.96% error-free seconds", BellSouth's MegaLink DS1 service goal is to provide 99.95% error free seconds (and if the service experiences a 60-second outage they refund that month's recurring charges), Bell Atlantic's Digipath II Digital Service offers 5 speeds up to 56kbps with 99.5% error-free seconds of transmission at all five transmission speeds, based on past performance, UniDial Network Services give 99.96% measured over 24 consecutive hours for DS-0 DDS access and a similar value for DS3 services, Ameritech guarantees a performance objective of 99.995% error-free seconds for its DS1 service. The Automotive Network eXchange sets their availability or access to network metric at 99.7% which means less than 3 hours/year of downtime, and allow less than 10 events of 30 seconds or greater per year.

Overall error free seconds distribution

The plot below shows the distribution of error free seconds / month measured between SLAC, FNAL, CMU and CERN outage frequency for the data for all pairs added together for November 1998 thru July 1999. It includes data from about 284 million probes or 142 million seconds. The table of numbers in the plot indicate the median error free seconds, the 25 and 75 percentiles and the inter quartile range (IQR). The magenta line indicates the cumulative frequency.

Monthly error free seconds by site pair and by month

The plot below shows the percentage of errored seconds per month per site pair from November 1998 thru July 1999.

Comparison of Errored seconds and packet loss

The PingER analysis of Surveyor data reports on packet loss rather than errored seconds, so it is useful to understand how well the two metrics compare with one another. We therefore calculated both the % errored seconds per month for each pair and the % packet loss per month for each pair. We then scatter plotted the points with the x coordinate given by monthly % errored seconds and y coordinate by % packet loss as shown below.

In the above plot, the points are fitted to a straight line (black line) and a power series (red line). The parameters of the fits are also shown together with the R2 which indicates there is a strong correlation (R2 > 0.6).
Created: 23 September 1999; last update 28 September 1999
URL: http://www-iepm.slac.stanford.edu/monitoring/surveyor/outage.html
Comments to iepm-l@slac.stanford.edu