Stop 08 · Serverless & Event-Driven ArchitectureTopics · 9/9

When Serverless Isn't the Right Fit

Part of the Cloud Computing Roadmap

In short

Serverless is a great default for many things, but not everything. Knowing when to say "this needs a container or a VM" is what interviewers listen for.

Long-running processes, workloads needing consistent low latency, or applications requiring fine-grained infrastructure control are usually better served by VMs or containers than serverless functions.

The blob character stares at a very long receipt beside a parked scooter.

Think of it like this

Cab every day, or a scooter?

Booking a cab for a weekend trip makes sense. Booking one for your daily 9-to-6 commute, every single day, adds up to a scary monthly total. At some point, buying a scooter is the smarter deal.

Watch it happen

Pick by workload shape

Serverless fits

  • Spiky or unpredictable traffic
  • Short jobs, seconds to minutes
  • Event glue between services

Containers or VMs fit

  • Steady high load all day
  • Jobs longer than 15 minutes
  • WebSockets, GPUs, custom OS needs

Match the tool to the workload. Many real systems mix both: containers for the core, functions for the edges.

The big ideas

What you need to know

Time limits

Lambda stops at 15 minutes. Long video encoding or ML training belongs on containers, Batch or VMs.

Steady load costs

At high, constant traffic, per-request pricing can overtake a right-sized always-on server. Do the maths.

Latency-critical paths

If every millisecond matters, cold starts and network hops to managed services can hurt.

Lock-in and control

Triggers and event formats are provider-specific. Some teams need more portability or OS-level control.

Save yourself the pain

Common mistakes

Oops

Splitting a long job into chained functions just to dodge the 15-minute limit.

Do this instead

If you're fighting the limit, it's the wrong tool. Use a container task, Step Functions or Batch.

Oops

Choosing serverless because it's trendy, without checking traffic shape.

Do this instead

Look at expected requests per second over a day. Flat and high often favours containers.

Your mission

Back-of-envelope cost check

~30 min

Quick check

Which job is the worst fit for a FaaS function?

Go deeper

  1. 1List 3 workload types that would fit poorly into a serverless function's execution model.
  2. 2Learn why a long-running WebSocket connection typically doesn't fit a standard FaaS model well.
  3. 3Practice matching workloads to serverless versus container/VM compute honestly, not defaulting to hype.

More topics at this stop

Stuck on this topic? Ask an Insider

Get 1:1 guidance from people who've walked this exact path — on the InsideEdge app.

Download
InsideEdge