# This topology (SimpleRelayWithCongestion) is designed for testing simple # network congestion on a InteriorRouter. You can change its behaviour by # defining multiple variables for EFCP, RMT and FA. Several implications needs # to be taken into account when defining these variables and their mutual relations. # For Flow Allocator it is mainly the createRequestTimeout. It specifies the # upper limit in which it expects response for createFlow request. So if you want # to avoid re-sending createFlow request, set it high enough to accomodate for # recursive Flow creation. It depends heavily on chosen topology. # Default value is 10s. You can change it in .ini. # # Example: # **.fa.createRequestTimeout = 12s # # For EFCP it is # initialSenderCredit (default 10) # rcvCredit (default 10) # closedWindowQLen (default 4) # mpl (default 50s) # rtt (default 2s) # initialSenderCredit specifies the initial credit on the sender direction. # Setting this variable high (might) cause the # InteriorRouter get congested. Setting it too low will on # the other hand prevent the initial overload. After first # exchange of FlowControlPDU, the initialSenderCredit looses # its importance. # rcvCredit is the value that is send (in form of RcvRightWindowEdge) # in mentioned FlowControlPDU. # closedWindowQLen takes place when SndRightWindowEdge permits from sending # another PDU, so DTP starts to put them on closedWindowQ. # When lenght of this queue reaches closedWindowQLen, # DTP signals Push-Back to upper flow. # mpl is maximum PDU lifetime and is used for computation few # inner timers. This variable is fixed throughout simulation run. # rtt is initial value for round-trip time. RTT is then gradually # updated through RTTEstimator policy. Set it high enough, # so the first DataTransferPDU - Ack exchange have chance # to updated it before retransmission timer expiration. # # In RMT you can set defaultThreshQLength, defaultMaxQLength and maxQPolicyName. # defaultThreshQLength sets initial threshold length of dynamically initiated RMT # queues (10 by default). # defaultMaxQLength sets initial maximum length of dynamically initiated RMT # queues (20 by default, defaultMaxQLength >= defaultThreshQLength). # maxQPolicyName specifies the RMT MaxQueue policy to be used in the scenario. # This policy is invoked by a queue each time number of # queued PDUs exceeds the defaultThreshQLength. The sample # MaxQueue policy used for invocation of congestion notifications # is called "UpstreamNotifier". # # CongestionPing: # We are trying to congest InteriorRouter so the connection between InteriorRouter # and Host2 has bigger latency, queues from InteriorRouter to Host2 are shorter, etc. # # Used AE: AEPing - behave in same sense as ICMP Echo Request/reply # **.host1.applicationProcess1.applicationEntity.iae.dstApName = "App2" # **.host1.applicationProcess1.applicationEntity.iae.dstAeName = "Ping" # **.host1.applicationProcess1.applicationEntity.iae.startAt = 10s -> start of Application entity (not start of sending PINGs) # **.host1.applicationProcess1.applicationEntity.iae.pingAt = 60s -> AE starts sending PINGs # **.host1.applicationProcess1.applicationEntity.iae.rate = 50 -> number of PINGs that will be sent # **.host1.applicationProcess1.applicationEntity.iae.stopAt = 250s -> time of deallocation # **.interiorRouter.relayIpc.relayAndMux.defaultMaxQLength = 5 -> shortened output queue towards Host2 # **.interiorRouter.relayIpc.relayAndMux.defaultThreshQLength = 3 -> lowered threshold # **.efcp.rtt = 25s -> high enough # **.host1.ipcProcess1.efcp.efcp.initialSenderCredit = 50 -> Host1 can send up to 50 PDUs before getting Ack or FlowControl update # **.interiorRouter.ipcProcess1.efcp.efcp.initialSenderCredit = 3 -> But IPC towards Host2 on InteriorRouter in the lower DIF can send only up to 3 PDUs before getting Ack or FlowControl update # **.interiorRouter.ipcProcess1.efcp.efcp.maxClosedWinQueLen = 4 -> after reaching 4 PDUs on closedWindowQ this IPC emits Push-Back # **.host2.ipcProcess0.efcp.efcp.rcvCredit = 3 -> thanks to this, the sender credit on Interior router in ipcProcess1 stays same as initialSenderCredit even after FlowControl update. # **.relayAndMux.maxQPolicyName = "UpstreamNotifier" -> Name of policy. # # Important Events: # t=10s - created connection between Host1.ipcp1 - Host1.ipcp0 # t=10.000Xs - created connection between Host1.ipcp0 - interiorRouter.ipcp0, interiorRouter.ipcp0 - interiorRouter.relayIpc, interiorRouter.relayIpc - interiorRouter.ipcp1. # t=15s - created connection between interiorRouter.ipcp1 - Host2.ipcp0. # t=25s - created connection Host2.ipcp0 - Host2.ipcp1, Host2.ipcp1 - Host2.irm # t=30s - created connection Host1.ipcp1 - Host1.irm # t=60s - start of sending PINGs # t=62s - interiorRouter.ipcp1 - senderCredit gets depleted (see SndRightWindowEdge -5 and NextSeqNumToSend -6) # t=63s - interiorRouter.ipcp1 - first PDU is put on closedWindowQ # t=66s - interiorRouter.ipcp1 - closedWindowQ is full - initiate PushBack (block upper flow), # - .relayIpc - RMT shuts down port towards ipcp1 and it starts to fill # t=70s - interiorRouter.ipcp1 - Ack is received -> there is space in closedWindowQ -> RMT port is unblocked and 1 PDU is released. # t=70s - interiorRouter.ipcp1 - closedWindowQ is full - initiate PushBack (block upper flow) (this happens several times # t=95s - interiorRouter.relayIpc - RMT port is full - SlowDown mechanism is invoked # - - RIBd sends CDAP message to Host1.ipcp1 to "SlowDown" # t=95s - Host1.ipcp1 - RIBd receives CDAP message to "SlowDown" and ECNSlowDownPolicy in DTCP is initiated. # # # CongestionStream # Used AE: AEStream - Sends messages to the other side. Unlike AEPing, it does not send response back. # Important Events: # t=10s - created connection between Host1.ipcp1 - Host1.ipcp0 # t=10.000Xs - created connection between Host1.ipcp0 - interiorRouter.ipcp0, interiorRouter.ipcp0 - interiorRouter.relayIpc, interiorRouter.relayIpc - interiorRouter.ipcp1. # t=15s - created connection between interiorRouter.ipcp1 - Host2.ipcp0. # t=25s - created connection Host2.ipcp0 - Host2.ipcp1, Host2.ipcp1 - Host2.irm # t=30s - created connection Host1.ipcp1 - Host1.irm # # t=60s - start of sending PINGs # t=62s - interiorRouter.ipcp1 - senderCredit gets depleted (see SndRightWindowEdge -5 and NextSeqNumToSend -6) # t=63s - interiorRouter.ipcp1 - first PDU is put on closedWindowQ # t=66s - interiorRouter.ipcp1 - closedWindowQ is full - initiate PushBack (block upper flow), # - .relayIpc - RMT shuts down port towards ipcp1 and it starts to fill # t=70s - interiorRouter.ipcp1 - Ack is received -> there is space in closedWindowQ -> RMT port is unblocked and 1 PDU is released. # t=70s - interiorRouter.ipcp1 - closedWindowQ is full - initiate PushBack (block upper flow), # t=84s - interiorRouter.relayIpc - RMT port is full - SlowDown mechanism is invoked # - - RIBd sends CDAP message to Host1.ipcp1 to "SlowDown" # t=84s - Host1.ipcp1 - RIBd receives CDAP message to "SlowDown" and ECNSlowDownPolicy in DTCP is initiated.