Systems and methods for initializing packet transfers
A method of initiating a new data transfer across a network, such as across a persistent network having historical data transfer performance information, is disclosed. Other embodiments of the invention comprise networks which implement methods of the invention. In accordance with one aspect of the invention, functional adjustments are made to an initial data window based upon an analysis of a historical data transfer performance and idle time since a last data transfer.
1. A method of initiating a new data transfer across a persistent network having historical data transfer performance information, comprising the steps of:
identifying, at a transmitting end of the persistent network, an entropy in connection with one or more connection metrics, the entropy is based on an amount of time since a last transfer across the persistent network;
adjusting, at the transmitting end of the persistent network, a congestion window for the new data transfer based on a selected one of a plurality of adjustment states, a different adjustment is associated with each adjustment state of the plurality, the adjusting comprising:
in a first adjustment state where the identified entropy satisfies an entropy threshold and an available bandwidth estimate satisfies an available bandwidth threshold, adjusting the congestion window in accordance with a maximum receive rate and a round trip time; and
in a second adjustment state where the identified entropy satisfies the entropy threshold and a queueing delay estimate satisfies a queueing delay threshold, adjusting the congestion window in accordance with a receive rate and a minimum round trip time; and
in a third adjustment state where the identified entropy exceeds the entropy threshold, adjusting the congestion window in accordance with a short-term receive rate and a minimum round trip time and further adjusting the congestion window as a function of the identified entropy; and
using, at the transmitting end of the persistent network, the adjusted congestion window for the new data transfer.
2. The method of claim 1 , further comprising:
in the second and third adjustment states, setting a congestion window threshold equal to a previous congestion window setting.
3. The method of claim 1 , further comprising:
receiving, at transmitting end of persistent network, at least one of the one or more connection metrics from a receiving end in the persistent network.
4. The method of claim 1 , further comprising:
using short-term receive rate estimate as the available bandwidth estimate; and
using a receive rate percentage as the available bandwidth threshold.
5. The method of claim 4 , the receive rate percentage used for the available bandwidth threshold comprising a short-term and long-term convergent estimate of an available data rate.
6. The method of claim 4 , the short-term receive rates slope estimate comprises a linear least squares regression estimate of the short-term receive rate.
7. The method of claim 1 , further comprising:
using a ratio of the round trip time to the minimum round trip time as the queueing delay estimate.
8. The method of claim 1 , further comprising:
using a multiple of the round trip time as the entropy threshold.
9. The method of claim 8 , the multiple is used to control a timing for entering the third adjustment state.
10. A method of initiating a new data transfer across a persistent network having historical data transfer performance information, comprising the steps of:
making a first determination, at a transmitting end of the persistent network, whether to accept one or more existing connection metrics as accurately measuring a persistent network's state, the first determination using an idle time of the persistent network;
determining, at a transmitting end of the persistent network, a first setting for an initial congestion window for a new data transfer using the first determination that accepts the one or more existing connection metrics as accurately measuring the persistent network's state and a second determination whether a current congestion window is efficiently using available bandwidth of the persistent network;
determining, at a transmitting end of the persistent network, a second setting for the initial congestion window for the new data transfer based on the first determination that accepts the one or more existing connection metrics as accurately measuring the persistent network's state, the second setting is based on at least a third determination, which differs from said second determination, as to whether the current congestion window is efficiently using the available bandwidth of the persistent network;
determining, at a transmitting end of the persistent network, a third setting for the initial congestion window for the new data transfer based upon the first determination that rejects the one or more existing connection metrics as accurately measuring the persistent network's state, the third setting is determined by decreasing a determined initial congestion window logarithmically as a function of the idle time; and
initiating, at a transmitting end of the persistent network, the new transfer using the initial adjusted congestion window.
11. The method of claim 10 , further comprising:
receiving, at the transmitting end of persistent network, at least one of the one or more existing connection metrics from a receiving end of the persistent network.
12. The method of claim 10 , making a first determination further comprising:
making the first determination that accepts the one or more existing connection metrics as accurately measuring the persistent network's state where the idle time is less than a multiple of a weighted average measure of round trip packet latencies; and
making the first determination that rejects the one or more existing connection metrics as accurately measuring the persistent network's state where the idle time is at least equal to the multiple of the weighted average measure of round trip packet latencies.
13. The method of claim 12 , the multiple operates to postpone using the third setting of the initial configuration window.
14. The method of claim 10 , further comprising:
using a slope of a short-term receive rate as the short-term receive rate accuracy; and
using a percentage of an available bandwidth estimate.
15. The method of claim 14 , the available bandwidth estimate comprises a short-term and long-term convergent available bandwidth estimate.
16. The method of claim 14 , further comprising:
receiving the available bandwidth estimate from a receiving end of the persistent network.
17. The method of claim 14 , the slope comprises a linear least squares regression estimate of the short-term receive rate.
18. The method of claim 10 , determining the first setting for the initial congestion window further comprising:
determining the first setting for the initial congestion window using a maximum receive rate and a round trip time.
19. The method of claim 10 , further comprising:
determining a queueing delay; and
making the third determination using the determined queueing delay and a queueing delay threshold.
20. The method of claim 10 , determining a third setting for the initial congestion window for the new data transfer further comprising:
determining the third setting for the initial congestion window using a receive rate and a minimum round trip time.
21. The method of claim 20 , further comprising:
setting a congestion window threshold equal to a previous congestion window setting.
22. The method of claim 20 , determining a third setting for the initial congestion window further comprising:
determining the initial congestion window using a short term receiving rate and a minimum round trip time; and
decreasing the determined initial congestion window logarithmically as a function of the idle time.
23. The method of claim 22 , further comprising:
setting a congestion window threshold equal to a previous congestion window setting.