HCIA-Big Data study plan
HCIA-Big Data · Give each week a clear focus.
8 weeks · About 6 hours per week
Big Data overview and ecosystem & Hadoop and MapReduce fundamentals
Define big data's core characteristics and distinguish batch, streaming, and interactive processing models with one real example each
Map the Hadoop ecosystem components (HDFS, YARN, MapReduce, Hive, HBase, ZooKeeper, Kafka, Spark, Flume, Sqoop) to the problem each solves
Explain HDFS architecture: NameNode, DataNode, block size, replication factor, and the read/write path at a conceptual level
Describe how a job's input splits become map tasks and reduce tasks, and why the shuffle is expensive
Your checkpoint
A one-page ecosystem diagram you draw from memory plus a 300-word written walkthrough of a file write and a MapReduce word-count job
Hadoop and MapReduce fundamentals
Write and run a MapReduce program locally (or in pseudo-distributed mode) that counts words and inspect its output part files
Configure a Combiner and a custom Partitioner, then explain how each changes network traffic and reducer load
Trace a job through its lifecycle states and identify which YARN component (ResourceManager, NodeManager, ApplicationMaster, Container) does what
Diagnose three common failures: data skew on one key, too few reducers, and small-file input
Your checkpoint
Working word-count code plus a short report comparing job counters and run time with and without a Combiner
Big Data overview and ecosystem & Hadoop and MapReduce fundamentals
Compare HDFS with a local filesystem and with object storage on durability, access pattern, and small-file handling
Explain YARN scheduling policies (FIFO, Capacity, Fair) and pick one for a shared cluster scenario
Practice the core HDFS and YARN shell commands for upload, permission change, block inspection, queue listing, and killing an application
Summarize data locality levels (node, rack, off-switch) and how speculative execution helps stragglers
Your checkpoint
A command cheat sheet you have executed at least once per line, with captured output for five commands
Hive and data warehousing basics
Explain Hive's architecture: metastore, HiveServer2, execution engine, and how SQL becomes distributed jobs
Distinguish managed (internal) from external tables and predict what happens to data on DROP in each case
Design a partitioned and bucketed table for a given query pattern and justify the choice
Write queries using joins, GROUP BY, window functions, and lateral views, and explain why a small-table join may become a map-side join
Your checkpoint
A Hive script that creates one external partitioned table, loads sample data, and answers five analytical questions with correct results
Hive and data warehousing basics & HBase and NoSQL basics
Compare the Hive data model (database, table, partition, bucket) with the HBase data model (namespace, table, row, column family, cell, timestamp)
Explain row-key design in HBase and why monotonically increasing keys cause hotspotting
Describe HBase architecture: HMaster, RegionServer, Region, WAL, MemStore, and StoreFile/HFile
State when to choose HBase over Hive and over a relational database, using access-pattern reasoning
Your checkpoint
A written design note for one workload that compares a Hive and an HBase implementation, including a proposed row key and a column-family layout
HBase and NoSQL basics
Perform HBase shell operations: create a table with a pre-split key, put, get, scan with filters, and delete
Explain why HBase provides only single-row atomicity and how that constrains application design
Describe compaction (minor versus major) and the read path through MemStore, block cache, and HFiles
Compare HBase with other NoSQL families (key-value, document, column-oriented, graph) and place each on the CAP spectrum
Your checkpoint
A runnable HBase shell script with a salted or pre-split row key plus 200 words on how your key choice avoids hotspots
FusionInsight and big data operations
Describe FusionInsight Manager's role: cluster deployment, service configuration, monitoring, alarms, and multi-tenant resource management
Explain how FusionInsight components relate to their open-source equivalents and what HA means for NameNode, ResourceManager, and Hive metastore
Outline routine operations: adding a node, expanding storage, backing up the metadata database, and applying a patch
Interpret typical alarms and map each to a first diagnostic step
Your checkpoint
An operations runbook covering five routine tasks and five alarms, each with a numbered procedure and a rollback note
Big Data overview and ecosystem & Hadoop and MapReduce fundamentals & Hive and data warehousing basics & HBase and NoSQL basics & FusionInsight and big data operations
Answer timed practice questions across all five domains and log every miss by domain and root cause
Re-derive from memory the three comparisons you are weakest on: Hive versus HBase, MapReduce versus Spark, and managed versus external tables
Re-run your week 2 and week 6 scripts end to end without notes to confirm retention
Build a one-page formula and fact sheet covering block size, replication, partitioning, row-key rules, and HA roles
Your checkpoint
A scored 60-question mock attempt with a written error log and a finalized one-page fact sheet
Milestones
Week 2: You can explain, without notes, how an input split becomes a map task and why the shuffle dominates cost, and your word-count job runs successfully
Week 4: You can create a partitioned external table, load data, and query it correctly, and you can state what happens to the data on DROP for both table types
Week 6: You can write a pre-split or salted HBase row key and justify it against hotspotting, and your shell script runs cleanly
Week 8: Mock score is stable across two attempts and no single domain accounts for more than a third of your errors