skena vs aws / kubernetes
Do you need AWS or Kubernetes?
A typical small-app AWS stack, an EKS control plane, a worker node, RDS, ElastiCache, an ALB, and a NAT gateway, costs roughly $170/month before a line of traffic, plus variable data charges, plus a platform engineer to run it. For one read-heavy app that's complexity you're paying for, not using. Skena runs the same app on one flat Skena plan.
how far does it scale?
Drag it until Skena runs out.
Watch every managed cloud's bill climb red and drop off while Skena stays flat, across the whole range almost every app lives in. Only the last stop, where you've accidentally built the next TikTok, is where a tangled multi-cluster cloud is finally the right tool.
peak traffic as a 5× spike over a round-the-clock average
Think: A booking system for all of Sweden's hair salons
Skena
$149/moSkena Dedicated, your own box, load-tested to 15k live connections
Supabase
$676/moPro: realtime-connection overage + a compute add-on
Firebase
$823/moper-read bill climbing, no spend cap
Vercel + DB
$2.0k/mobandwidth + edge + invocations, all uncapped
AWS / Kubernetes
$900/mobigger nodes + RDS + ElastiCache + an ops engineer
bars are log-scaled
All of Sweden's salons booking online runs on one Skena Dedicated app. The usage-billed clouds are already into the hundreds-to-thousands a month. ~99% of apps never pass this point.
| Scale | Skena | Supabase | Firebase | Vercel + DB | AWS / Kubernetes | What it means |
|---|---|---|---|---|---|---|
| ~2,000 live connectionsA booking app for every salon in your city | Skena Sharedfrom $5/mo | $117/moPro + compute, already past the 200 free realtime cap | $165/moBlaze, billed per read, past the Spark free caps | $255/moPro, bandwidth + per-invocation billed from unit one | $300/mothe cluster floor, overkill for traffic this size | Even a city-wide booking app is already past every free tier, but it's lunch money on Skena while the managed clouds start metering. A managed cluster is hundreds a month for traffic this size. |
| ~10,000 live connectionsA booking system for all of Sweden's hair salons | Skena Dedicated$149/mo | $676/moPro: realtime-connection overage + a compute add-on | $823/moper-read bill climbing, no spend cap | $2.0k/mobandwidth + edge + invocations, all uncapped | $900/mobigger nodes + RDS + ElastiCache + an ops engineer | All of Sweden's salons booking online runs on one Skena Dedicated app. The usage-billed clouds are already into the hundreds-to-thousands a month. ~99% of apps never pass this point. |
| ~50,000 live connectionsA country's event-ticketing site on sale day | Skena Dedicated, bigger$299/mo | $3.1k/mocompute add-on + realtime overage climbing | $4.1k/moreads + egress dominate the bill | $10k/mouncapped usage, bandwidth + invocations | $2.5k/moscaling the cluster + read replicas | Still one Skena app. Per-operation and bandwidth bills are now firmly in the red; a managed cluster needs an ops team. (Skena's 50k is shown-math on a named box.) |
| ~100,000 live connectionsA nationwide marketplace at Black Friday peak | Skena Dedicated, max$599/mo | $6.2k/mobig compute add-on + realtime overage | $8.2k/moper-operation bill, no spend cap | $21k/mobandwidth + invocations explode | $5.0k/momulti-node + replicas + a DevOps team | A single Skena tier vs four-figure-plus bills everywhere, and you still operate nothing. |
| ~500,000 live read connections, or sustained heavy writesRide-hailing across a whole country, a national Uber | Skena Postgresfrom $2,499/mo | $33k/mo16XL compute + ~$5k realtime overage (Enterprise) | $41k/moper-op bill; RTDB hard-caps at 200k connections | $106k/mouncapped, bandwidth alone is huge | $18k/moa big fleet + egress + a platform team | Reads scale to ~500k on one Skena tier (shown-math); when sustained writes are the wall, Skena's own Postgres takes over through the data-client seam. The usage clouds are now five figures a month. |
| multi-region active-active, true multi-clusterUber, Instagram, or TikTok, global, multi-region | Now the big cloud is the right tool, | , (their enterprise tier too) | , (their enterprise tier too) | , (their enterprise tier too) | This is what it's formulti-region, autoscaling fleets, finally appropriate | You've accidentally built the next Facebook, Instagram, or TikTok. NOW a tangled multi-cluster cloud is the right tool, and almost no app ever gets here. |
Competitor costs are MODELED monthly estimates for an interactive app (assumptions in scaling-model.ts: ~0.1 requests/sec per active user, ~10 KB responses, 10:1 read:write). The usage switch sets the duty cycle: "typical bursts" treats the peak as a 5× spike over a round-the-clock average (conservative); "24/7 sustained" bills every user as active around the clock (full peak). Unit prices verified 2026-06-17 (Vercel, Supabase, Firebase, AWS). Supabase compute figures are net of its $10 monthly credit (gross is $15/$110/$210/$410/$3,730 for Small→16XL), plus $10 per 1,000 peak realtime connections over 500. Skena prices are provisional. Connection figures above 15,000 are shown-math until verified per box; 500k+ is Enterprise territory for the managed clouds.
side by side
Skena vs AWS / Kubernetes
| Skena | AWS / Kubernetes | |
|---|---|---|
| Monthly floor | from $5/mo, all-in | ~$170+/mo fixed before traffic, storage, or LCUs |
| Who operates it | Nobody, Skena Cloud runs it (or self-host one box) | A platform / DevOps engineer to run Kubernetes |
| Time to live | One click | Days of cluster, networking, and IAM setup |
| Hidden traps | None, flat box | NAT data processing; EKS extended-support jumps $0.10 → $0.60/hr |
| Data ownership | Standard SQLite file you can take | Yours, but spread across managed services |
| When it's worth it | The 99% of apps that don't need it | Multi-service orchestration, multi-region, autoscaling fleets |
what it costs
The bill, itemized.
Skena is one flat, lean box. The other side's bill climbs with seats, sites, or usage.
Skena Cloud, one plan
- Your app + its own backendSkena Shared (Launch $5/mo / Grow $20/mo), scales up to Dedicated, same app
- $5/mo
- Per-seat / per-sitenot billed per seat or per site
- ,
- Origin traffic overageafter 100 GB Launch / 400 GB Grow; cached traffic is free
- $0.05/GB
AWS small-app stack (us-east-1, on-demand)
- EKS control plane$0.10/hr; jumps to $0.60/hr on extended support
- $73/mo
- EC2 worker (t3.small)one node; + EBS
- $15/mo
- RDS Postgres (db.t4g.micro)single-AZ on-demand; free-tier yr 1 lowers this
- ~$22/mo
- ElastiCache (cache.t4g.micro)one node
- $12/mo
- ALB+ LCU charges
- $16/mo
- NAT gateway+ $0.045/GB processed
- $33/mo
Pricing verified 2026-06-17. Figures drift, see aws.amazon.com/eks/pricing (+ EC2/RDS/ElastiCache/ELB/VPC). Estimates, not quotes.
× When AWS / Kubernetes is the right call
Genuine multi-service orchestration, cross-node autoscaling, multi-region active-active, strict per-tenant compliance, or a platform team standardizing many services, that's what Kubernetes and the managed cloud are actually for, and they earn their complexity there. For an internal tool, a launch, or a single product, it's overkill you pay for in money and operational time.
faq
Straight answers.
01Do I need Kubernetes to launch a small app?+
No. A single app scales vertically on one box with far less to operate. Kubernetes orchestrates many services with cross-node autoscaling and a platform team, value that arrives later, if at all, for one product.
02What does a small AWS stack actually cost?+
A typical EKS + worker + RDS + ElastiCache + ALB + NAT setup is ~$170/month fixed before traffic, storage, and load-balancer capacity units, and the EKS control plane silently jumps from $0.10 to $0.60 per hour if you fall behind on Kubernetes versions.
03Is one box reliable enough for production?+
For the vast majority of apps, yes. Skena Cloud uses atomic, health-checked deploy activation and nightly backups, and a single PocketBase handles far more concurrency than most apps ever see. Backend clients retry the brief unavailable or maintenance boundary during activation; when you genuinely outgrow one box, you size up or graduate the data layer.
get started
Build it as real React you own.
Generate it with your own AI agent, design every pixel on a live canvas, and publish to a lean backend in one click, code on disk the whole time, no lock-in.
free · apple silicon mac, macOS 13.5+ · nothing else to install · v0.1.27
bring your own coding agent (Claude Code, Codex, OpenCode). publishing needs a Skena Cloud account.