Method and apparatus for identifying delay causes in traffic traversing a network
View Patent ↗A monitoring technique is provided for diagnosing delays in file transfers between a server and a client via a network, which monitoring technique operates to make measurements of ordinary network traffic on a link unique to a given client/server pair connection. In particular, the invention operates to apply a set of rules to partition the file transfer time into increments attributable to specific identified causes. To that end, the invention intercepts and sorts traffic traversing a predetermined monitor point into a set of flows, wherein each flow consists of all of the traffic between one client and one server. For each flow the packet headers and timestamps of all packets passing in each direction are read. The partitioning of the file transfer times is based on the resulting data.
1. A method for identifying delay causes in a network established to provide connections between a first plurality of Clients and a second plurality of Servers, and wherein data is transmitted in said network via packets, said method comprising the steps of:
providing a Monitor means in said network operable to read and output header data for packets in a multiplicity of said Client-Server connections;
selecting a given connection from said multiplicity of connections and identifying packets belonging to said given connection based on data outputted by said Monitor means;
measuring a transmission time in said network for ones of said identified packets; and
determining an amount by which said network transmission time of said ones of said identified packets exceeds a threshold amount, wherein said threshold amount is related to a nominal transmission time for said identified packets.
2. The method of claim 1 wherein a time stamp is applied to said ones of said identified packets.
3. The method of claim 2 wherein said measured network transmission time is determined as a difference between time stamps for selected pairs of said ones of said identified packets.
4. The method of claim 1 wherein said determining step includes a substep of:
allocating said network transmission time for said ones of said identified packets among a set of identified delay causes.
5. The method of claim 4 wherein said identified packets are partitioned into ones of a set of predefined segments and said determining step is carried out on a segment by segment basis.
6. The method of claim 5 wherein said predefined segments include a connection-open segment, a connection-close segment and one or more data segments.
7. The method of claim 4 wherein said set of identified delay causes include delay attributable to a sub-network serving said Client, said Client host, a sub-network serving said Server, said Server host, and an Excess parameter representing time that a connection is open but not in use.
8. The method of claim 1 wherein said threshold amount in said determining step is approximately equal to an RTT for a network portion traversed by a given identified packet.
9. The method of claim 1 wherein said threshold amount in said determining step is established as a function of an ideal transfer time for said network.
10. The method of claim 9 wherein said ideal transfer time is determined in relation to a TCP slow start boundary curve.
11. The method of claim 9 wherein said ideal transfer time is determined in relation to a maximum of a TCP slow start boundary curve and a network bandwidth limit curve.
12. The method of claim 1 wherein amounts of network transmission time determined to exceed said threshold amount are classified as being caused by network congestion.
13. The method of claim 12 wherein network transmission time amounts classified as being caused by network congestion are output, along with identification of ones of said packets associated with said transmission time amounts, to a user process.
14. The method of claim 1 including the further steps of:
identifying connections accessed by said Monitor means that are characterized by a sub-optimum packet transfer rate; and
marking said connections for non-selection by said selecting step.
15. In a network established to provide connections between a first plurality of Clients and a second plurality of Servers, and wherein data is transmitted in said network via packets, a system for identifying delay causes comprising:
monitor means in said network operable to read and output header data for packets in a multiplicity of said Client-Server connections;
selection means for selecting a given connection from said multiplicity of connections and identifying packets belonging to said given connection based on data outputted by said Monitor means;
timing means for determining a network transmission time interval for ones of said identified packets; and
computing means for determining an amount by which a network transmission time of said ones of said identified packets exceeds a threshold amount.
16. The system of claim 15 wherein said timing means includes a time stamp applied to said ones of said identified packets.
17. The system of claim 15 wherein said computing means further operates to allocate said network transmission time for said ones of said identified packets among a set of identified delay causes.
18. The system of claim 17 wherein said set of identified delay causes include delay attributable to a sub-network serving said Client, said Client host, a sub-network serving said Server, said Server host, and an Excess parameter representing time that a connection is open but not in use.
19. The system of claim 15 wherein said threshold amount is approximately equal to an RTT for a network portion traversed by a given identified packet.
20. The system of claim 15 wherein said threshold amount is established as a function of an ideal transfer time for said network.
21. The system of claim 20 wherein said ideal transfer time is determined in relation to a TCP slow start boundary curve.
22. The system of claim 20 wherein said ideal transfer time is determined in relation to a maximum of a TCP slow start boundary curve and a network bandwidth limit curve.
23. The system of claim 15 wherein amounts of network transmission time determined to exceed said threshold amount are classified as being caused by network congestion.
24. The system of claim 23 wherein network transmission time amounts classified as being caused by network congestion are output, along with identification of ones of said packets associated with said transmission time amounts, to a user process.
25. The method of claim 15 further comprising:
means for identifying connections accessed by said monitor means that are characterized by a sub-optimum packet transfer rate; and
means for marking said connections for non-selection by said selection means.