This page provides explanations on each of the simulation engine metrics collected in the ROSS instrumentation layer.

Metric Description
nevent_processed Number of forward events processed.
nevent_abort Number of events aborted.
nevent_rb Number of reverse events processed.
rb_total Total number of rollbacks (primary + secondary).
rb_prim Total number of primary rollbacks (i.e., caused by receiving an event from the past).
rb_sec Total number of secondary rollbacks (i.e., caused by receiving an anti-message).
fc_attempts Number of fossil collection attempts.
pq_qsize Size of the priority queue (splay tree). This is the number of forward events to be processed at this time.
network_send Number of network sends.
network_recv Number of network recvs.
num_gvt Number of GVT computations.
event_ties Event ties detected on PEs.
efficiency Rollback efficiency: 1 - (events rolled back / net events). If this number is positive, more forward events are being processed than reverse events. If negative, we’re spending more time on rollback processing.
virtual_time_diff Difference in a KP’s local virtual clock and GVT. Typically positive (i.e., so KP is ahead of GVT), but can be negative if the KP has not processed any events since the last GVT.

The following metrics are based on cycle counters to determine number of cycles spent in each type of processing.

Metric Description
net_read_time Time spent polling network for received events
net_other_time Time spent performing other network tasks
gvt_time Time spent in GVT computation
fc_time Time spent in fossil collection
event_abort_time Time spent processing aborted events
event_proc_time Time spent in event processing
rb_time Time spent processing rollbacks
pq_time Time spent on priority queue operations
cancel_q_time Time spent processing events in the cancel queue
avl_time Time spent in AVL tree operations