Skip to content
Services · Platform Readiness Review

Platform Readiness Review.

An independent, fixed-fee assessment of your production Kubernetes platform, including how its components work together.

Opens a short form. An engineer replies within 1–2 business days, then sends a fixed-fee proposal.

  • Fixed fee
  • 2–4 weeks
  • Prioritized findings report

What it is

An evidence-based read on your platform, organized by priority.

The review gives your platform lead a defensible list of where to spend engineering time, with no partner tier or referral fee behind any recommendation. It is an assessment only: nothing on your platform changes while it runs, and what comes next is a separate decision, made with the report in hand.

Who it's for

  • Teams that inherited a platform from a departed team, an acquisition, or a previous vendor
  • Fast-growing organizations that want the risks found before growth exposes them
  • Teams about to hand operations to a partner, including us, that need an objective baseline
  • Engineering leaders who must show, with evidence, where to invest platform engineering time

No platform yet? Then start with an Implementation Project or a Managed Platform that includes the build.

What you receive

Clear findings and practical next steps.

Every finding cites the configuration, metric, or trace behind it, so anyone on your team can verify it and act on it.

  • Platform inventory

    Clusters, components, configuration, and dependencies, as observed.

  • Findings report

    Each finding with its evidence, impact, and priority.

  • Prioritized recommendations

    What to fix first, with the options and their trade-offs compared.

  • Closing meeting

    A walkthrough for your team and leadership, recorded on request.

Sample finding

From the findings report · Request paths

Illustrative example

Retries at three layers can turn one slow dependency into an outage.

Evidence

The ingress, the mesh, and the checkout service's HTTP client each make up to three attempts, so one request can reach the payments API 27 times. The ingress gives up first, after 15 seconds.

ingress configmesh policytraces

Impact

When the payments API slows, retries multiply its load until it fails outright, while it keeps working on requests users have already abandoned.

Priority

High

Checkout carries revenue, and the fix is configuration only, so it fits the next scheduled maintenance window.

Recommended next step

Let only the mesh retry, with a retry budget, and give each layer a shorter timeout than the one in front of it. Prove it with a fault-injection test.

Alternative consideredRetrying in each client library instead gives finer control per call, but every service team then maintains its own settings. We recommend it only for calls that leave the mesh.

What we look at

Three lenses on one connected system.

How the platform is built, how it is operated, and how healthy it runs, then how its components affect each other. Organizational and process review is a separate engagement.

Lens 1

Architecture and capacity

Whether the design will hold as workloads and teams grow.

  • Cluster topology, capacity, and version posture, including end-of-life dates
  • Networking: CNI, ingress, service mesh, DNS, traffic management
  • Workloads: tenancy, resource requests and limits, scheduling
  • Storage: volume types, backups, data locality
  • Availability: replicas, anti-affinity, multi-zone readiness

Lens 2

Operations and delivery

How the platform behaves in real use: the technical mechanics, not team process.

  • Delivery: GitOps promotion, rollback, drift control
  • Observability: metrics, logs, traces, dashboards, alerts
  • Resilience: failure recovery, dependency mapping, blast radius
  • Lifecycle: upgrade cadence, add-on currency, deprecations
  • Infrastructure as code: declarative coverage, secret management

Lens 3

Platform health

The day-to-day condition of the platform and the standing cost of running it.

  • Cost and waste: utilization, over-provisioning, autoscaling, idle resources, GPU utilization where applicable
  • Security hygiene: RBAC, Pod Security, network policies, secrets, image sources
  • Configuration drift between the source of truth and the running clusters

Cost findings are recommendations, not a FinOps program; deeper cost work is scoped separately.

Across all three lenses

How the components interact

Many of the findings that matter most sit between components that each look healthy on their own.

  • Compatibility of versions and features, from Kubernetes and the CNI to the mesh, operators, and GPU drivers
  • Request paths from the edge through ingress, mesh, and policy to the workload
  • Overlapping controls: retries, timeouts, and policies that duplicate or contradict each other
  • Upgrade dependencies: what must change together, or in order
  • Cross-boundary failures that no single component reports

How we work

Evidence first, with context from your team.

Practical terms

Duration
2–4 weeks, fixed at the start. The exact length depends on the size of the platform and how much evidence there is.
Fee
A fixed fee, agreed up front and billed in KRW. No time-and-materials meter.
Access
Read-only access to clusters, IaC, observability, and pipelines. If your policy rules that out, a screen-share where your engineer runs agreed commands, or exported artifacts.
Your team
A kickoff, a few focused discussions on intent, history, and constraints, and the closing walkthrough.
Afterwards
The report stands on its own. If an Implementation Project or a Managed Platform starts within 90 days, part of the review fee is credited toward it.
  1. Discovery

    We agree how we'll see the evidence, in a way that fits your security policy.

    OutputAgreed access to the evidence

  2. Assessment

    We review the platform through the three lenses and trace how its components interact.

    OutputEvidence-backed observations

  3. Report

    We write up each finding with its evidence, impact, priority, and recommended next step.

    OutputThe findings report, grouped by lens and priority

  4. Discussion

    We walk your platform team and engineering leaders through the report.

    OutputAgreed priorities: what matters now and what can wait

Frequently asked

What teams ask before they start.

Does anything on our platform change during the review?

No. The review is an assessment only, run from read-only access or, where your security policy rules that out, a screen-share or exported artifacts. Any change comes later, if you decide on one, with the report in hand.

Is the review independent?

Yes. There is no partner tier or referral fee behind any recommendation. Where there is more than one sensible fix, the report compares the options and their trade-offs, and sometimes the best fit is the tool you already run.

How is the review priced?

As a fixed fee, agreed up front and billed in KRW. There is no time-and-materials meter. If an Implementation Project or Managed Platform engagement starts within 90 days of the review, part of the review fee is credited toward it.

Does the review cover GPUs and AI workloads?

Yes, where your platform runs them. GPU utilization is part of the platform-health lens, and GPU drivers and serving components are part of the compatibility checks across the stack.

Do you review our team's processes?

No. The review covers the technical mechanics of how the platform is built and operated, not team process. An organizational and process review is a separate engagement.

What happens after the review?

The report stands on its own, and many teams fix the findings themselves. If you want help, the findings map directly onto an Implementation Project or a Managed Platform engagement.

Two to four weeks to a clear picture.

Start with a fixed-fee Platform Readiness Review.

Request a proposal

Opens a short form. An engineer replies within 1–2 business days, then sends a fixed-fee proposal.