Backblaze B2 — Goodput Calculator (Neocloud)

▶ Configure Profile · Cluster · Pipeline · Flash · Tier

B2 50 Gbps · $6.95/TB·mo Flash: — —

1 · Workload profile — pick a scenario; it fills the fields below

—

2 · Your cluster — auto-filled by the profile; tune as needed

Cluster size (GPUs)

GPU $/hr

Model size

Checkpoint size (GB)

MLPerf anchors (default = 70B): ···

Checkpoint cadence

Concurrent jobs

3 · What’s in B2 — pipeline sizes (TB)

Training data (data lake)

Checkpoints (per run)

Models + RAG + KV cache

Logs & exhaust

4 · Flash economics — what flash costs, and how many tenants like this

Drive price ($/TB hardware)

Amort. (months)

DC overhead (%)

Sell price ($/TB·mo)

Tenants like this (× N)

1 = single tenant · raise to project the fleet

⚙ Advanced assumptions — goodput / stall / flash-displacement knobs. Defaults are conservative.

Hot working set (% of dataset)

default — varies by working-set size + access pattern

Flash $/TB·mo (displacement)

Std I/O wait %

spec default — measure on your training workload

Overdrive I/O wait %

modeled — pilot to validate

Overdrive flash-replace %

default — depends on read pattern + dataset shape

Epochs per run

Run duration (days)

Checkpoint write

Async stall capture %

modeled — async only

Streamed (uncached) %

modeled — feed lever

Cold % rehydrated / run

Training runs / month

I/O wait deltas default to 4.0% / 0.8% per spec §4.2 — validate against live customer telemetry before externalizing.

6 · B2 throughput tier — throughput · storage rate

▶ Tier comparison Per-tier load time, GPU productivity, and net ROI — derived from Configure pipeline sizes

Standard 50 Gbps

Reframe ROI — goodput · stall · flash displacement vs premium, by throughput level

Capability column shows what fraction of the cluster's read demand each throughput level can serve — measured against the prefetch demand (dataset × epochs × 8 / run-time), the same basis as the ⚡ keeps-fed tier. Goodput & Flash are prorated by capability; Stall scales with bandwidth directly.

THROUGHPUT	STORAGE RATE	CAPABILITY	GOODPUT	STALL	FLASH $	PREMIUM	NET ROI/MO

Net ROI = goodput + stall + flash displacement − tier premium vs Standard. Drives Configure → Cluster size / GPU $/hr / Checkpoint cadence / Hot working set. Active tier highlighted in cyan.

Watch the goodput math execute against a real B2 bucket

Click to start. Hero tiles fill in as each stage finishes.

skip · just populate the math

live event stream

[ waiting for events ]

LIVE COST PROJECTION

Open B2 console ↗

▸ SCALE 25 PB · 256 GPUs · 70B model · 1 hr ckpts

Demo validates the pattern — costs shown at your configured scale.

Goodput recovered / mo

$0.00

—

Checkpoint stall eliminated

$0.00

—

Flash capacity reclaimed

0 PB

—

New-capacity revenue / mo

$0.00

—

Total monthly impact

$0.00

goodput + stall recovered + net flash displaced

▸ Storage line (secondary)

B2 at scale

$0.00

per month

Flash at scale

$0.00

per month

You save

$0.00

—

Capacity reclaimed for resale

0% of workload

0 GB freed · run stages → —

ACTUAL DEMO BYTES

Stored: 0 B

Egress: 0 B

PUTs: 0

GETs: 0

Actual B2 cost of this run: $0.00

Estimate only — not a quote. Figures shown here are projections derived from the inputs you've configured, including estimated storage allocation percentages across workloads, and Backblaze list pricing as of June 1, 2026. Actual results may vary based on your specific workload, cluster utilization, contracted rates, migration scope, storage distribution, retention policies, access patterns, and implementation. This dashboard is for discussion purposes only and does not constitute a sales quote, pricing commitment, service-level agreement, or warranty. The I/O wait and flash-replacement defaults are working defaults pending pilot validation — instrument your own cluster and override the Configure fields before relying on these numbers for budget decisions. Final pricing and terms require a signed agreement.