Automatic Bandwidth Measurement

The Automatic Bandwidth Measurement can ease enablement time and cost. The CoS Bandwidth Settings configuration includes the ability to specify automatic bandwidth configuration information including: the algorithm to use for the test, the IP addresses of the target test servers (intended to be located on the Internet), the frequency with which to execute the test, and whether the test is for the primary or non-primary interface. A test for the primary interface and a test for the non-primary interface are both allowed for a VPN Gateway configuration. The configured test servers are reported on the Advanced Information page of the VPN Gateway web interface and via the SNMP polling data.

There are two algorithms for bandwidth determination. A second, more refined algorithm (algorithm 2) handles variation between measurements to avoid swings in the value to which the upstream and downstream bandwidth is throttled. The differences between the two algorithms are described later, but first the common features to both algorithms will be described.

The WAN bandwidth is measured by applications local to the VPN Gateway. The older, “legacy”, application executes the FTP protocol to retrieve a file from a measurement server to determine the downstream bandwidth. The file size used varies based on last measured bandwidth, starting with a file size of 500K. The upstream test sends a 1MB stream of data towards the discard port on the same test server, while the newer method of bandwidth testing uses an application that sends and receives data from a dedicated process running on a measurement server. While much more data is sent and received than with the legacy method, more accurate bandwidth measurements are calculated.

When Automatic Bandwidth testing is configured, and the VPN Gateway reboots, a test will be scheduled to execute after the interface in question has been up consistently for 3 minutes. If the test is scheduled to execute but the interface in question is not up, then that test will be postponed until the interface is available.

The results of the test and the next scheduled execution of the test are published on the VPN Gateway web interface. All measurements from the previous tests since the last reboot are also available for review. The test results that are currently in-use are reported to the AT&T Administration Server when a standard device query is performed. Information about the test results being used and the next scheduled execution of the test is also reported via the SNMP polling data.

Since this test sends and receives a large amount of data in both directions it should be considered an intrusive test, from the standpoint that it will interfere with other data being sent and received on the Internet connection while the test is running. Competing data being transferred when the test is run can also result in artificially low measurements being taken. Algorithm 2 minimizes this possibility.

If the VPN Gateway is configured to use automatic bandwidth detection, it must be able to determine the bandwidth before it can set up any Class of Service parameters and do any sort of prioritization of traffic.

If the bandwidth detection test fails for some reason, then the previous settings are used. If a test has never successfully executed, then the VPN Gateway cannot be allowed to continue working until valid measurements have been obtained. For that reason, if a bandwidth test fails and there has been no prior test performed, there is no information to use in CoS configuration, so the VPN Gateway will be temporarily disabled. When in this mode:

  • The web-based logon page will show the discontinued status.
  • The FOCUS BOOT/ALIVE messages will show a status of ‘D’.
  • Most traffic through the VPN Gateway will be blocked.

NOTE: Traffic will still be allowed to maintenance addresses such as AT&T FOCUS, AT&T Administration Server, DNS servers, and STP servers.

  • AT&T Administration Server will still show the VPN Gateway as active.

As soon as a test successfully completes, the CoS settings will be configured in the VPN Gateway and it will return itself to full operation and reflect Active status in the AT&T Administration Server. If the VPN Gateway is rebooted when in dormant mode, it will temporarily be reactivated and a new broadband test will be run. If the test fails again, the VPN Gateway will be set in the dormant mode until the problem is resolved. If the bandwidth test continues to fail then a new test destination can be configured in the VPN Gateway’s profile or the upstream and downstream bandwidth will need to be set explicitly in order for the box to be restored to service. If neither of these is acceptable then CoS will have to be disabled in the AT&T Administration Server for the VPN Gateway.

Time of Day Windows

Time of Day windows for can be used to schedule when bandwidth measurements are allowed.

Bandwidth Measurement Algorithm 1

This is the original bandwidth measurement. It is recommended that this no longer be used if algorithm 2 is available.

With this algorithm, the VPN Gateway will immediately clamp the bandwidth based on the results of each test, whether bad or good. If the test happens to run at the same time there is competing data transfer, an abnormally low measurement of upstream or downstream bandwidth could result and the bandwidth set on the VPN Gateway accordingly. This can adversely affect performance if set too low, and with the default settings, it will be 24 hours before the test is run again and better values restored.

Bandwidth Measurement Algorithm 2

This algorithm includes several major improvements to avoid issues resulting from rogue bad measurements.

A high water mark algorithm is introduced. Instead of automatically using whatever measurement is taken, a running history of previous measurements is maintained, and the actual bandwidth clamping references these previous measurements. In this manner, any bad measurements can be ignored under the assumption that they are indeed outliers and do not indicate a real degradation in the line quality since the previous measurement. A threshold is configurable through the AT&T Administration Server to indicate how much lower than a previous measurement a set of results can be while still being referenced. Any new measurement that is less than the current settings by more than a threshold percentage will not be used immediately. Any measurement that is better than the current setting will be used immediately. If the current WAN configuration (DHCP, PPPoE, etc.) changes, then the existing high water marks that were measured under the old WAN configuration will be removed. This helps ensures accurate, current values that will reflect the new configuration.

A sliding window is used to expire the current setting if it is more than a configurable number of measurement periods old. Once a measurement is accepted and used to set traffic shaping parameters, it cannot remain in effect indefinitely, even if all subsequent measurements are lower. Once a measurement used for the current upstream or downstream setting has expired, the highest of the subsequent measurements that have already been taken is used and becomes a new high water mark.

Testing without CoS

The VPN Gateway can run Automatic Bandwidth tests without CoS being enabled. In this scenario the Automatic Bandwidth testing automatically takes place using the specified configuration. As when CoS is enabled, the results are reported via the VPN Gateway web interface as well as the SNMP polling data. The results are also reported to the AT&T Administration Server when a standard device query is performed. The test results are for informational purposes only and are not used in any way to control bandwidth on the VPN Gateway. Tests on the non-primary interface are also allowed.