Reliable ISP Access Cloud state detection method and apparatus

Abstract

A Multi-Homing System is equipped with an Adaptive ISP Access Cloud State Detection apparatus (ACSD) that improves the reliability of the availability of digital connections (links) between computer sites, such as a Computer Premises Network and the Internet, in which such connections are made by connecting through a multiplicity of ISP Access Clouds (links). Reliability is improved over prior art methods by using data elements of Internet Protocol datagrams, e.g. record fields or bits of fields, that are regularly and normally exchanged between the ISP Access Clouds and the CPN without creating additional data traffic. Data Elements from each ISP Access Cloud are used by processing functions of the by the ACSD to test for conditions that indicate that it may be in a DOWN status, when a DOWN status is suspected, other functions in the ACSD initiate transmission of a set of PROBE packets that can reliably determine if the suspect link is actually DOWN or merely giving a response that would be interpreted as DOWN by prior art methods.

1 . An adaptive network communications system communicating through a 1 st  plurality of communication access paths to a 2 nd  plurality of destinations each identified by a destination address that uses a subset of observed round trip times (RTT) for a set of messages sent through said 1 st  plurality of communication access paths to detect UP or DOWN status of communication through any of said communication access paths, said system comprising:
 a. means for determining said Up or Down status of one of said communications paths as Up when more than one message directed to one of said destinations through said one of said communication paths is acknowledged by said destination within a wait delay period. TIMEOUT, and determining said Up or Down status of said one of said communications paths as Down when said more than one message directed to said one of said destinations through said one of said communication paths is not acknowledged by said destination within said wait delay period. TIMEOUT, wherein TIMEOUT equals a multiple of the maximum RTT of said subset of observed times   
   
   
       2 . The network communications system set forth in  claim 1 , in which said multiple is a small number. 
   
   
       3 . The network communications system set forth in  claim 2 , in which said number is about 3. 
   
   
       4 . An adaptive MULTIHOMING SYSTEM for reliable monitoring and reporting UP/DOWN status of a plurality N of ISP-ACCESS CLOUDs connected to said adaptive MULTIHOMING SYSTEM, comprising:
 a. an ACCESS CLOUD STATE DETECTION apparatus comprising:   b. a permanent memory store PM- 1 ;   c. a dynamic memory store DM- 1 ;   d. a computing resource CR- 1 ; comprising:   e. a CPU-& I/O chip set;   f. a Control Program and a plurality of Program_Modules in cooperation with said CPU-& I/O chip set that control the ACCESS CLOUD STATE DETECTION to execute operations, comprising:
 i. for each one of said N ISP-A/C's. A continually and continuously read packet data flow information for packets directed to and from each one of said N ISP-A/C's; 
 ii. on a first condition for a packet directed to and from each ISP-A/C-n of said N ISP-A/C's, store said directed packet's flow information in a TCP-Flow-Table; 
 iii. on a second condition for a packet directed to each ISP-ACCESS CLOUD-n of said N ISP-ACCESS CLOUDs, add TCP destination IP address of said packet to and delete an old TCP destination IP address from, a list of TCP destination IP addresses allocated to a memory portion DYNAMIC SEED LIST in said dynamic memory; 
 iv. on a third condition for packets directed to an ISP-ACCESS CLOUD-n of said N ISP-ACCESS CLOUDs, update a permanent memory portion DEFAULT SEED LIST in said memory store PM- 1  with said list of TCP destination IP addresses stored in Dynamic Memory- 1 ; 
 v. for each inbound packet from each ISP-ACCESS CLOUD-n of said N ISP-ACCESS CLOUDs, determine the inbound byte count, and add said count to a corresponding ISP-ACCESS CLOUD-n inbound-byte-counter, and on a regular, periodic interval of integer k-seconds, compare its inbound-byte-counter value at time=t. [ISP-n(t)] with its inbound-byte-counter value at time=t+k. [ISP-n(t+k)], and if said values are the same, set a Blackout_Hint variable to YES, and initiate an UP/DOWN status test of said ISP-ACCESS CLOUD-n to determine if said ISP-ACCESS CLOUD-n is UP or DOWN. 
 vi. on a fourth condition for packets directed to a TCP destination IP address, send a set PROBE of packet sequences, to said TCP destination IP address through each one of said N ISP-ACCESS CLOUDs, measure a Round Trip Time value to and from said TCP destination IP address for each one of said N ISP-ACCESS CLOUDs, and store said value for each one of said N ISP-ACCESS CLOUDs and each said destination address in a ROUND TRIP TIME table; 
 vii. wherein:
 1. each PROBE set probe packet has a TCP source port number that is a number P, such that flow information <X, Y, P, Q> for each said probe packet is not found in said TCP-Flow-Table. 
 2. wherein P is found by:
 a. said ACCESS CLOUD STATE DETECTION computing resource CR- 1  sets variables X, Y, Q given by X=TCP source address. Y=TCP destination IP. Q=a TCP destination port; 
 b. said ACCESS CLOUD STATE DETECTION computing resource CR- 1  instantiates a label GENERATE: and calls a random number generator procedure RANDOM (P); 
 c. said ACCESS CLOUD STATE DETECTION computing resource CR- 1  calls a Program Module SEARCH Flow Table for <X, Y, P, Q> in which variable flow-found=YES, if <X, Y, P, Q> is found in said Flow Table and then CR- 1  sends control back to said label GENERATE. ELSE CR- 1  sets flow-found=NO; SET TCP source port=P.