Hands-on AWS knowledge base for developers Managed Ops
Talk to Us

DEA-C01 Exam Guide and Study Plan for Working Engineers

Export PDF Print-friendly layout with a copyright watermark; choose Save as PDF in the print dialog
DEA-C01 Exam Guide and Study Plan for Working Engineers

TL;DR

  • Ingestion & Transformation (34%) and Data Store Management (26%) are 60% of DEA-C01 — spend your hours in that ratio, not evenly across four domains.
  • It is $150 USD, 65 questions in 130 minutes, scaled pass at 720/1000. Budget 40–60 study hours if Glue, Athena and Redshift are already your day job; closer to 90 if they aren't.
  • Almost every storage question hides a second layer: Lake Formation grants, KMS key policy, or partition layout. Answering with "attach an IAM policy" fails.

The engineer who gets paged at 02:00 is the target audience

DEA-C01 is not a general cloud badge. It assumes you've already debugged a failed Glue job at an unhelpful hour, or argued with a Redshift query planner about a missing sort key. AWS's own prerequisite language is 2–3 years of data engineering experience and 1–2 years hands-on with AWS, and the questions read that way — they describe a broken or expensive pipeline and ask for the smallest fix.

If your job is mostly dashboarding in QuickSight over tables someone else built, the exam will feel abstract. Do 3–4 weeks of hands-on ingestion work first, or sit SAA-C03 instead and come back. If you own the pipelines end to end and the bill that comes with them, this is the most directly useful associate exam AWS has published.

The four domains and where the 34% actually lives

Domain Weight What actually shows up
1. Data Ingestion and Transformation 34% Glue jobs, crawlers and job bookmarks; Kinesis Data Streams vs Firehose vs MSK; DMS replication; EMR and EMR Serverless; file formats, compression, partitioning
2. Data Store Management 26% S3 layout and lifecycle, Data Catalog, Redshift distribution/sort keys and Spectrum, DynamoDB capacity modes, schema evolution
3. Data Operations and Support 22% Step Functions vs MWAA, CloudWatch metrics and alarms, DPU and worker sizing, cost tuning, failure triage
4. Data Security and Governance 18% IAM vs Lake Formation, KMS key policies, TLS in transit, PII discovery with Macie, row/column-level filtering, CloudTrail auditing

The weighting looks lopsided until you sit the exam and notice that security questions are baked into the other three domains rather than isolated in domain 4. A "why is this Glue job slow" question can hinge on a KMS-decrypt-per-object cost, and a "cheapest way to query S3" question can hinge on whether the caller has Lake Formation SELECT at all.

Domain 1 rewards knowing defaults cold: a Firehose delivery buffer won't flush under 60 seconds or 1 MiB, Glue crawlers re-classify a partition and can rename a column, and a Spark job writing 8,000 small Parquet files to S3 will make Athena miserable regardless of how correct the data is.

Booking: what $150 buys, and which exam goes first

Exam Fee (USD) Format Scaled pass Sit it first if
DEA-C01 $150 65 Q / 130 min 720 You already write ETL and touch Glue, Athena or Redshift weekly
SAA-C03 $150 65 Q / 130 min 720 Your job is VPCs, accounts and landing zones
DVA-C02 $150 65 Q / 130 min 720 Your day is Lambda, API Gateway and CI/CD

Fees are per attempt and vary by region and local tax; scheduling runs through Pearson VUE, either at a test centre or online-proctored. Certification is valid 3 years. AWS marks some of the 65 questions as unscored, so a handful of items you find bizarre don't count against you — don't let one bad question derail the next 20.

Order matters less than overlap. If you already hold SAA-C03, roughly a third of the networking and IAM material transfers, and DEA-C01 becomes a 5-week job. If you hold nothing, take DEA-C01 first only when data is genuinely your daily work; otherwise SAA-C03 gives you broader vocabulary that makes the data-domain questions easier to parse.

A six-week plan that survives on-call

Seven to eight hours a week, front-loaded on hands-on labs rather than video.

Weeks 1–2 are ingestion: build one Glue crawler plus a PySpark job that reads CSV, writes partitioned Parquet, and re-runs twice so you can see bookmarks work. Weeks 3–4 are storage and performance: the same dataset in Redshift with two different sort keys, then query both and compare. Weeks 5–6 are operations and security, plus two timed practice runs.

export BUCKET="dea01-lab-$(aws sts get-caller-identity --query Account --output text)"
aws s3 mb "s3://$BUCKET" --region us-east-1
aws s3 cp ./sample/orders.csv "s3://$BUCKET/raw/orders/" --recursive

aws glue create-database --database-input '{"Name":"dea01"}'
aws glue create-crawler --name dea01-crawler --role "$GLUE_ROLE_ARN" 
  --database-name dea01 
  --targets '{"S3Targets":[{"Path":"s3://'$BUCKET'/raw/"}]}'
aws glue start-crawler --name dea01-crawler
aws glue get-crawler --name dea01-crawler --query 'Crawler.LastCrawl.Status'


That last command returns `"SUCCEEDED"` after roughly 60–90 seconds on a small CSV. Then query it from Athena:

```sql
SELECT partition_0 AS dt, count(*) AS rows
FROM dea01.orders
GROUP BY 1 ORDER BY 1 DESC LIMIT 10;


The whole lab costs under $5 in us-east-1 if you tear down the crawler and bucket the same evening. Skip the Redshift cluster at your peril though — a single `ra3.xlplus` for 2 hours is the cheapest way to internalise distribution styles, and it is the one topic people consistently blank on.

## The 02:00 Lake Formation denial that taught me domain 4

On 2025-11-04 at 02:10 UTC, a nightly Glue job that had run for five months started failing on write. Nothing in the job had changed; a teammate had spent the previous afternoon registering the curated prefix in Lake Formation. Athena, used as a fast repro, returned this immediately:


Insufficient Lake Formation permission(s) on orders (Database: analytics, Table: orders)


My first configuration was wrong in an instructive way: I granted `SELECT` on the table to the Glue execution role and assumed that was sufficient because the role already had `s3:GetObject` on the bucket. It wasn't. The curated prefix had been registered as a Lake Formation location under a different role — the one that ran the registration — and Lake Formation vends temporary S3 credentials for that location's registering role, not for the caller.

The diagnosis came from CloudTrail, not from the error text. Filtering on `eventSource = lakeformation.amazonaws.com` and `eventName = GetDataAccess` showed the vended credentials tied to the registration role. Once I granted the Glue role data-location permission on the same S3 path and re-ran, the job succeeded in 6m12s against a 4-minute baseline. The extra minutes were the job retrying the first few partitions.

Fix, if you hit this:

```bash
aws lakeformation grant-permissions 
  --principal '{"DataLakePrincipalIdentifier":"arn:aws:iam::123456789012:role/GlueETLRole"}' 
  --resource '{"DataLocation":{"ResourceArn":"arn:aws:s3:::analytics-curated/orders/"}}' 
  --permissions DATA_LOCATION_ACCESS


Verify by re-running the job and confirming the CloudWatch log ends in `Job run Succeeded` — and by reapplying the same grant in Terraform so the next registration doesn't silently break it again.

## Exam technique: 130 minutes, 65 questions

That's two minutes per question, and the long scenario items will eat four if you let them. Read the last sentence of the stem first — it usually says "MOST cost-effective" or "LEAST operational overhead", and those two phrases have different right answers. On a Glue-vs-EMR question, the cheap answer is often EMR Serverless; the low-ops answer is often Glue.

Flag anything you can't decide in 90 seconds and move. My first pass typically leaves 8–12 flagged; the second pass closes most of them because a later question often clarifies an earlier one. AWS does not penalise wrong answers, so never leave a multiple-response item with fewer selections than it asks for — partial credit doesn't exist, but guessing costs nothing.

## When this study plan is the wrong tool

If your deadline is under three weeks and you haven't run a Spark job, this schedule will produce a stressed candidate and a $150 retake. Either push the date six weeks out or sit a fundamentals exam first. Equally, if you only need credibility for a sales conversation rather than for your own work, the free official practice question set on Skill Builder plus the public exam guide will tell you what you need without the labs.

We keep a Terraform baseline for exactly this lab setup (bucket, crawler, Lake Formation grants) so it can be torn down and rebuilt in one apply; that's part of what awsgood.com's managed services team maintains for clients who'd rather not hand-roll the governance layer. Tear down your lab tonight, then book the exam for a date 5 weeks out and put it on the calendar as a real appointment.

## Sources

- [AWS Certified Data Engineer – Associate exam guide (DEA-C01)](https://docs.aws.amazon.com/aws-certification/latest/data-engineer-associate-01/data-engineer-associate-01.html)
- [AWS Certification: Certified Data Engineer – Associate](https://aws.amazon.com/certification/certified-data-engineer-associate/)
- [What is AWS Glue?](https://docs.aws.amazon.com/glue/latest/dg/what-is-glue.html)
- [What is AWS Lake Formation?](https://docs.aws.amazon.com/lake-formation/latest/dg/what-is-lake-formation.html)

FAQ

How long does it take to prepare for DEA-C01?

If you already build Glue or Redshift pipelines at work, 40–60 study hours over about 6 weeks is realistic. Without hands-on data engineering experience, budget 80–90 hours, because the exam assumes you know crawler, bookmark and partition behaviour from having debugged it rather than having read about it.

What is the passing score for DEA-C01?

AWS uses a scaled score from 100 to 1000, and the minimum passing score is 720. The exam has 65 questions in 130 minutes, and some questions are unscored during development, so you cannot calculate a pass from a raw question count. Roughly 75% correct is the working target.

Is DEA-C01 harder than the Solutions Architect Associate?

They are differently hard. SAA-C03 rewards breadth across networking, identity and cost; DEA-C01 rewards depth in a narrow set of services and punishes surface knowledge of Glue, Kinesis and Lake Formation. Engineers with daily ETL work usually find DEA-C01 faster to prepare for.

How much does the DEA-C01 exam cost and how long is it valid?

The fee is $150 USD per attempt, though local currency and tax vary by region. The exam is delivered through Pearson VUE at a test centre or online-proctored, lasts 130 minutes, and the certification is valid for 3 years before you need to recertify.

Do I need to know Lake Formation for DEA-C01?

Yes, and not only in the security domain. Roughly 18% of the exam is Security and Governance, but Lake Formation permissions also appear in ingestion and storage questions, because a Glue job or Athena query that loses catalog access fails with a permission error rather than a data error.

Can I pass DEA-C01 with only video courses?

Unlikely. The exam asks for the smallest correct fix to a described failure, and that judgement comes from running the services. Build one crawler, one partitioned Parquet job and one Redshift table with two sort keys. The lab costs under $5 in us-east-1 and covers the highest-weight topics.

Leave a Reply

Your email address will not be published. Required fields are marked *