Exam Tips - AWS Data Analytics Specialty Overview This guide covers common patterns, decision trees, and quick-reference tips for the DAS-C01 exam. Use this as a final review before the exam.
Service Selection Decision Trees Streaming Ingestion Need to stream data into AWS?
βββ Need custom consumers or ordering? -> Kinesis Data Streams
βββ Just load to S3/Redshift/OpenSearch? -> Kinesis Data Firehose
βββ Real-time SQL on streams? -> Kinesis Data Analytics (SQL)
βββ Complex stream processing (stateful)? -> Kinesis Data Analytics (Flink)
βββ IoT device data? -> IoT Core -> Rules Engine -> Kinesis/S3/DynamoDB
Batch Processing Need to process batch data?
βββ Simple transform on single file? -> Lambda (< 15 min, < 10 GB)
βββ Serverless ETL pipeline? -> Glue ETL
βββ Need Spark/Hive at PB scale? -> EMR
βββ One-time schema conversion? -> Glue Crawler + Glue ETL
βββ Orchestrate multi-step pipeline? -> Step Functions
Query and Analysis Need to query data?
βββ Ad-hoc SQL on S3 data? -> Athena
βββ Complex analytics with joins? -> Redshift
βββ Extend Redshift to S3? -> Redshift Spectrum
βββ Full-text search or log analytics? -> OpenSearch
βββ Interactive queries across sources? -> Athena Federated Query
βββ Business dashboards? -> QuickSight
Database Selection for Analytics What type of analytics data store?
βββ Structured, complex queries? -> Redshift
βββ Semi-structured, full-text search? -> OpenSearch
βββ Key-value with real-time access? -> DynamoDB
βββ Time-series data? -> Timestream
βββ Graph relationships? -> Neptune
βββ Data lake (all formats)? -> S3 + Lake Formation
High-Frequency Exam Topics Kinesis Data Streams vs Firehose Question Keyword Answer Custom processing logic Data Streams Load directly to S3/Redshift Firehose Real-time (milliseconds) Data Streams Near real-time (60 sec minimum) Firehose Manage scaling manually Data Streams (provisioned) Fully managed, no scaling Firehose Multiple consumers needed Data Streams Transform before delivery Firehose (Lambda)
Question Keyword Answer Athena query performance Parquet Hive on EMR ORC Schema evolution needed Avro Human-readable CSV/JSON Write-heavy streaming Avro or JSON Read-heavy analytics Parquet or ORC
Glue vs EMR Question Keyword Answer Serverless ETL Glue Custom Spark libraries EMR PB-scale processing EMR Schema discovery Glue Crawler Incremental ETL Glue (bookmarks) Interactive notebooks EMR (or Athena Spark) Hive, Presto, HBase EMR Cost-sensitive small jobs Glue
Key Numbers to Remember Metric Value Kinesis shard write capacity 1 MB/sec or 1,000 records/sec Kinesis shard read capacity 2 MB/sec (shared), 2 MB/sec per consumer (enhanced) Kinesis max record size 1 MB Kinesis retention 24 hours - 365 days Firehose buffer interval 60 - 900 seconds Firehose buffer size 1 - 128 MB Lambda timeout 15 minutes Lambda memory Up to 10,240 MB Athena query timeout 30 minutes Athena cost $5 per TB scanned Redshift Spectrum cost $5 per TB scanned S3 Standard-IA min duration 30 days Glacier Flexible retrieval 1 - 12 hours Glacier Deep Archive retrieval 12 - 48 hours DynamoDB export Does not consume read capacity QuickSight SPICE 10 GB/user Standard, 500 GB/user Enterprise
Common Traps and Distractors "Real-time" does not always mean Kinesis Data Streams - If loading to S3, Firehose is sufficient even though it has a 60-sec buffer Athena LIMIT does not reduce cost - Full partition scan still occurs; use partitions and columnar formats to reduce cost Redshift COPY is always preferred over INSERT - COPY is parallel, INSERT is single-row Glue Crawler does not transform data - It only discovers schema and updates the catalog EMR core nodes should not use Spot - HDFS data is on core nodes; losing them loses data Kinesis Analytics SQL is being replaced by Flink - But SQL questions still appear on the exam Lake Formation does not replace IAM - It works alongside IAM; users still need basic IAM permissions S3 Select is not a replacement for Athena - S3 Select works on single objects; Athena queries across objects Redshift Enhanced VPC Routing - Does not improve performance; it routes traffic through VPC for security DynamoDB Streams vs Kinesis adapter - Streams has 24h retention; use Kinesis adapter for longer retention needs Last-Minute Review Checklist Kinesis family: Streams vs Firehose vs Analytics - when to use each S3 storage classes and lifecycle transitions Glue components: Crawler, Data Catalog, ETL, bookmarks EMR node types and Spot instance strategy Redshift: distribution styles, sort keys, COPY command, WLM Athena: cost optimization (Parquet, partitions, compression) Lake Formation: permission model, LF-Tags, cross-account Encryption: SSE-S3 vs SSE-KMS vs SSE-C - when to use each VPC endpoints: Gateway (S3, DynamoDB) vs Interface (everything else) QuickSight: SPICE, RLS, CLS, embedding OpenSearch: UltraWarm, log analytics pipeline Step Functions vs Glue Workflows for orchestration