Abhijit Mazumder

LAB 01 / MIX-AXIS EXPLORER

Freeze the box.
Sweep the mix.

The stock benchmark freezes the workload. This instrument holds the measured system fixed and lets you change the question an operator actually faces.

LIVE MODEL / MIX-AXIS EXPLORER

What happens when the workload writes more?

Move the mix. The curve is a two-parameter capacity model; the three points are measured runs.

20%
16 read units4 write units

Each tile represents 5% of the operation mix—not a row, request timeline, or individual latency.

MODELLED CAPACITY80,409operations / second
RELATIVE COST / OP1.27×read cost = 1.00×
MODEL STATUSAnchor runMeasured: 80,409 ops/s
The capacity model predicts lower throughput as the write fraction increases. The measured points are provided in the table below.

MODEL T(x) = K / ((1 − x)r + xw)

FIT w / r = 2.33

ANCHOR 20% writes = 80,409 ops/s

Claim boundary. The tiles explain the mixture; they do not depict Cassandra internals. The dots are measured. The curve between and beyond them is modelled.

Read the measurement →

WHAT IT SHOWS

Cost is inferred from capacity.

The model treats a workload as a weighted mixture of read and write costs. Increasing the write share raises average operation cost, so the modelled throughput falls. The measured points let you see where that simple model holds and where reality bends away from it.

WHAT IT DOES NOT SHOW

A curve is not a mechanism.

The fit does not prove why writes cost more, and it does not claim that every individual write has 2.33 times the latency of a read. Mechanism claims need configuration, traces, logs, and further measurements.