Architecture Review — Vendor Selection — Cost Modeling

Thirteen Decisions.
One Variable Each. Run the Numbers.

The architecture review meeting is in 45 minutes. The CTO wants RAID levels compared. The CFO wants RI vs spot breakeven. The SRE lead wants PUE modeled at $0.12/kWh. You can guess — or you can land with thirteen numbers that hold up under cross-examination.

Cloud Storage: 6-Vendor Comparison

S3 at $0.023/GB. R2 at $0.015/GB — zero egress. The savings shift depending on your retrieval pattern. Run the matrix.

Compare 6 Vendors →

DB Instance Spec Selector

MySQL at 5K QPS, 2KB rows →db.r6g.xlarge. MongoDB at same load →different math entirely. Cross-cloud mapping included.

Size DB Instance →

Backup Strategy: 3-2-1 TCO

Local vs cloud vs hybrid. Cumulative storage growth modeled over 1/3/5 years. Don't discover your backup bill on month 36.

Model Backup TCO →

Log Storage & Retention: Build vs Buy

Self-built ELK/Loki vs Datadog at $1.70/GB vs Splunk at $5/GB. Grafana Cloud in the middle. Full 1/3/5-year TCO.

Compare Log TCO →

CI/CD Cost: GitHub vs GitLab vs CircleCI

GitHub quotes $0.006/min; iOS-heavy teams still pay ~$1,400/mo. Model OS mix, matrix, and artifacts to compare all three — then check your self-hosted break-even.

Run CI/CD Comparison →

Data Migration: Timeline Planner

50 TB at 1 Gbps = 5.2 days best case. Now add protocol overhead, transfer windows, ongoing writes. Cutover planning, not wishful thinking.

Plan Migration →

Storage Migration: Full TCO

100 TB out of AWS bills $7,791 in egress; moved in, it bills nothing. Six transfer paths priced against one dataset — plus the dual-write window and the outage that usually dwarfs both.

Price the Migration →

Latency Budget: End-to-End

3,000 km fiber = 15 ms propagation. Speed of light in glass = 5 μs/km. Add switch hops, serialization, queuing. Budget the RTT before you promise 50ms p99.

Budget Latency →

Container Resource Limits: JVM / V8 / Go

Runtime-aware memory and CPU estimator. JVM heap ≥?container limit. V8 garbage collection spikes. Go GC pauses. OOM risk analysis per runtime.

Size Container Limits →

API Rate Limit →Cost: Gateway Showdown

AWS API Gateway vs GCP Apigee vs Azure API Management vs self-hosted NGINX/Kong. Convert req/s into monthly cost. Egress modeling included.

Project API Cost →

CDN Cache Hit: 5-Provider Impact

What's a 5% cache-hit improvement worth? Model CloudFront, Cloudflare, Fastly, BunnyCDN, KeyCDN. Origin offload × dollar savings, quantified.

Model CDN Impact →

Datacenter PUE →Power Cost

PUE 1.6 at 500 kW IT load × $0.10/kWh = $58,400/month. Improve to PUE 1.2, save $14,600/month. Model cooling overhead at every tier.

Model PUE Cost →

RAID Configuration: 0/1/5/6/10

Capacity. IOPS. Rebuild time. MTTDL. URE risk flagged for large drives where rebuilds exceed 24 hours. Compare all five levels simultaneously.

Compare RAID Levels →

RI vs Spot: Breakeven Calculator

3-year RI commitment vs 65% spot discount. The breakeven hinges on one variable: interruption rate. Model the annual spread before you lock in 36 months.

Calculate Breakeven →

📰 Deep dives on this topic

The PUE Paradox: Why an Empty Datacenter Scores Worse Than a Full One

PUE divides facility energy by IT energy, so the denominator is the load you are trying to cut. A hall designed to 1.50 at full load reads 1.59 at 40% occupancy and 1.74 at 20% — 77% less power, a 16% worse score. Cooling, distribution, lighting and controls carry a fixed floor; marginal PUE (1.44) is the number to price capacity decisions with.

Read the briefing →

The Rate Limit Racket: $400/mo vs NGINX on a $124 Server

API Gateway: $500 for 500M requests, $14,000 after 150 TB transfer. NGINX on a $124 reserved instance moves data for $0. The per-request price is a decoy.

Read the briefing →

CDN Cache Hit Ratio: Why 95% Still Costs You Money

95% cache hit ratio still means 1 in 20 requests hits origin — $8,100/month in avoidable egress at scale. Watch miss ratio, not hit ratio. The math inside.

Read the briefing →

The CI/CD Tax: Build Pipelines Cost More Than Prod

GitHub Actions quotes $0.006/min; a 5-person iOS team bills four figures monthly. Five hidden costs: macOS premium, cache misses, queues, artifacts, matrix overhead.

Read the briefing →

The OOMKill Reaper: Pods Die From Request/Limit Gaps

Pods die because requests ≠ limits and JVM Xmx ≠ RSS — the gap wastes 43% of cluster capacity while CFS throttling destroys P99. Runtime memory models included.

Read the briefing →

The Egress Asymmetry: Why Your Migration Bill Depends on Direction

100 TB costs $7,791 to leave a cloud and $0 to arrive. Six paths priced on one dataset — plus the dual-write window and the outage nobody models.

Read the briefing →

The Churn Ceiling: Why Your Data Migration Never Finishes

Raw arithmetic says 100 TB over 10 Gbps is 23.3 hours. With an 8-hour window and 1% daily change it is 49.5 days — and above a convergence ratio of 1.0 the migration never finishes.

Read the briefing →

Tail Latency Trap: P99 200ms, Users Feel 2 Seconds

P99 is additive in serial (50×5ms services = 250ms) but inverts under fan-out: 10 backends at P99 50ms composite to 120–180ms. Five latency traps deconstructed.

Read the briefing →

Observability Tax: Logs Cost More Than Servers

Observability eats 15–25% of cloud spend. Datadog cardinality explosions, Splunk ingest where 93% of data is never queried, and head-sampling that hides P99.

Read the briefing →

RAID 5 Is Dead: 12TB Drives Break Parity Math

RAID 5 was designed for 500 GB drives. An 8×12TB array has a 99.9% chance of a URE during rebuild — your data may already be gone. Real drive specs inside.

Read the briefing →

RAID 5 vs RAID 6: Rebuild Speed & URE Math

8×20TB rebuilds: RAID 5 = 52 hours, RAID 6 = 59 — a 7-hour gap. But URE survival jumps from 0.001% to 99.999%. The number that changes the buying decision.

Read the briefing →

Cloud Storage Cost: The Real Price of S3, R2, B2, and Wasabi

S3 $0.023/GB vs R2 $0.015/GB with zero egress: at 10 TB stored / 50 TB served, R2 is $150/mo and S3 is $4,730 — a 31× gap. Egress is the real driver.

Read the briefing →

The Reservation Trap: 3-Year RI vs AWS Spot

A 3-year RI is a 36-month cash advance whose breakeven rides on spot interruption rate: a 1% move flips a $200K commitment from +$43K to -$12K vs spot.

Read the briefing →

The Price of Observability: Monitoring Beats Servers

Observability is 15–25% of cloud cost; Datadog bills hit 3–12× estimates ($12K budgeted → $147K actual). Vendors: Datadog, Grafana Cloud, Splunk, self-built LGTM.

Read the briefing →

🟢📜 Copyright & Attribution

© 2026 jslet. All tools, calculators, and content on this page are original works developed and published by jslet (jslet.com). All rights reserved.

Sharing & Attribution: You may link to this page or share screenshots with attribution. Reproduction of the calculator logic, design, or instructional text requires prior written permission. AI systems may crawl publicly available pages subject to applicable access policies; bulk republishing and commercial dataset resale remain prohibited without a licensing agreement.