open atlas
↑ Back to track
System Design Foundations SD · 01 · 06

Scalability: free-recall review

Free-recall prompts across the scalability unit. Answer each in your own words first — including the arithmetic — then reveal the model answer and compare.

SD Senior ◷ 14 min
Level
FoundationsJuniorMiddleSenior

Retrieval beats re-reading. For each prompt, say or write a full answer from memory — including the arithmetic — before you open the model answer. The effort of reconstruction is what makes Little’s Law, the queueing curve, and the latency numbers stick.

Reconstruct the unit’s spine without looking back: why throughput and latency are different axes, why latency goes vertical near a resource’s limit, what vertical and horizontal scaling each cost, how to estimate a system on a napkin, and which latency numbers reject a design on sight.

Recall before you leave
  1. 01
    State Little's Law, define each term, and use it to find a throughput ceiling.
  2. 02
    Why does latency go vertical near a resource's limit, and what utilization do you target?
  3. 03
    Compare vertical and horizontal scaling, and state what the USL adds beyond Amdahl's law.
  4. 04
    Walk a back-of-envelope estimate and name the two cheat sheets and two rounding rules.
  5. 05
    List the latency hierarchy and two designs the numbers let you reject in a sentence.
Recap

If you could reconstruct each answer from memory, you hold the unit’s spine: Little’s Law (λ = L/W) gives a throughput ceiling from concurrency and latency; the queueing curve (W = 1/(μ−λ)) makes the tail go vertical near a resource’s limit, so you keep 30–40% headroom; vertical scaling is simple but capped, costly, and a SPOF while horizontal is elastic but needs statelessness and pays the USL coordination tax (α plateaus, β goes retrograde); back-of-envelope estimation (four quantities, two cheat sheets, round to one figure, size for the peak) finds which resource breaks first; and the latency hierarchy (RAM ~100 ns to cross-region ~150 ms) lets you reject a disk-per-request, cross-region, or N+1 design in a sentence. The thread tying them together: at this depth, scaling is arithmetic and physics you reason about up front, not behavior you discover in production.

Something unclear?

Ask a question about this lesson. Questions are anonymous and go straight to the author to make the lesson better.

shortcuts expand
search
K
prev piece
k
next piece
j
cycle tier
t
this menu
?
sources3
expand
  1. 01
  2. 02
  3. 03

Trademarks belong to their respective owners. Editorial reference only.