#141 Advanced queueing extensions

[Giuliano Casale]

Server vacation: the server goes offline for some time

Breakdown and repair: the server goes offline and then returns after
some time online. Similar to vacation, measures may be different.

Bulk service and arrival (batches): jobs arrive or get service k at a time
Service changeover:

Time dependent arrivals: distribution that depends on simulation time.

Periodic arrivals: distribution, a special case of Time dependent

Age dependent service: for example Multilevel PS serves jobs according
to their age in the system and a set of K user-specified thresholds

Blocking probability: measure

Loss systems: infinite server that drop jobs unless they respect a set
of linear constraints on the class jobs in the station (this can
already be done with FCR with some effort)

Queue-dependent number of active servers: this can be done with some
effort with LD

Busy period: measure, e.g., mean and variance of the busy period and idle period
Setup time: a server takes some time before being able to process a
new job, similar to switchover time

Random number of servers: a job is served by a number of servers at random

General blocking and starvation: The mechanism of general blocking
allows the server to process a limited number of jobs when the buffer
downstream is full, and that of general starvation allows the server
to perform a limited number of services in anticipation of jobs that
are yet to arrive. T

Impolite customers (aka interjections): an arriving job can enter in a
random place of the FCFS queue

Age of Information: metric time elapsed since the freshest received
job still in the system was generated (at system level or
queue-level).

Transient analysis: measure

Multiple waiting lines: jobs of the same class can sit in different
waiting buffers at entry of the queue. Can already be done with
classes with some effort.

Negative jobs, triggers, catastrophes: missing, requirer negative jobs

Coupled queues: the service of a queue depends on the state of another queue

Crossover feedback: there are two priority classes, and the Poisson
arrival process for each class can be subdivided into two groups: one
group which only requires service at the priority level to which it
arrives, and another group which requires subsequent service after it
feeds back to the other queue.

Age process: measure for the age of the jobs inside a queue

Fluid queues: engine change, the moving unit in the simulation is continuous

Discrete time: engine change, the simulation proceeds at equi-spaced
time instants

Shuffling: each arrival, departure or movement of a customer from one
queue to another triggers a shuffle of the other customers at each
queue. The shuffle distribution may depend on the network state and on
the customer that triggers the shuffle.

Reservations/appointments: scheduling policy, when the job in service
completes, it is the one with the reservation that enters next

Ticket queues: a job arrives and takes a ticket, it can renege while
in the queue, but this is observed only when the job is called to
enter service (it's its ticket's turn), so the mean queue-length is
overestimated until then.

Matching system: station where jobs of one type needs to be matched
with jobs of another type, can already be done with advanced join
strategies

Cache: state-dependent class switcher (Giuliano's paper in ToN)

Server affinity / compatibilities: a job can only be processed by
certain servers of the queue, which could be heterogeneous in
processing rate

Random environment: a background process switches the parameters of
the model at random times

Job redundancy and cancellation: a job is forked and then various
cancellation policies are implemented for the tasks that are still
running when k of the tasks complete

Split merge: like fork-join but there is a queue before the fork and
the fork sends only to queues with finite capacity 1 (including job in
service)

Inventory: station type, triggers replenishment orders according to its level