Data Systems & Infrastructure

Database Administrator

Keep the system of record correct, fast, and recoverable.

This is usually a second role — most people arrive from support, systems or networking.

What you actually do

  • Tune queries and indexes so an application stops timing out
  • Design and test backup and recovery for data nobody can afford to lose
  • Plan capacity and growth before it becomes urgent
  • Manage replication and failover
  • Control who can read and change what, at the data layer

Technologies you may touch

Nobody uses all of these. Which ones depends entirely on the employer.

  • SQL The language for asking a database a precise question.
  • Windows Server The Microsoft server platform running file, directory, application and print services in most enterprises.
  • Linux The operating system most servers, network devices and cloud workloads run on, driven from a command line.
  • Microsoft Azure Microsoft’s cloud — deeply tied to Windows and Entra ID identity, which is why enterprises land there.
  • AWS The largest cloud platform — servers, networks, storage and identity created through an API instead of a purchase order.
  • PowerShell Microsoft’s scripting language — the standard way to do something to a thousand Windows machines at once.

Concepts you will learn

These transfer between employers and outlast any particular product.

  • Indexing and query plans
  • Transactions and consistency
  • Replication
  • Point-in-time recovery
  • Data retention

Where this work happens

Financial services Healthcare Retail and e-commerce Any organisation with a critical application

You may enjoy this if…

  • You like precision and being certain rather than probably right
  • You enjoy performance work with measurable before-and-after
  • You are comfortable owning something where mistakes are permanent

The honest downside

Recovery testing is unglamorous and skipping it is how careers end. On-call is common. The role is narrowing on-premises as managed cloud databases absorb parts of it.

What it pays

Pay is the thing people most often get misled about, so here is where every number comes from — including the ones that are not really numbers.

Official occupation

Median for the occupation

~$104,000

Database Administrators (15-1242)

What you actually start on

~$60,000–$80,000 to start

Outlook

About +8%, though managed cloud databases are absorbing parts of the role

This role maps cleanly onto a published government occupation, so the median below is a real statistic you can go and check.

Maps to Database Administrators (15-1242). The on-premises version of the role is narrowing as managed cloud databases absorb parts of it; database engineering and reliability work is where it is heading.

A median is the midpoint of everyone employed in that occupation, including people decades into their careers — so starting pay runs materially lower. Verify against the live BLS page before making a decision.

Source: U.S. Bureau of Labor Statistics · BLS OES, May 2024 release · figures are national and United States–specific; pay varies substantially by region and employer. Nothing here is adjusted for cost of living.

CertBlueprint earns affiliate commission on some study resources. It earns nothing from Database Administrator salaries, and no role is ranked, recommended or presented more favourably because of what it pays.

Where this leads

IT careers are a graph, not a ladder. These are the moves people actually make from here — each one reuses most of what you already know.

See where Database Administrator sits on the whole map →

Data Engineer

The pipeline-building direction, with more code.

Systems Engineer

The broader infrastructure the database sits on.

Often confused with

Titles overlap heavily in IT and tell you very little. Each of these puts Database Administrator beside another role on the same dimensions, with the practical differences underneath.

vs Systems Engineer vs Data Engineer

Try it this weekend, before you spend anything

Reading about work and doing it are different. Each of these is free, runs on the machine you already have, and takes under an hour and a half. Finding out you dislike it is a genuinely useful result — and far cheaper here than after an exam voucher.

Try the work · 60–90 minutes

Ask data a question

Free

Take a messy public dataset nobody prepared for you, find one true thing in it, and say that thing plainly — including what it does not prove.

Nothing to install · no account · open the steps

Before you start

Use openly published data, and check the licence — most public portals permit reuse with attribution, some do not. Avoid datasets containing personal information about identifiable people; if you use your own records instead, such as a bank export, keep them local and do not paste them into an online tool to "quickly clean them up". Handling data carefully when nobody is watching is the actual professional standard here, not a formality.

A spreadsheet is enough — Excel, Numbers, or free LibreOffice. There is an optional SQL path at the end if you want it. No account needed.

  1. Get something real and unhelpful

    Download a CSV from a public data portal — data.gov.uk, data.gov, or your national statistics office. Pick a subject you actually care about: crime near you, rainfall, energy prices, hospital waiting times. Deliberately do not pick a tutorial dataset. Real data is messy, and the mess is the job.

  2. Write your question down before you look

    One sentence, on paper, before you open the file. "Has burglary in my area gone up or down over five years?" Doing this first is what stops you finding whatever the data happens to show and calling it a finding afterwards.

  3. Meet the mess

    Open it. Find the problems: blank cells, dates in three formats, a total row hiding at the bottom that will corrupt every average you take, categories renamed halfway through the period. Write a list of every defect you find before you fix any of them.

  4. Clean it, and keep a record of what you changed

    Fix the defects — but write down each decision as you go. "Dropped 214 rows with no date." "Treated blank as zero." Those decisions change the answer, and being unable to say what you did to the data is the difference between analysis and guessing.

  5. Answer the question

    Pivot table, chart, or a formula. Get to a number. Then do the part almost everyone skips: try to disprove it. Is the trend just one anomalous year? Did the collection method change? Would a different but equally reasonable cleaning decision have flipped the result?

    Any
    Optional SQL path — sqlite3 ships with macOS and Linux:

sqlite3 data.db
.mode csv
.import yourfile.csv t
SELECT year, COUNT(*) FROM t GROUP BY year ORDER BY year;

  6. Say it in three sentences

    What you found, how confident you are, and what it does not tell you. Show it to someone who has not seen the data. If they ask a question you cannot answer, that is not a failure — that is you discovering the limit of your own finding, which is the most valuable output of the whole exercise.

This suits you if…

  • The cleaning was satisfying rather than infuriating
  • You tried to disprove your own finding without being told to
  • You wanted a second dataset to cross-check it against
  • Stating the limits felt like rigour rather than weakness

If you disliked it

If the cleaning felt like pointless drudgery blocking the interesting part, be honest about that — because in real data work the cleaning IS most of the job, frequently eighty percent of it. Enjoying the answer but not the preparation usually points toward analyst-adjacent roles that consume prepared data rather than data engineering, which is almost entirely preparation.

Try the work · 60–90 minutes

Walk the physical layer

Free

Stop treating the network as an abstraction. Measure a real radio signal in a real building, find out where it dies and why, and put your hands on the hardware everything else runs on.

Nothing to install · no account · open the steps

Before you start

For the hardware section: shut down fully, unplug from the mains, and hold a metal part of the case before touching components. Never open a power supply or a CRT — they store lethal charge even unplugged. Do not open equipment you rent, lease, or that belongs to an employer, and be aware that opening some laptops voids warranty. If in any doubt, do the wireless half and skip the rest; it is the smaller part of the trial.

Built-in wireless tools plus your own feet. The optional hardware section needs a computer you own and are willing to open — no purchase required.

  1. Measure what you have been assuming

    Find your current signal strength, right now, where you are sitting. Note the number and the units — dBm is a negative figure where closer to zero is better, and −50 is excellent while −80 is barely usable.

    Windows
    netsh wlan show interfaces
    macOS
    Hold Option, click the Wi-Fi menu — or open Wireless Diagnostics from Spotlight
    Linux
    iw dev wlan0 link    # or: nmcli -f IN-USE,SSID,SIGNAL,CHAN dev wifi

  2. Walk the building and write numbers down

    Take a reading in every room, and at the far corner of each. Write them on a rough floor plan. You are producing a coverage survey — the same artefact, at smaller scale, that a wireless engineer produces before a hospital or a warehouse deployment.

  3. Find the wall that eats your signal

    Look for the sharp drop, not the gradual fade. A gentle decline is distance; a cliff between two adjacent readings is something physical — a wall with foil-backed insulation, a mirror, a boiler, a fridge, a chimney. Go and identify the actual object. This is the moment radio stops being theoretical.

  4. Discover you are not alone on the air

    List the networks around you and the channels they occupy. Count how many are on the same channel as yours. Wi-Fi is a shared medium in a public commons — your neighbour's router degrades yours, and no configuration on your side can fully fix it. That constraint is the defining fact of wireless work.

    Windows
    netsh wlan show networks mode=bssid
    Linux
    nmcli -f SSID,CHAN,SIGNAL dev wifi list
    macOS
    Wireless Diagnostics → Window → Scan

  5. Trace one cable end to end

    Pick any cable in your home — the one from the wall to the router. Follow it physically to both ends. Where does it enter the building, and what is on the other side of that wall? Almost nobody knows. Documenting exactly this, across a building, is a real job that real people are paid to do because the records are always wrong.

  6. Optional — open something and name every part

    Power off, unplug, ground yourself, then open a desktop or an old laptop you own. Identify: the processor and its cooler, the memory, the storage, the power supply, the fans. Trace where the air enters and leaves. Ask where the heat goes. Now imagine ten thousand of these in one room — that question of heat, power and airflow is most of what data centre design actually is.

This suits you if…

  • Finding the exact wall that killed the signal was satisfying
  • You enjoyed being on your feet with a number in your hand
  • The shared-medium constraint interested you rather than annoying you
  • You wanted to open the case more than you wanted to read about it

If you disliked it

If the physicality felt like a chore — the walking, the writing down, the fiddling with hardware — that is worth knowing before pursuing wireless, field or data centre work, all of which are substantially physical jobs in real buildings, sometimes at height, in the cold, or at 4am. Plenty of superb network engineers want nothing to do with the physical layer and work entirely in routing and policy instead.

Certifications that fit this path

These come last for a reason. A certification is evidence for a direction you have already chosen — it is not the direction itself.

Adjacent Microsoft PL-300 Useful if reporting and analysis is part of your remit. Core Microsoft AZ-104 Managed database services increasingly live here. Foundation CompTIA Linux+ Most open-source databases run on Linux.

Not sure whether this is actually you?

Reading about a role and enjoying the work are different things. The scenarios show you the problems instead of describing them, and tell you which ones you keep reaching for.

Try the scenarios →