retrieval_chain.m

% Cache with a chained retrieval system (delayed hits).
clc; clear solver AvgTable;
 
accessProb   = [0.6, 0.3, 0.1];   % per-item access probabilities (3 items)
serviceRates = [2.0, 2.0, 2.0;    % Queue_1 per-item retrieval rates
                3.0, 3.0, 3.0];   % Queue_2 per-item retrieval rates
nQueues      = 2;
 
model = Network('DelayedHits');
 
n = size(serviceRates, 2);        % number of items
capacity = [1];                   % per-level cache capacity
 
source    = Source(model, 'Source');
cacheNode = Cache(model, 'Cache', n, capacity, ReplacementStrategy.FIFO);
queues = cell(1, nQueues);
for i = 1:nQueues
    queues{i} = Queue(model, sprintf('Queue_%d', i), SchedStrategy.FCFS);
end
sink      = Sink(model, 'Sink');
 
jobClass  = OpenClass(model, 'InitClass', 0);
hitClass  = OpenClass(model, 'HitClass', 0);
missClass = OpenClass(model, 'MissClass', 0);
 
source.setArrival(jobClass, Exp(1));
 
pAccess = DiscreteSampler(accessProb);
cacheNode.setRead(jobClass, pAccess);
cacheNode.setHitClass(jobClass, hitClass);
cacheNode.setMissClass(jobClass, missClass);
 
cacheNode.setRetrievalSystem(jobClass, missClass, queues, serviceRates);
 
% Per-item chain: Cache -> Queue_1 -> Queue_2 -> Cache 
for item = 1:n
    cacheNode.setItemQueueEntryProbability(jobClass, item, queues{1}, 1.0);
    for q = 1:(nQueues - 1)
        cacheNode.setItemRoutingProbability(jobClass, item, queues{q}, queues{q + 1}, 1.0);
    end
    cacheNode.setItemQueueExitProbability(jobClass, item, queues{nQueues}, 1.0);
end
 
P = model.initRoutingMatrix();
P{jobClass, jobClass}(source, cacheNode)  = 1.0;
P{hitClass, hitClass}(cacheNode, sink)    = 1.0;
P{missClass, missClass}(cacheNode, sink)  = 1.0;
model.link(P);
 
% Simulation
SSA(model, 'samples', 5000, 'method', 'serial', 'seed', 1).getAvgCacheTable
LDES(model, 'samples', 1e6, 'seed', 1).getAvgCacheTable
 
% Analytical
MVA(model).getAvgCacheTable
NC(model).getAvgCacheTable
 
% Item-level cache occupancy
MVA(model).getAvgItemTable
NC(model).getAvgItemTable