…

Sequenziell VS Wahlfrei

• Sequential VS random access

• Sl.5 - Sequentieller Zugriff kann schneller sein als aufeinanderfolgende wahlfreie Zugriffe auf verteilte Adressen

• TODO - RAM = “Hauptspeicher”?

  • dict.cc says “MainMemory”
  • Is this terminology “historisch bedingt”?ainMemory"?
  • Is this terminology “historisch bedingt”?

Scale-in VS scale-out cost-benefit-analysis

• Sl.13
  • Chances that the data is in the current box
  • then count the cost (=time) of accessing local data (=SMP access) VS external (in one of the other servers)
  • for $n$ servers: $$1ms + f*(100ns/n + 100\micro s * (n-1/n))$$
  • cost for creating the info, $f$ ignored, cost for local access and cost for external access by the Wahrscheinlichkeit the data is somewhere else (1/n chances is in the current box, (n-1)/n chances it’s not)
  • $f$: high/medium/light communication (Sl.15)
• TL;DR when you have to cluster high-end servers, their advantage sinks

Seeks VS Scans

• (L) Disk Throughput == Durchsatz

• Sl.17 - TL;DR sometimes going through all records is faster than through 1/10 of them randomly, because looking for them takes time, and if you do it sequentially the path is small (and therefore time is negligible) and it’s faster at the end

• (L) Bottleneck = der Flaschenhals

• (L) Volumen – Wiktionary:

  •