Littleの法則 L=λE[T] は、平均系内数、流量、平均滞在時間を結びます。強みは境界を揃えた記帳にあり、Poisson到着を必要としません。ただし平均時間をp99に置き換えることはできません。
面積が滞在時間の総和になる理由
区間 [0,H] における要求 i の占有寄与は指示関数 1{ai≤t<di} です。要求を足し合わせて積分すると:
∫0HN(t)dt=i∑∣[ai,di)∩[0,H]∣.
これは確率を仮定しない標本路の恒等式です。両辺の単位は要求·s。窓の前からいる要求や、窓の後に完了する要求も、共通部分だけを数えれば正しく寄与します。
始点と終点で空になり、窓内に到着した m 要求がすべて退出するなら:
NˉH=H1i=1∑mTi=Hm⋅m1i=1∑mTi.
長期到着率と平均滞在時間が存在し、境界の残余寄与を H で割った値が0へ向かうなどの条件で、L=λE[T] が得られます。定常状態はよく使う解析の枠組みであり、有限窓の面積恒等式の前提ではありません。短時間落ち着いて見えるだけで極限条件は保証されません。
観測の打切りが落とすもの
実験は到着 [0,1,2,6] s、処理時間 [3,1,2,1] sです。H=5 sでは最初の2件が完了しています。占有面積は 3+3+3=9 要求·s、平均系内数は1.8。完了率 2/5 と完了者の平均3 sの積は1.2で、差は未完了の3件目が既に寄与した3要求·sです。
図を準備しています
Littleの法則と統計の境界 · 実験到着[0,1,2,6] s、処理時間[3,1,2,1] s。観測終点を変え、全占有面積と完了ジョブだけの寄与を比べます。初期は空、拒否はありません。
{"id":"queueing-02","title":"Littleの法則と統計の境界 · 実験","summary":"到着[0,1,2,6] s、処理時間[3,1,2,1] s。観測終点を変え、全占有面積と完了ジョブだけの寄与を比べます。初期は空、拒否はありません。","height":1000,"html":"<p class=\"intro\" data-i18n=\"scope\"></p><div class=\"controls\"><div class=\"control\"><label for=\"parameter-0\" data-i18n=\"windowEnd\"></label><output id=\"value-0\" for=\"parameter-0\"></output><input type=\"range\" id=\"parameter-0\" data-parameter=\"0\" min=\"3\" max=\"9\" step=\"0.5\" value=\"5\"/></div></div><div class=\"tools\"><button id=\"reset\" data-i18n=\"reset\"></button></div><div id=\"timeline\"></div><div id=\"plots\"></div><div id=\"metrics\"></div><p id=\"status\" role=\"status\" aria-live=\"polite\"></p>","css":".plotarea{display:grid;grid-template-columns:max-content minmax(0,1fr);gap:8px}.yticks{display:flex;flex-direction:column;justify-content:space-between;padding:3px 0;font-size:13px;font-variant-numeric:tabular-nums;color:var(--muted)}*{box-sizing:border-box}.intro{margin:0 0 16px;color:var(--muted);line-height:1.65}.controls{display:grid;gap:14px}.control{display:grid;grid-template-columns:1fr auto;gap:6px 12px;align-items:center}.control label{font-size:16px;line-height:1.5}.control input{grid-column:1/-1;width:100%;min-height:28px}.control output{font-variant-numeric:tabular-nums}.control select{grid-column:1/-1;width:100%;max-width:100%;padding:8px;font:inherit;background:var(--surface);color:var(--ink);border:1px solid var(--rule);border-radius:6px}.tools{margin:14px 0}.tools button{font:inherit;padding:8px 14px;min-height:42px}.chart{margin:22px 0}.chart h3{font-size:16px;margin:0 0 6px;line-height:1.5}.chart svg{display:block;width:100%;height:160px;overflow:hidden}.scale,.axes{font-size:14px;color:var(--muted);font-variant-numeric:tabular-nums}.axes{display:flex;justify-content:space-between;align-items:start;gap:8px;margin-top:6px}.axes span:nth-child(2){text-align:center;flex:1;min-width:0}.legend{display:flex;flex-wrap:wrap;gap:8px 18px;margin-top:8px;font-size:15px}.legend span{display:inline-flex;align-items:center;gap:7px}.legend i{display:inline-block;width:22px;flex-shrink:0}#metrics{border-top:1px solid var(--rule);padding-top:12px;display:grid;gap:8px}#metrics>div{display:flex;justify-content:space-between;gap:16px;font-size:15px;line-height:1.5}#metrics strong{font-weight:600;font-variant-numeric:tabular-nums;flex-shrink:0}#status{font-size:14px;color:var(--muted);min-height:3em;margin:14px 0 0}button:focus-visible,input:focus-visible,select:focus-visible{outline:2px solid var(--accent);outline-offset:3px}@media(max-width:420px){.chart svg{height:145px}#metrics>div{flex-wrap:wrap;gap:3px 12px}}\n.job{margin:16px 0}.job-title{font-size:16px;font-weight:600}.job-times{display:flex;flex-wrap:wrap;gap:4px 16px;font-size:14px;color:var(--muted);line-height:1.6}.job svg{display:block;width:100%;height:26px;margin:6px 0}.timeline-key{display:flex;flex-wrap:wrap;gap:12px;font-size:15px;line-height:1.5}.timeline-key span{border-left:12px solid var(--accent);padding-left:6px}.timeline-key span:first-child{border-color:var(--muted)}\n.presets{display:flex;flex-wrap:wrap;grid-column:1/-1;gap:8px}.presets button{font:inherit;line-height:1.5;min-height:44px;max-width:100%;padding:8px 12px;border:1px solid var(--rule);border-radius:6px;color:var(--ink);background:var(--surface);text-align:left;white-space:normal}.presets button[aria-pressed=\"true\"]{border-color:var(--accent);box-shadow:inset 0 0 0 1px var(--accent)}\n","js":"const MODEL_ID=2, DEFAULTS=[5];\n/* Independent teaching models. Seconds, jobs, and jobs/second unless stated.\n Pure functions: no DOM, clock, timers or external packages. */\nconst mean=a=>a.reduce((s,x)=>s+x,0)/a.length;\nconst curve=(name,points)=>({name,points});\nconst chart=(title,xlabel,curves,extra={})=>({title,xlabel,curves,...extra});\nfunction fifo(arrivals,services){let end=0;return arrivals.map((a,i)=>{const b=Math.max(a,end);end=b+services[i];return {id:i+1,a,b,d:end,s:services[i],w:b-a,t:end-a};});}\nfunction stepCounts(jobs,end,key){const events=jobs.map(j=>j[key]).filter(t=>t<=end).sort((a,b)=>a-b),p=[[0,0]];let n=0;for(const t of events)p.push([t,n],[t,++n]);p.push([end,n]);return p;}\nfunction occupancy(jobs,end){const ev=jobs.flatMap(j=>[[j.a,1],[j.d,-1]]).filter(e=>e[0]<=end).sort((a,b)=>a[0]-b[0]||a[1]-b[1]);let n=0;const p=[[0,0]];for(const [t,v]of ev){p.push([t,n],[t,n+v]);n+=v;}p.push([end,n]);return p;}\nfunction erlangC(c,rho){const a=c*rho;let term=1,sum=1;for(let n=1;n<c;n++){term*=a/n;sum+=term;}const tail=term*a/c/(1-rho);return tail/(sum+tail);}\nfunction finiteQueue(lambda,mu,K){const rho=lambda/mu;const w=Array.from({length:K+1},(_,n)=>rho**n),z=w.reduce((s,x)=>s+x,0),prob=w.map(x=>x/z),block=prob[K],through=lambda*(1-block),L=prob.reduce((s,x,n)=>s+n*x,0);return {prob,block,through,L,W:L/through,busy:1-prob[0]};}\nfunction seeded(seed){let s=seed>>>0;return()=>{s=(Math.imul(1664525,s)+1013904223)>>>0;return (s+.5)/4294967296;};}\nfunction simulateMM1(rho,count,seed){const random=seeded(seed),arr=[],srv=[];let t=0;for(let i=0;i<count;i++){t+=-Math.log(random())/(100*rho);arr.push(t);srv.push(-Math.log(random())/100);}return fifo(arr,srv);}\nfunction queueModel(id,p){let plots=[],metrics={},rows;\n if(id===1){const [gap,long]=p,jobs=fifo([0,gap,2*gap,3*gap],[long,1,1,1]),end=jobs.at(-1).d;\n rows=jobs;plots=[chart('cumulative','timeAxis',[curve('arrivals',stepCounts(jobs,end,'a')),curve('starts',stepCounts(jobs,end,'b')),curve('departures',stepCounts(jobs,end,'d'))])];metrics={meanWait:mean(jobs.map(j=>j.w)),meanTime:mean(jobs.map(j=>j.t)),lastDeparture:end};\n }else if(id===2){const [end]=p,jobs=fifo([0,1,2,6],[3,1,2,1]),overlap=j=>Math.max(0,Math.min(j.d,end)-j.a),area=jobs.reduce((s,j)=>s+overlap(j),0),done=jobs.filter(j=>j.d<=end),completedArea=done.reduce((s,j)=>s+j.t,0);\n plots=[chart('occupancy','timeAxis',[curve('inSystem',occupancy(jobs,end))],{yrange:[0,4]})];metrics={area,meanN:area/end,completedProduct:completedArea/end,boundaryArea:area-completedArea,completed:done.length};\n }else if(id===3){const [rate,window]=p,m=rate*window,prob=[];let q=Math.exp(-m);for(let n=0;n<=Math.ceil(m+7*Math.sqrt(m)+5);n++){if(n)q*=m/n;prob.push([n,q]);}\n plots=[chart('countProbability','countAxis',[curve('poisson',prob)],{bars:true,yrange:[0,Math.max(...prob.map(x=>x[1]))*1.12]}),chart('gapSurvival','timeAxis',[curve('exponential',Array.from({length:101},(_,i)=>{const t=5/rate*i/100;return [t,Math.exp(-rate*t)];}))],{yrange:[0,1]})];metrics={countMean:m,countVariance:m,emptyProbability:Math.exp(-m),meanGap:1/rate};\n }else if(id===4){const [rho]=p,delta=100*(1-rho),q99=Math.log(100)/delta;\n plots=[chart('tailProbability','timeAxis',[curve('systemTail',Array.from({length:101},(_,i)=>{const t=q99*1.2*i/100;return [t,Math.exp(-delta*t)];})),curve('waitingTail',Array.from({length:101},(_,i)=>{const t=q99*1.2*i/100;return [t,rho*Math.exp(-delta*t)];}))],{yrange:[0,1]}),chart('latency','rhoAxis',[curve('meanTime',Array.from({length:95},(_,i)=>{const r=(i+1)/100;return [r,1/(100*(1-r))];})),curve('p99',Array.from({length:95},(_,i)=>{const r=(i+1)/100;return [r,Math.log(100)/(100*(1-r))];}))],{event:rho})];metrics={utilization:rho,meanTime:1/delta,meanWait:rho/delta,p99:q99,zeroWait:1-rho,meanN:rho/(1-rho)};\n }else if(id===5){const [ca2,cs2]=p,points=f=>Array.from({length:95},(_,i)=>{const r=(i+1)/100;return[r,f(r)];}),pk=r=>r/(1-r)*(1+cs2)/2*.01,king=r=>r/(1-r)*(ca2+cs2)/2*.01;\n plots=[chart('waitingSeconds','rhoAxis',[curve('poissonReference',points(pk)),curve('kingman',points(king))],{event:.8})];metrics={pkWait:pk(.8),kingmanWait:king(.8),secondMoment:(1+cs2)*.0001,residualBusy:(1+cs2)*.005};\n }else if(id===6){const [c,rho]=p,C=erlangC(c,rho),wait=C/(c*100*(1-rho));\n const pts=f=>Array.from({length:95},(_,i)=>{const r=(i+1)/100;return[r,f(r)];});\n plots=[chart('waitingSeconds','rhoAxis',[curve('pooled',pts(r=>erlangC(c,r)/(c*100*(1-r)))),curve('split',pts(r=>r/(100*(1-r))))],{event:rho})];metrics={arrivalRate:c*100*rho,waitProbability:C,pooledWait:wait,splitWait:rho/(100*(1-rho)),meanTime:wait+.01};\n }else if(id===7){const [long,policy]=p,sizes=[long,1,1,1,1],order=policy?[1,2,3,4,0]:[0,1,2,3,4];let end=0;rows=order.map(i=>{const b=end;end+=sizes[i];return{id:i+1,a:0,b,d:end,s:sizes[i],w:b,t:end};});\n plots=[];metrics={meanWait:mean(rows.map(j=>j.w)),meanTime:mean(rows.map(j=>j.t)),longWait:rows.find(j=>j.id===1).w,maxSlowdown:Math.max(...rows.map(j=>j.t/j.s)),makespan:end};\n }else if(id===8){const [lambda,K]=p,q=finiteQueue(lambda,100,K);\n plots=[chart('stateProbability','countAxis',[curve('stationary',q.prob.map((v,n)=>[n,v]))],{bars:true})];metrics={offeredRatio:lambda/100,blocking:q.block,acceptedRate:q.through,meanN:q.L,meanTime:q.W,utilization:q.busy};\n }else if(id===9){const [lambda,feedback]=p,visit=1/(1-feedback),local=lambda*visit,stable=local<100,demands=[visit*.01,visit*.004];\n plots=[chart('utilizationChart','nodeAxis',[curve('nodeLoad',[[1,lambda*demands[0]],[2,lambda*demands[1]]])],{bars:true,yrange:[0,Math.max(1.2,lambda*demands[0]*1.1)]})];metrics={visits:visit,internalRate:local,bottleneckBound:100/visit,stable,networkTime:stable?visit*(1/(100-local)+1/(250-local)):null};\n }else if(id===10){const [rho,count,seed]=p,jobs=simulateMM1(rho,count,seed),warm=Math.floor(count/5),cohort=jobs.slice(warm),end=jobs.at(-1).a,done=cohort.filter(j=>j.d<=end),times=cohort.map(j=>j.t).sort((a,b)=>a-b),q99=times[Math.ceil(.99*times.length)-1],theory=1/(100*(1-rho)),max=Math.max(q99*1.2,theory*Math.log(100)*1.2);\n let ix=0;const ecdf=Array.from({length:151},(_,i)=>{const t=max*i/150;while(ix<times.length&×[ix]<=t)ix++;return[t,1-ix/times.length];});\n plots=[chart('tailProbability','timeAxis',[curve('empirical',ecdf),curve('stationaryTheory',Array.from({length:151},(_,i)=>{const t=max*i/150;return[t,Math.exp(-t/theory)];}))],{yrange:[0,1]})];metrics={sampleCount:times.length,censored:cohort.length-done.length,cohortMean:mean(times),completedMean:done.length?mean(done.map(j=>j.t)):null,theoryMean:theory,sampleP99:q99,theoryP99:theory*Math.log(100)};\n }else if(id===11){const [capacity,prob]=p,base=60,factor=1+prob+prob*prob,normal=base*factor,burst=120*factor,Qpeak=Math.max(0,burst-capacity)*30,drain=capacity>normal?Qpeak/(capacity-normal):Qpeak===0?0:null;\n const pts=Array.from({length:241},(_,i)=>{const t=i,rate=t<20||t>=50?normal:burst;let Q;if(t<20)Q=Math.max(0,normal-capacity)*t;else if(t<50)Q=Math.max(0,normal-capacity)*20+Math.max(0,burst-capacity)*(t-20);else Q=Math.max(0,Math.max(0,normal-capacity)*20+Qpeak+(normal-capacity)*(t-50));return [t,Q];});\n plots=[chart('backlog','timeAxis',[curve('queuedWork',pts)]),chart('rateChart','timeAxis',[curve('attemptRate',[[0,normal],[20,normal],[20,burst],[50,burst],[50,normal],[240,normal]]),curve('capacity',[[0,capacity],[240,capacity]])])];metrics={attemptFactor:factor,normalAttempts:normal,burstBacklog:pts[50][1],canDrain:capacity>normal,drainSeconds:capacity>normal?pts[50][1]/(capacity-normal):null,endBacklog:pts.at(-1)[1]};\n }\n return {plots,metrics,...(rows?{rows}:{})};\n}\nif(typeof module!=='undefined')module.exports={queueModel,fifo,erlangC,finiteQueue,simulateMM1};\n\nconst $=s=>document.querySelector(s), colors=['var(--accent)','#b77818','#269282'],dash=['','7 4','2 4'];\nconst number=x=>Number.isFinite(x)?viz.number(x,Number.isInteger(x)?0:Math.abs(x)<.01&&x!==0?5:3):'—';\nfunction plot(parent,spec){\n const all=spec.curves.flatMap(s=>s.points),xx=all.map(p=>p[0]),yy=all.map(p=>p[1]);\n const zero=true;\n let lo=zero?Math.min(0,...yy):Math.min(...yy),hi=zero?Math.max(0,...yy):Math.max(...yy),xmin=Math.min(...xx),xmax=Math.max(...xx);\n if(hi===lo)hi=lo+1;if(xmin===xmax)xmax=xmin+1;\n const pad=(hi-lo)*.08;lo=Math.max(0,lo-pad);hi+=pad;\n if(spec.bars){xmin-=.5;xmax+=.5;}if(spec.xrange)[xmin,xmax]=spec.xrange;if(spec.yrange)[lo,hi]=spec.yrange;\n const sx=x=>8+(x-xmin)/(xmax-xmin)*584,sy=y=>8+(hi-y)/(hi-lo)*144;\n const section=document.createElement('section');section.className='chart';\n const title=document.createElement('h3');title.textContent=viz.t(spec.title);section.append(title);\n const scale=document.createElement('div');scale.className='scale';scale.textContent=viz.t('vertical')+' '+number(lo)+' … '+number(hi);section.append(scale);\n const ns='http://www.w3.org/2000/svg',svg=document.createElementNS(ns,'svg');svg.setAttribute('viewBox','0 0 600 160');svg.setAttribute('preserveAspectRatio','none');svg.setAttribute('role','img');svg.setAttribute('aria-label',viz.t(spec.title));\n function node(name,attrs){const n=document.createElementNS(ns,name);for(const [k,v]of Object.entries(attrs))n.setAttribute(k,v);svg.append(n);return n;}\n for(let j=0;j<=4;j++)node('line',{x1:8,x2:592,y1:8+j*36,y2:8+j*36,stroke:'var(--rule)','stroke-width':1});\n if(lo<=0&&hi>=0)node('line',{x1:8,x2:592,y1:sy(0),y2:sy(0),stroke:'var(--muted)','stroke-width':1});\n if(spec.scatter)node('line',{x1:sx(0),x2:sx(0),y1:8,y2:152,stroke:'var(--muted)','stroke-width':1});\n if(spec.event!==undefined&&spec.event>=xmin&&spec.event<=xmax)node('line',{x1:sx(spec.event),x2:sx(spec.event),y1:8,y2:152,stroke:'var(--muted)','stroke-dasharray':'3 3'});\n spec.curves.forEach((s,i)=>{if(spec.bars){for(const [x,y]of s.points)node('line',{x1:sx(x),x2:sx(x),y1:sy(0),y2:sy(y),stroke:colors[i%3],'stroke-width':Math.min(16,400/s.points.length)});}else if(spec.scatter||spec.dots===i){for(const [x,y] of s.points)node('circle',{cx:sx(x),cy:sy(y),r:spec.scatter?5:2.8,fill:colors[i%3]});}else{let pts=s.points;if(spec.title==='replicas')pts=pts.flatMap((p,j)=>j?[[p[0],pts[j-1][1]],p]:[p]);node('path',{d:pts.map(([x,y],j)=>(j?'L':'M')+sx(x).toFixed(2)+' '+sy(y).toFixed(2)).join(' '),fill:'none',stroke:colors[i%3],'stroke-width':2,'stroke-dasharray':dash[i%3],'vector-effect':'non-scaling-stroke'});}});\n const area=document.createElement('div');area.className='plotarea';const ticks=document.createElement('div');ticks.className='yticks';for(const v of [hi,(hi+lo)/2,lo]){const label=document.createElement('span');label.textContent=number(v);ticks.append(label);}area.append(ticks,svg);section.append(area);\n const axes=document.createElement('div');axes.className='axes';const fmt=x=>spec.logx?number(10**x):number(x);\n for(const text of [fmt(spec.bars?xmin+.5:xmin),viz.t(spec.xlabel),fmt(spec.bars?xmax-.5:xmax)]){const a=document.createElement('span');a.textContent=text;axes.append(a);}section.append(axes);\n const legend=document.createElement('div');legend.className='legend';spec.curves.forEach((s,i)=>{const el=document.createElement('span'),sw=document.createElement('i');sw.style.borderTop=`3px ${i===0?'solid':i===1?'dashed':'dotted'} ${colors[i%3]}`;el.append(sw,document.createTextNode(viz.t(s.name)));legend.append(el);});section.append(legend);parent.append(section);\n}\nfunction timeline(rows){\n const host=$('#timeline');host.replaceChildren();if(!rows.length)return;\n const key=document.createElement('div');key.className='timeline-key';for(const k of ['waiting','service']){const e=document.createElement('span');e.textContent=viz.t(k);key.append(e);}host.append(key);\n const max=Math.max(...rows.map(r=>r.d)),ns='http://www.w3.org/2000/svg';\n for(const r of rows){const box=document.createElement('div');box.className='job';const title=document.createElement('div');title.className='job-title';title.textContent=viz.t('job')+' '+r.id;box.append(title);\n const svg=document.createElementNS(ns,'svg');svg.setAttribute('viewBox','0 0 600 26');svg.setAttribute('preserveAspectRatio','none');svg.setAttribute('role','img');svg.setAttribute('aria-label',viz.t('waiting')+' '+number(r.w)+' s; '+viz.t('service')+' '+number(r.s)+' s');\n for(const [a,b,color]of [[r.a,r.b,'var(--muted)'],[r.b,r.d,'var(--accent)']]){const rect=document.createElementNS(ns,'rect');rect.setAttribute('x',a/max*600);rect.setAttribute('y','3');rect.setAttribute('width',(b-a)/max*600);rect.setAttribute('height','20');rect.setAttribute('fill',color);svg.append(rect);}box.append(svg);\n const text=document.createElement('div');text.className='job-times';for(const [k,val]of [['arrivalTime',r.a],['startTime',r.b],['endTime',r.d]]){const e=document.createElement('span');e.textContent=viz.t(k)+' '+number(val)+' s';text.append(e);}box.append(text);host.append(box);\n }const axis=document.createElement('div');axis.className='axes';for(const s of ['0',viz.t('timeAxis'),number(max)]){const e=document.createElement('span');e.textContent=s;axis.append(e);}host.append(axis);\n}\n\nfunction update(){\n const values=[...document.querySelectorAll('[data-parameter]')].map(e=>+(e.matches('.presets')?e.dataset.value:e.value));\n document.querySelectorAll('input[data-parameter]').forEach(e=>$('#value-'+e.dataset.parameter).textContent=number(+e.value));\n document.querySelectorAll('[data-choice]').forEach(b=>b.setAttribute('aria-pressed',String(b.dataset.choice===b.parentElement.dataset.value)));\n const r=queueModel(MODEL_ID,values);document.body.dataset.result=JSON.stringify(r.metrics);document.body.dataset.parameters=JSON.stringify(values);\n timeline(r.rows||[]);$('#plots').replaceChildren();r.plots.forEach(s=>plot($('#plots'),s));$('#metrics').replaceChildren();\n for(const [k,v]of Object.entries(r.metrics)){const line=document.createElement('div'),label=document.createElement('span'),value=document.createElement('strong');label.textContent=viz.t(k);value.textContent=typeof v==='boolean'?viz.t(v?'yes':'no'):number(v);line.append(label,value);$('#metrics').append(line);}\n $('#status').textContent=viz.t('updated');\n}\ndocument.querySelectorAll('input[data-parameter]').forEach(e=>e.addEventListener('input',update));\ndocument.querySelectorAll('[data-choice]').forEach(b=>b.addEventListener('click',()=>{b.parentElement.dataset.value=b.dataset.choice;update();}));\n$('#reset').addEventListener('click',()=>{document.querySelectorAll('[data-parameter]').forEach((e,i)=>{if(e.matches('.presets'))e.dataset.value=DEFAULTS[i];else e.value=DEFAULTS[i];});update();});update();\n","audio":false,"strings":{"area":"系内数の曲線面積 / ジョブ·s","boundaryArea":"未完了ジョブの寄与 / ジョブ·s","completed":"完了したジョブ数","completedProduct":"完了率 × 完了ジョブ平均","inSystem":"システム内","meanN":"平均系内ジョブ数 L","no":"いいえ","occupancy":"システム内ジョブ数 N(t)","reset":"実験をリセット","scope":"到着[0,1,2,6] s、処理時間[3,1,2,1] s。観測終点を変え、全占有面積と完了ジョブだけの寄与を比べます。初期は空、拒否はありません。","timeAxis":"時間 / s","updated":"再計算しました。— はこの条件では適用できない値です。","vertical":"縦軸の範囲","windowEnd":"観測の終了時刻 / s","yes":"はい"}}
終点を7 s以後へ動かすと全件完了し、境界寄与が0になります。さらに観測を延ばすと空の時間だけが増え、二つの計算結果は一緒に下がります。窓の長さも統計の定義に含まれます。
境界、対象群、期間を揃える
安定したプールで受付率200/s、平均接続保持時間40 msなら、平均占有接続数は 200×0.040=8。これは接続を8本に設定すれば待ち時間目標を満たすという意味ではありません。平均占有と同じ容量では変動への余裕を別途確保できません。
待機区では Lq=λE[Wq]、全体では L=λE[T]。外部1000/sのうち800/sだけ入るなら、内部占有には800/sを使います。失敗や取消も退出に含めるなら、その群の滞在時間も含めます。成功スループットだけと全試行の在途数を組み合わせてはいけません。
再試行では一つの業務要求に複数の試行が対応します。業務単位なら最初の発行から最終結果まで、試行単位なら各回の入退場を測ります。どちらも有用ですが、個数と時間を混ぜられません。
確認問題
- 平均系内数50、完了率500/sなら、p99は100 msですか。
考え方
長期の境界が一致すれば平均系内時間が100 msと分かるだけです。同じ平均を持つ分布でも裾は異なります。分位点には要求単位の分布か追加モデルが必要です。
- 障害中に系内数が増えても完了者平均が低いとき、法則が破れたのでしょうか。
考え方
遅い要求が未完了で統計から抜けている可能性があります。瞬時の系内数も時間平均とは違います。窓内面積、未完了者の既経過時間、窓の前に入った要求を確認し、長期モデルの条件も検討します。
続けて読む
MIT 2026年講義はLittleの法則と部分システムへの適用を扱います。上の打切り例は独自の時刻列から直接導いたものです。