One station of the network. More...

#include <line/lang/qn/network_struct.h>

Collaboration diagram for line::qn::Station< T >:

Classes
struct	BalkingThreshold
	One balking threshold: with min_jobs <= n <= max_jobs at the station, an arriving job of the class refuses to join with probability. More...
struct	BalkingParam
	Per class; strategy == NONE is a class that declares no balking. More...
struct	ServerType
	A heterogeneous server pool: count servers that serve only compatible classes, each with its own service law. More...

Public Attributes
std::string	name
NodeType	nodetype = NodeType::Queue
SchedStrategy	sched = SchedStrategy::FCFS
double	nservers = 1.0
	may be infinite (a Delay, or an inf-scheduled task)
bool	attr_ishost = false
std::size_t	attr_idx = 0
	LQN element this station stands for.
std::vector< T >	schedparam
	sn.schedparam, per class: the DPS / GPS weight, or the SEPT / LEPT rank.
double	cap = std::numeric_limits<double>::infinity()
	Station capacity in Kendall's K, as setCapacity sets it.
std::vector< double >	classcap
	Per-class buffer from setChainCapacity; infinite where unset.
std::vector< bool >	immfeed
	Node-level immediate feedback, per class; empty when the station sets none.
std::vector< int >	droprule
	Per-class blocking rule as an INT, with 0 meaning "not set".
std::vector< T >	lldscaling
	sn.lldscaling for this station: the multiplier at population 1, 2, ... Empty when the station is not load dependent.
CdScaling< T >	cdscaling
	sn.cdscaling for this station: the class-dependence map, empty when unset.
std::vector< T >	cdscalingpeak
	sn.cdscalingpeak for this station: the DECLARED peak rate scaling per class, empty when the station is not class dependent.
CdScaling< T >	jdscaling
	sn.jdscaling for this station: MATLAB's Station.ljdScaling, the JOINT dependence map eta_i(n), empty when unset.
std::vector< T >	jdscalingpeak
	sn.jdscalingpeak for this station: the declared peak joint-dependent scaling per class.
std::function< T(const std::vector< std::size_t > &)>	svc_rate_fun
	sn.nodeparam{ind}.svcRateFun for a PAS / OI station: the TOTAL service rate as a function of the ordered microstate, a 1-based list of class indices in queue order.
std::vector< lang::PollingType >	polling_type
	Polling parameters for a POLLING station, MATLAB's pollingType, switchoverTime and pollingPar on the Queue.
std::vector< Distrib< T > >	switchover
int	polling_par = 0
std::vector< std::vector< Distrib< T > > >	switchover_pair
	Queue.setSwitchover(fromClass, toClass, distrib): the walk the server takes when it turns from serving class r to serving class s, MATLAB's (K x K) switchoverTime cell.
Matrix< T >	swap_graph
	sn.nodeparam{ind}.swapGraph: which class a departing job promotes the jobs behind it into.
std::vector< Distrib< T > >	patience
	Queue.setPatience(class, dist): the abandonment timer of a WAITING job, with impatience[r] naming which rule it is.
std::vector< lang::ImpatienceType >	impatience
std::vector< Distrib< T > >	orbit_impatience
	Queue.setOrbitImpatience(class, dist): abandonment from the RETRIAL ORBIT, which is a different population from the waiting line above – a job that gave up retrying never occupied a buffer slot.
std::vector< T >	batch_reject
	Queue.setBatchRejectProbability: per-class rejection of a whole batch.
std::vector< BalkingParam >	balking
std::vector< ServerType >	server_types
lang::HeteroSchedPolicy	hetero_policy = lang::HeteroSchedPolicy::ORDER
std::vector< std::size_t >	server_parallelism
	Queue.setServerParallelism(class, n): the servers a job seizes for the whole of its service, JMT's job parallelism.
std::vector< Distrib< T > >	arrival_batch
	Source.setArrivalBatch(class, dist): the batch-size law released at each arrival epoch.
std::vector< std::size_t >	marked_classes
	Source.markedClasses: the 1-based class of each mark of an MMAP arrival.
std::vector< lang::DepartureDiscipline >	departure_discipline
	Place.departureDiscipline, per class.

Detailed Description

template<class T>
struct line::qn::Station< T >

One station of the network.

Definition at line 593 of file network_struct.h.

Member Data Documentation

◆ arrival_batch

template<class T>

std::vector<Distrib<T> > line::qn::Station< T >::arrival_batch

Source.setArrivalBatch(class, dist): the batch-size law released at each arrival epoch.

It does NOT space the epochs – the arrival process in service does – so the two are separate and both are needed.

Definition at line 791 of file network_struct.h.

Referenced by line::io::network_to_json(), line::qn::Network< double >::set_arrival_batch(), and line::api::sn_remove_class().

◆ attr_idx

template<class T>

std::size_t line::qn::Station< T >::attr_idx = 0

LQN element this station stands for.

Definition at line 599 of file network_struct.h.

◆ attr_ishost

template<class T>

bool line::qn::Station< T >::attr_ishost = false

Definition at line 598 of file network_struct.h.

◆ balking

template<class T>

std::vector<BalkingParam> line::qn::Station< T >::balking

Definition at line 760 of file network_struct.h.

Referenced by line::io::network_to_json(), line::qn::Network< double >::set_balking(), line::api::sn_remove_class(), and line::qn::used_lang_features().

◆ batch_reject

template<class T>

std::vector<T> line::qn::Station< T >::batch_reject

Queue.setBatchRejectProbability: per-class rejection of a whole batch.

Definition at line 743 of file network_struct.h.

Referenced by line::io::network_to_json(), line::qn::Network< double >::set_batch_reject(), and line::api::sn_remove_class().

◆ cap

template<class T>

double line::qn::Station< T >::cap = std::numeric_limits<double>::infinity()

Station capacity in Kendall's K, as setCapacity sets it.

Infinite means unbounded; sn.cap is derived from it and from the class capacities.

Definition at line 611 of file network_struct.h.

Referenced by line::qn::binding_capacity_reason(), line::fes::fes_aggregate(), line::io::network_to_json(), line::qn::NetworkStruct< T >::refresh_capacity(), and line::api::sn_aggregate_chains().

◆ cdscaling

template<class T>

CdScaling<T> line::qn::Station< T >::cdscaling

sn.cdscaling for this station: the class-dependence map, empty when unset.

Definition at line 636 of file network_struct.h.

Referenced by line::qn::cd_factor(), line::qn::NetworkStruct< T >::is_mm1k_loss(), line::io::network_to_json(), line::qn::Network< double >::set_class_dependence(), line::mva::sn_is_mm1k_loss(), line::ctmc::solver_ctmc_avg_from_pi(), line::nc::solver_nc_conv(), and line::qn::used_lang_features().

◆ cdscalingpeak

template<class T>

std::vector<T> line::qn::Station< T >::cdscalingpeak

sn.cdscalingpeak for this station: the DECLARED peak rate scaling per class, empty when the station is not class dependent.

It is not derivable from cdscaling: finding max_n beta_r(n) would mean sweeping the whole population lattice, and the reference does not. It is what utilization at a class-dependent station is normalized by, so that U = T*S/peak keeps the T*S/c convention of an ordinary multiserver station; without it a beta emulating two servers reports twice the true utilization. MATLAB's setClassDependence requires it.

Definition at line 648 of file network_struct.h.

Referenced by line::qn::fj_tag(), line::io::network_to_json(), line::qn::Network< double >::set_class_dependence(), line::api::sn_remove_class(), line::ctmc::solver_ctmc_avg_from_pi(), line::nc::solver_nc_conv(), and line::qn::tag_chain().

◆ classcap

template<class T>

std::vector<double> line::qn::Station< T >::classcap

Per-class buffer from setChainCapacity; infinite where unset.

Definition at line 613 of file network_struct.h.

Referenced by line::qn::binding_capacity_reason(), line::qn::fj_tag(), line::io::network_to_json(), line::qn::NetworkStruct< T >::refresh_capacity(), line::qn::Network< double >::set_class_capacity(), line::api::sn_remove_class(), and line::qn::tag_chain().

◆ departure_discipline

template<class T>

std::vector<lang::DepartureDiscipline> line::qn::Station< T >::departure_discipline

Place.departureDiscipline, per class.

Definition at line 795 of file network_struct.h.

Referenced by line::io::network_to_json(), line::qn::Network< double >::set_departure_discipline(), line::api::sn_remove_class(), and line::qn::used_lang_features().

◆ droprule

template<class T>

std::vector<int> line::qn::Station< T >::droprule

Per-class blocking rule as an INT, with 0 meaning "not set".

The sentinel is MATLAB's: DropStrategy has no member with value 0, so a zero entry is what an unset rule looks like there, and the refresh derives those from the capacity. It matters that the two are distinguishable, because an EXPLICIT WAITQ for an open class at a finite buffer is rejected while the derived one is not.

Definition at line 629 of file network_struct.h.

Referenced by line::qn::fj_tag(), line::mva::list_valid_methods(), line::io::network_to_json(), line::qn::NetworkStruct< T >::refresh_capacity(), line::qn::Network< double >::set_drop_rule(), line::api::sn_remove_class(), and line::qn::tag_chain().

◆ hetero_policy

template<class T>

lang::HeteroSchedPolicy line::qn::Station< T >::hetero_policy = lang::HeteroSchedPolicy::ORDER

Definition at line 777 of file network_struct.h.

Referenced by line::io::network_to_json().

◆ immfeed

template<class T>

std::vector<bool> line::qn::Station< T >::immfeed

Node-level immediate feedback, per class; empty when the station sets none.

Only a Queue carries it in the reference, which is also the only node the JSON writer emits it for.

Definition at line 619 of file network_struct.h.

Referenced by line::io::network_to_json(), line::qn::Network< double >::set_immediate_feedback(), and line::api::sn_remove_class().

◆ impatience

template<class T>

std::vector<lang::ImpatienceType> line::qn::Station< T >::impatience

Definition at line 735 of file network_struct.h.

Referenced by line::io::network_to_json(), line::qn::Network< double >::set_class_patience(), line::qn::Network< double >::set_patience(), line::api::sn_patience_handles(), line::api::sn_remove_class(), and line::qn::used_lang_features().

◆ jdscaling

template<class T>

CdScaling<T> line::qn::Station< T >::jdscaling

sn.jdscaling for this station: MATLAB's Station.ljdScaling, the JOINT dependence map eta_i(n), empty when unset.

It has the same signature as cdscaling and is folded into it multiplicatively wherever a rate is scaled, exactly as State.afterEventInit does. It is kept as a SEPARATE field rather than pre-multiplied into cdscaling because the two carry different modelling claims: a cdscaling beta_r(n) keeps the product form (it is the class-dependent rate lattice pfqn_cdfun evaluates), while an eta_i(n) does not, and pfqn_mvajd / pfqn_ncjd are selected on that distinction.

Definition at line 661 of file network_struct.h.

Referenced by line::qn::cd_factor(), line::qn::NetworkStruct< T >::is_mm1k_loss(), line::io::network_to_json(), line::qn::Network< double >::set_joint_dependence(), line::mva::sn_is_mm1k_loss(), line::ctmc::solver_ctmc_avg_from_pi(), line::nc::solver_nc_conv(), and line::qn::used_lang_features().

◆ jdscalingpeak

template<class T>

std::vector<T> line::qn::Station< T >::jdscalingpeak

sn.jdscalingpeak for this station: the declared peak joint-dependent scaling per class.

setJointDependence makes it mandatory for the same reason setClassDependence does – utilization is reported as T*S/peak.

Definition at line 667 of file network_struct.h.

Referenced by line::qn::fj_tag(), line::io::network_to_json(), line::qn::Network< double >::set_joint_dependence(), line::api::sn_remove_class(), line::ctmc::solver_ctmc_avg_from_pi(), and line::nc::solver_nc_conv().

◆ lldscaling

template<class T>

std::vector<T> line::qn::Station< T >::lldscaling

sn.lldscaling for this station: the multiplier at population 1, 2, ... Empty when the station is not load dependent.

Definition at line 634 of file network_struct.h.

Referenced by line::qn::NetworkStruct< T >::is_mm1k_loss(), line::io::network_to_json(), line::mva::sn_is_mm1k_loss(), line::nc::solver_nc_solve(), and line::qn::used_lang_features().

◆ marked_classes

template<class T>

std::vector<std::size_t> line::qn::Station< T >::marked_classes

Source.markedClasses: the 1-based class of each mark of an MMAP arrival.

Definition at line 793 of file network_struct.h.

Referenced by line::io::network_to_json(), and line::api::sn_remove_class().

◆ name

template<class T>

std::string line::qn::Station< T >::name

Definition at line 594 of file network_struct.h.

Referenced by line::qn::Network< double >::add_delay(), line::qn::Network< double >::add_join_unbound(), line::qn::Network< double >::add_place(), line::qn::Network< double >::add_queue(), line::qn::Network< double >::add_source(), line::qn::NetworkStruct< T >::add_station(), line::qn::binding_capacity_reason(), line::fes::fes_aggregate(), line::mva::fj_ht(), line::mva::fj_mmt(), line::opt::open_sensitivities(), and line::qn::NetworkStruct< T >::refresh_capacity().

◆ nodetype

template<class T>

NodeType line::qn::Station< T >::nodetype = NodeType::Queue

Definition at line 595 of file network_struct.h.

Referenced by line::qn::Network< double >::add_delay(), line::qn::Network< double >::add_join_unbound(), line::qn::Network< double >::add_place(), line::qn::Network< double >::add_queue(), line::qn::Network< double >::add_source(), line::qn::NetworkStruct< T >::add_station(), line::autosolver::auto_find_solver(), line::qn::binding_capacity_reason(), line::fes::fes_aggregate(), line::mva::fj_ht(), line::mva::fj_mmt(), line::qn::from_marginal_and_started_core(), line::qn::from_marginal_core(), and line::qn::NetworkStruct< T >::refresh_capacity().

◆ nservers

template<class T>

double line::qn::Station< T >::nservers = 1.0

may be infinite (a Delay, or an inf-scheduled task)

Definition at line 597 of file network_struct.h.

Referenced by line::qn::Network< double >::add_delay(), line::qn::Network< double >::add_join_unbound(), line::qn::Network< double >::add_place(), line::qn::Network< double >::add_queue(), line::qn::Network< double >::add_source(), line::fes::fes_aggregate(), line::mva::fj_ht(), line::mva::fj_mmt(), line::qn::from_marginal_and_started_core(), line::qn::from_marginal_core(), line::mva::mva_mvac_reason(), line::io::network_to_json(), line::opt::open_sensitivities(), line::io::qn2java(), line::io::qn2matlab(), line::qn::Network< double >::set_server_parallelism(), line::api::sn_aggregate_chains(), line::mva::solver_mva_polling_analyzer(), line::nc::solver_nc_analyzer(), line::nc::solver_nc_solve(), and line::io::tikz_graph().

◆ orbit_impatience

template<class T>

std::vector<Distrib<T> > line::qn::Station< T >::orbit_impatience

Queue.setOrbitImpatience(class, dist): abandonment from the RETRIAL ORBIT, which is a different population from the waiting line above – a job that gave up retrying never occupied a buffer slot.

Definition at line 741 of file network_struct.h.

Referenced by line::io::network_to_json(), line::qn::Network< double >::set_orbit_impatience(), line::api::sn_remove_class(), and line::qn::used_lang_features().

◆ patience

template<class T>

std::vector<Distrib<T> > line::qn::Station< T >::patience

Queue.setPatience(class, dist): the abandonment timer of a WAITING job, with impatience[r] naming which rule it is.

A disabled entry is a class that declares none.

Definition at line 734 of file network_struct.h.

Referenced by line::io::network_to_json(), line::qn::Network< double >::set_class_patience(), line::qn::Network< double >::set_patience(), line::api::sn_patience_handles(), and line::api::sn_remove_class().

◆ polling_par

template<class T>

int line::qn::Station< T >::polling_par = 0

Definition at line 693 of file network_struct.h.

Referenced by line::qn::NetworkStruct< T >::effective_polling(), line::io::network_to_json(), line::qn::Network< double >::set_polling_type(), and line::mva::solver_mva_polling_analyzer().

◆ polling_type

template<class T>

std::vector<lang::PollingType> line::qn::Station< T >::polling_type

Polling parameters for a POLLING station, MATLAB's pollingType, switchoverTime and pollingPar on the Queue.

polling_type[r] is the discipline of class r's buffer (the reference assumes it is identical across buffers), switchover[r] its switchover distribution, and polling_par the K of a K-limited discipline. Empty polling_type means the station is not a polling station.

Definition at line 691 of file network_struct.h.

Referenced by line::qn::NetworkStruct< T >::effective_polling(), line::io::network_to_json(), line::qn::Network< double >::set_polling_type(), and line::mva::solver_mva_polling_analyzer().

◆ sched

template<class T>

SchedStrategy line::qn::Station< T >::sched = SchedStrategy::FCFS

Definition at line 596 of file network_struct.h.

◆ schedparam

template<class T>

std::vector<T> line::qn::Station< T >::schedparam

sn.schedparam, per class: the DPS / GPS weight, or the SEPT / LEPT rank.

Empty means the discipline takes no parameter; the refresh fills it with ones for DPS and GPS, which is MATLAB's default weight.

Definition at line 606 of file network_struct.h.

Referenced by line::qn::fj_tag(), line::io::network_to_json(), line::qn::NetworkStruct< T >::refresh_sched_param(), line::qn::Network< double >::set_sched_param(), line::api::sn_remove_class(), and line::qn::tag_chain().

◆ server_parallelism

template<class T>

std::vector<std::size_t> line::qn::Station< T >::server_parallelism

Queue.setServerParallelism(class, n): the servers a job seizes for the whole of its service, JMT's job parallelism.

Per class, empty or all ones when every job seizes one server.

Definition at line 784 of file network_struct.h.

Referenced by line::io::network_to_json(), line::qn::Network< double >::set_server_parallelism(), line::api::sn_remove_class(), and line::qn::used_lang_features().

◆ server_types

template<class T>

std::vector<ServerType> line::qn::Station< T >::server_types

Definition at line 776 of file network_struct.h.

Referenced by line::io::network_to_json(), line::api::sn_remove_class(), and line::qn::used_lang_features().

◆ svc_rate_fun

template<class T>

std::function<T(const std::vector<std::size_t>&)> line::qn::Station< T >::svc_rate_fun

sn.nodeparam{ind}.svcRateFun for a PAS / OI station: the TOTAL service rate as a function of the ordered microstate, a 1-based list of class indices in queue order.

A pass-and-swap or order-independent queue is parameterized by mu(c) as a whole; there is no per-class service distribution, and MATLAB's setServiceRateFunction rejects one. The refresh still derives a representative per-class rate mu([r]) so the ordinary rate machinery stays consistent, exactly as Queue.setServiceRateFunction does.

Definition at line 680 of file network_struct.h.

Referenced by line::mva::find_oi_stations(), line::mva::nc_is_oi_model(), line::qn::Network< double >::set_pas(), and line::qn::Network< double >::set_service_rate_function().

◆ swap_graph

template<class T>

Matrix<T> line::qn::Station< T >::swap_graph

sn.nodeparam{ind}.swapGraph: which class a departing job promotes the jobs behind it into.

All zero is the ORDER-INDEPENDENT case, where no swapping happens and the station is product-form; a nonzero entry makes it a genuine pass-and-swap station, which the OI analyzer refuses.

Definition at line 718 of file network_struct.h.

Referenced by line::qn::Network< double >::set_pas(), line::qn::Network< double >::set_service_rate_function(), and line::qn::station_swap_graph().

◆ switchover

template<class T>

std::vector<Distrib<T> > line::qn::Station< T >::switchover

Definition at line 692 of file network_struct.h.

Referenced by line::qn::NetworkStruct< T >::effective_polling(), line::io::network_to_json(), line::qn::Network< double >::set_switchover(), and line::mva::solver_mva_polling_analyzer().

◆ switchover_pair

template<class T>

std::vector<std::vector<Distrib<T> > > line::qn::Station< T >::switchover_pair

Queue.setSwitchover(fromClass, toClass, distrib): the walk the server takes when it turns from serving class r to serving class s, MATLAB's (K x K) switchoverTime cell.

Empty means none declared at all; otherwise it is (K x K) and an undeclared pair is DISABLED, so it is not serialized. MATLAB fills its cell with Immediate and writes all K^2 entries instead; nothing reads either, and the sparse form is the one that round-trips a document unchanged.

IT IS CARRIED, NOT CONSUMED, and that is parity rather than an omission. No solver in any codebase reads a pairwise switchover: the MVA polling analyzer and the state machinery reach switchover above through the polling buffers, and writeJSIM warns and DROPS the pairwise times on an ordinary Server because JMT's Server has no switchover of its own. What the field buys is that the model can be declared, serialized and read back unchanged – which is what switchover_basic exercises.

Definition at line 711 of file network_struct.h.

Referenced by line::io::network_to_json(), and line::qn::Network< double >::set_switchover().

The documentation for this struct was generated from the following file:

include/line/lang/qn/network_struct.h

Classes

Public Attributes

Detailed Description

Member Data Documentation

◆ arrival_batch

◆ attr_idx

◆ attr_ishost

◆ balking

◆ batch_reject

◆ cap

◆ cdscaling

◆ cdscalingpeak

◆ classcap

◆ departure_discipline

◆ droprule

◆ hetero_policy

◆ immfeed

◆ impatience

◆ jdscaling

◆ jdscalingpeak

◆ lldscaling

◆ marked_classes

◆ name

◆ nodetype

◆ nservers

◆ orbit_impatience

◆ patience

◆ polling_par

◆ polling_type

◆ sched

◆ schedparam

◆ server_parallelism

◆ server_types

◆ svc_rate_fun

◆ swap_graph

◆ switchover

◆ switchover_pair