AWS Consulting in New Zealand

Ask for the monthly running cost before the project cost

The consulting fee is one-off. The AWS bill is forever, and it is usually the larger number over three years.

A serious proposal states what the environment will cost to run each month, what that assumes about traffic and storage, and what it looks like at twice the volume. Ask for the workings from the pricing calculator, not a figure.

Then check what is included, because these are the lines people forget:

• Data transfer out. Traffic leaving AWS is charged, and it surprises people running anything media-heavy.
• Non-production environments. Test and staging running around the clock often cost as much as production.
• Backups and snapshots, which accumulate quietly.
• Support plan, which is a percentage of spend and is worth having.
• NAT gateways and load balancers, which are charged hourly whether or not anything is using them.

Auckland, Sydney, or somewhere else

AWS has a New Zealand region, and it is a genuine option rather than an obvious answer.

It is right when latency to New Zealand users matters, when a customer or a government agency requires data to stay onshore, or when your own policy commits you to it.

It is not automatically the cheapest. Pricing varies by region and Sydney is frequently less expensive; newer regions also receive some services later than established ones.

Ask each consultant which region they propose, why, and what the monthly difference is. If onshore is a hard requirement for you, say so in the brief — it changes the design and the price, and it should not be discovered later.

Insist on infrastructure as code

An environment built by clicking through the AWS console works and is a liability. Nobody can reproduce it, nobody can review a change, nobody can tell you what was different last Tuesday, and rebuilding it after a serious incident means somebody remembering.

Terraform or AWS CDK describes the whole environment in files that live in a repository you own. That gives you review, repeatability, a real disaster recovery position, and — not incidentally — the ability to change consultant without starting again.

If it is not in the quote, ask why. A consultant who prefers the console is telling you something about how the engagement ends.

Get the account structure right early

This is boring, it is cheap now, and it is painful to retrofit.

• Separate accounts for production and non-production, so a mistake in test cannot touch live.
• The root account locked away with multi-factor authentication, used by nobody day to day.
• People and services with only the permissions they need.
• Billing alerts at thresholds you choose, going to a person who will act.
• Tags on everything, so the bill can be read by project or by team.

That last one turns "AWS costs us $6,000 a month" into "the reporting environment costs us $1,900 a month", which is the sentence that lets you make decisions.

A cost review is usually the best first engagement

If you are already on AWS and the bill has been climbing, buy the review before anything else. It is the highest-return work in this category and it regularly pays for itself within a quarter.

What it reliably finds:

• Instances sized for a peak that never arrives.
• Test and staging environments running at 3am and at the weekend.
• Old snapshots and unattached storage volumes nobody has deleted.
• Data transfer between regions that a design change removes.
• Savings Plans or Reserved Instances not taken up on workloads that have run unchanged for two years.
• Services someone enabled for a trial and forgot.

Ask for the output as a prioritised list with a dollar figure and an effort estimate against each item, so you can choose. A report without numbers is an opinion.

Migration: the copying is not the hard part

Moving data and standing up servers is routine. What decides whether a migration goes well:

• A tested cutover plan with a rollback, rehearsed at least once.
• Knowing what depends on what — the integration nobody remembered is the usual cause of a bad Monday.
• DNS and certificates handled deliberately, with time-to-live reduced beforehand.
• A decision on what is not moving, and what happens to it.
• Someone watching for a week afterwards, not just for a weekend.

Ask how many rehearsals are in the quote. One attempt on the night is not a plan.

Know how you would leave

Not because you intend to, but because the answer tells you how tightly the design binds you. Virtual machines, containers and standard databases move without much drama. Deep use of proprietary managed services is genuinely convenient and genuinely harder to unpick.

That is a trade worth making deliberately rather than by accident. Ask which parts of the proposed design would be difficult to move, and make sure somebody in your business heard the answer.

Comparing five quotes

Read for:

• Who gave a monthly running cost with their assumptions?
• Who proposed infrastructure as code and offered you the repository?
• Who asked about onshore requirements before choosing a region?
• Who included monitoring, backups and a tested restore rather than only the build?
• Who explained what happens to root credentials and access when the engagement ends?

Describe what you run today, where it runs, what your current bill is if you have one, and any requirement about where data sits. Consultants who do this work will ask about the bill and the dependencies before they ask about architecture.