Task 1.4: Designing data migrations.

• AStorage Transfer Service agents with a bandwidth cap, running for six weeks

  Incorrect

  At 1 Gbps, 100 TB takes about 12 days, so 400 TB needs roughly seven weeks at full line rate, and a cap to protect production traffic makes it longer.

• BTransfer Appliance, copying the archive locally and shipping it to Google

  Correct

  Google recommends Transfer Appliance when the available bandwidth cannot move the data before the deadline and more bandwidth cannot be acquired in time.

• CParallel gcloud storage uploads from several hosts over the existing link

  Incorrect

  More parallel uploads do not add bandwidth; the shared 1 Gbps link is still the limit.

• DAn event-driven Storage Transfer Service transfer from the source share

  Incorrect

  Event-driven transfers react to change notifications from cloud sources such as Amazon S3 or Cloud Storage; they do not speed up a bulk copy over a slow link.

The choice between online and offline transfer is a bandwidth calculation. Google's guidance puts 100 TB at about 12 days over 1 Gbps, so 400 TB over a shared 1 Gbps link cannot finish in six weeks. When bandwidth cannot be bought in time, Transfer Appliance moves the data offline.

• ADatabase Migration Service with a continuous migration job

  Correct

  A continuous migration copies the full database first and then replicates changes until the team promotes the destination, which keeps downtime to the cutover itself.

• BDatastream streaming the changes from MySQL into a BigQuery dataset

  Incorrect

  Datastream replicates into BigQuery, Cloud Storage or Apache Iceberg tables for analytics and events; reaching Cloud SQL needs extra Dataflow templates, and it has no promotion step.

• CA mysqldump export to Cloud Storage and then a Cloud SQL import

  Incorrect

  An export and import is a one-time, point-in-time copy; writes made after the export are missed unless the source stops taking writes for the whole copy.

• DStorage Transfer Service copying the MySQL data files to Cloud Storage

  Incorrect

  Copying raw database files to a bucket does not create a running Cloud SQL database or keep one in sync.

Database Migration Service migrates MySQL, PostgreSQL, SQL Server and Oracle databases to Cloud SQL or AlloyDB. A continuous migration starts with a full initial copy of the database and keeps replicating changes, so the cutover happens when source and destination are in sync and the destination is then promoted to a standalone primary.

• AThe interactive SQL translator in BigQuery Studio, one script at a time

  Incorrect

  The interactive translator is for reviewing and debugging a single query; 6,000 scripts one at a time is the case the batch tool exists for.

• BScheduled queries in BigQuery that run the scripts unchanged

  Incorrect

  A scheduled query must be valid GoogleSQL, so Teradata BTEQ and SQL have to be translated before BigQuery can run them, and nothing reports what needs fixing.

• CThe batch SQL translator, with the scripts uploaded to Cloud Storage

  Correct

  The batch translator converts many scripts at once from files in Cloud Storage and writes one output file per input plus a summary report of warnings and errors.

• DGemini in BigQuery, rewriting each script from a natural-language prompt

  Incorrect

  Gemini assistance helps write individual queries in the editor; it is not a bulk dialect translator with a migration report.

BigQuery's migration tooling has a batch SQL translator for bulk work and an interactive translator for single queries. The batch job reads the source files from Cloud Storage, writes the GoogleSQL output alongside, and produces a CSV summary of warnings and errors that the team uses to plan the fixes.

• ATransfer Appliance, shipped once for the initial copy of the filer

  Incorrect

  The link can move 50 TB in well under a day, and an appliance does not provide the nightly incremental copies.

• BA cron job on one host that runs gcloud storage rsync each night

  Incorrect

  A script on one host has no management interface, agent fault tolerance or transfer logs, and is limited to that host's throughput.

• CThe BigQuery Data Transfer Service, with a scheduled file transfer

  Incorrect

  The BigQuery Data Transfer Service loads data into BigQuery tables; it does not copy files from an NFS filer into Cloud Storage.

• DStorage Transfer Service with agents installed near the filer

  Correct

  Storage Transfer Service for file systems uses agents in the data center for full and incremental copies on a schedule, with bandwidth caps and a record of every file moved.

When bandwidth is sufficient, an online transfer is the right tool. For on-premises file systems, Storage Transfer Service runs fault-tolerant agents near the data, supports recurring incremental transfers and bandwidth limits, and logs every file it moves.

• AA BigQuery migration assessment on extracted metadata and query logs

  Correct

  The migration assessment analyzes the source warehouse's metadata and query logs and produces a report with BigQuery storage and workload cost estimates.

• BA trial Data Transfer Service transfer of the largest Redshift tables

  Incorrect

  A trial transfer moves data, which stakeholders wanted to avoid, and says little about how the whole warehouse is used.

• CThe batch SQL translator on the warehouse's stored procedures

  Incorrect

  Translation converts SQL; it does not inventory the warehouse or estimate BigQuery costs.

• DThe BigQuery pricing calculator with storage estimated by hand

  Incorrect

  A hand estimate leaves out how the workload actually uses the warehouse, which the assessment reads from the query logs.

The BigQuery migration assessment is the first step of a warehouse migration. Metadata and query logs extracted from the source, such as Amazon Redshift, Teradata or Snowflake, are analyzed to describe the existing warehouse and estimate BigQuery costs, with results in a Data Studio report.

• AA nightly Cloud Run job that exports the tables to CSV and loads them

  Incorrect

  A nightly export and load is batch code to maintain, and it cannot apply changes within minutes.

• BDatastream, with a stream from PostgreSQL to a BigQuery destination

  Correct

  Datastream is a serverless change data capture service that replicates CDC and backfill data into BigQuery for near real-time analytics.

• CA nightly BigQuery Data Transfer Service transfer from PostgreSQL

  Incorrect

  A nightly batch transfer cannot apply changes within minutes.

• DFederated queries from BigQuery to the on-premises database

  Incorrect

  Federated queries reach Cloud SQL, AlloyDB and Spanner, not an on-premises database, and they replicate nothing.

Datastream reads changes from MySQL, PostgreSQL, Oracle, SQL Server and other sources and applies them to BigQuery, together with a backfill of existing data, with no servers or pipelines to manage.

• ARerun the migration assessment against the BigQuery tables

  Incorrect

  The assessment analyzes the source warehouse before the move; it does not compare source and destination tables.

• BTranslate the reports again with the interactive SQL translator

  Incorrect

  Translation converts SQL; it says nothing about whether the data matches.

• CRun the data validation tool to compare source and destination

  Correct

  After migrating, Google's guidance is to run the data validation tool to confirm that the source and destination tables match.

• DCheck each table's row count on its BigQuery Details tab

  Incorrect

  Looking at BigQuery alone gives no comparison with the source, and row counts by hand do not scale or cover column values.

A migration is not finished when the data is loaded. Google's migration overview ends with validation: the data validation tool compares source and destination tables, so that consumers move to BigQuery with evidence that the data is complete.