Prepare and test your skills
Prepare and test your skills
Differentiate Between Object, File, and Block Storage means matching how an application accesses data to the right Google Cloud storage service. Cloud Storage (opens in a new tab) provides object storage in buckets for unstructured data, Filestore provides shared file storage that many virtual machines can mount at the same time, and Persistent Disk and Google Cloud Hyperdisk provide block volumes attached to a single Compute Engine virtual machine. The choice depends on access protocol, consistency and replication needs, and whether the workload needs objects, shared files, or high-performance disks.
Cloud Storage is object storage that holds data as objects inside regional, dual-region, or multi-region buckets. Data in regional buckets is replicated synchronously across the zones in the region, so every zone copy stays consistent for reads. Data in dual-region or multi-region buckets is stored redundantly in at least two separate geographic locations, where metadata is written synchronously across regions and object data is replicated asynchronously. For dual-region buckets, turbo replication ensures newly written objects are replicated between the region pair with a recovery point objective (RPO) of 15 minutes. Cloud Storage is low-cost storage for data replicated across locations, including database backups and other unstructured data that must survive zone or region problems.
Filestore is fully managed network file storage that multiple virtual machines in a region can mount and share at the same time. A Filestore regional instance replicates data synchronously across three zones within the region, which provides high availability and robustness against zone outages. Applications use the shared instance for files needed by a whole tier, such as shared configuration files, common tools and utilities, and centralized logs across web-tier or application-tier virtual machines. For robustness against a region outage, a Filestore instance can be replicated to a different region. When a workload needs continuous availability Server Message Block (SMB) file storage, such as for a Microsoft SQL Server failover cluster instance, Google Cloud NetApp Volumes provides fully managed Network File System (NFS) and SMB storage.
Persistent Disk and Google Cloud Hyperdisk are high-performance block volumes attached to Compute Engine virtual machines. Zonal Persistent Disk volumes live in one zone, while regional Persistent Disk volumes provide synchronous replication of data across two zones within a region for more durability. Data in Persistent Disk volumes is not replicated across regions, so a regional disk protects against zone failure but does not by itself provide cross-region copies. Hyperdisk provides better performance, flexibility, and efficiency than Persistent Disk, and Hyperdisk Balanced lets you provision input/output operations per second (IOPS) and throughput separately and dynamically to tune one volume to many workloads.
Exam tip: Regional Persistent Disk replicates synchronously across two zones but not across regions; use Cloud Storage dual-region or multi-region buckets when data must be redundantly stored in two geographic locations.
Choosing between the three types starts from how the workload accesses data and how far replication must reach. The access path determines the service and the replication scope determines the configuration, so sharing model and resilience scope must be decided together. Balance functional characteristics, resilience requirements, performance expectations, and cost goals when you design storage.
A decision tree branching on objects, shared files, or block access to choose Cloud Storage, Filestore, or Persistent Disk and Hyperdisk with replication notes.
| Service | Access model | Replication and consistency |
|---|---|---|
| Cloud Storage | Objects in buckets for unstructured data | Regional buckets replicate synchronously across zones; dual-region and multi-region buckets write metadata synchronously and replicate data asynchronously |
| Filestore | Shared file system mounted on multiple virtual machines | Regional instance replicates synchronously across three zones; optional replication to a different region |
| Persistent Disk and Hyperdisk | Block volume attached to a virtual machine | Regional volumes replicate synchronously across two zones; data is not replicated across regions |
Choose Cloud Storage when low-cost, multi-location durability for objects matters most, Filestore when many machines must read and write the same files, and Persistent Disk or Hyperdisk when a virtual machine needs durable, tunable block performance.
Select Relational and Non-Relational Database Services means choosing an operational database that balances transactional guarantees, scale, schema control, and read and write performance. Cloud SQL provides managed relational databases for standard online transaction processing (OLTP), Cloud Spanner provides a mission-critical, global-scale relational database with SQL and transactions, Firestore in Datastore mode provides a schemaless NoSQL document database with strong consistency and automatic scaling, and Cloud Bigtable provides a NoSQL option when workloads do not require ACID transactions or highly structured data.
Cloud SQL is a fully managed relational database service that delivers MySQL, PostgreSQL, and SQL Server engines in the cloud. Cloud SQL stores related data in row-based tables so related elements from different tables connect without storing the same data multiple times. Cloud SQL is valued for openness, ease of use, security, cost-efficiency, and integration with Google Cloud.
Cloud SQL separates routine operation from application logic because Google handles setup, maintenance, and administration while applications connect with standard SQL. For higher demands, the Enterprise Plus edition for PostgreSQL and MySQL provides a 99.99% availability SLA with near-zero downtime maintenance, optimized hardware and software configurations, intelligent data caching for read-intensive transactional workloads, a configurable data cache option, and 35 days of log retention. The MySQL Enterprise Plus edition also adds advanced disaster recovery capabilities like orchestrated failover and switchback. Choose Cloud SQL when the workload needs a relational database with full SQL support for OLTP.
Cloud SQL protects durability through backups and replicas (opens in a new tab) that flow from the primary instance to separate recovery targets. Automated backups plus binary logging enable point-in-time recovery, which restores the database from a backup and recovers it to a fresh Cloud SQL instance. A high availability (HA) configuration protects against zonal failure and cross-region replicas protect against regional failure by maintaining copies that can serve traffic after failover. For Enterprise Plus, advanced disaster recovery simplifies recovery and fallback with zero data loss after a cross-regional failover.
Cloud Spanner is a mission-critical, global-scale database service that combines relational structure with global distribution. Spanner speaks GoogleSQL, which is an ANSI-compliant SQL dialect formerly called Google Standard SQL. Applications send query statements that scan tables and return rows, definition statements that create tables, views, and roles, manipulation statements that insert, update, and delete rows, access-control statements that grant and revoke privileges at the table and column level, and transaction-control statements that manage data modifications.
Spanner depends on Cloud Storage and Dataflow for full backup and restore workflows. A Dataflow export template copies the full Spanner database to a set of Avro files in a Cloud Storage bucket, and a separate import template loads those exported files into a new Spanner database. For more controlled backups, the Dataflow connector lets custom pipeline code read from Spanner and write to Cloud Storage as the backup target. Choose Spanner when the workload needs relational transactions and SQL beyond a single zone or region.
Firestore in Datastore mode, often called Datastore, is a NoSQL document database built for automatic scaling, high performance, and ease of application development. Datastore is schemaless, so entities of the same kind can have different properties and the same property name can hold different value types in different entities. Unlike a relational table that enforces a schema, Datastore leaves consistent property sets to application code when the application wants them.
Datastore maps naturally to object-oriented languages and exposes data through multiple clients plus a SQL-like query language. Its operational characteristics include strong consistency, automatic scaling, and managed security:
Datastore reads scale with the size of the result set, not the size of the dataset, so a query returning 100 entities performs similarly whether it searches hundreds or millions of entities. Because every query is served by indexes, Datastore does not support join operations, inequality filtering on multiple properties, or filtering based on subquery results. Datastore uses different names for familiar ideas: a kind corresponds to a table or collection group, an entity corresponds to a row or document, a property corresponds to a column or field, and a key corresponds to a primary key or document ID. Datastore fits product catalogs with real-time inventory, user profiles based on past activity, and ACID-based transactions such as transferring funds between bank accounts. Firestore protects against accidental deletion through a managed export and import service where the export flow moves entities to a Cloud Storage bucket and the import flow restores them back into Firestore.
Cloud Bigtable is a NoSQL database for workloads that do not require ACID transactions or highly structured data. Bigtable suits large operational datasets where flexible structure and high-throughput reads and writes matter more than relational joins and multi-row guarantees. Choose Bigtable when the data is not highly structured or when strict transactional consistency is not required.
Bigtable depends on Cloud Storage for portable backups and on asynchronous replication for redundancy. An export flow writes a Bigtable table to a series of Hadoop sequence files, those files are stored in Cloud Storage, and an import flow later loads the files back into another Bigtable instance. A replication relationship copies the Bigtable dataset asynchronously across zones within a Google Cloud region. This design separates live serving from durable backup copies without tying backup lifetime to the source instance.
Comparing operational database choices means weighing data model, consistency, and scaling behavior against workload needs. The four services split between relational guarantees and flexible NoSQL scale, so the need for SQL, transactions, and global distribution determines the fit.
| Service | Data model | Consistency and transactions | Scaling and fit |
|---|---|---|---|
| Cloud SQL | Relational tables with MySQL, PostgreSQL, or SQL Server | Full SQL for OLTP with point-in-time recovery, HA, and cross-region replicas | Scale within managed instances with 99.99% SLA and caching in Enterprise Plus; best for standard relational OLTP |
| Cloud Spanner | Relational tables with GoogleSQL | Transaction-control statements and table and column level access control at global scale | Global-scale mission-critical serving; best when relational guarantees must span regions |
| Firestore in Datastore mode | Schemaless NoSQL entities with kinds, properties, and keys | Atomic transactions and strong consistency without joins or multi-property inequality filters | Automatic scaling where query cost follows result size; best for catalogs, profiles, and flexible documents |
| Cloud Bigtable | NoSQL for large, less-structured data | No ACID requirement; async cross-zone replication | High-throughput operational scale; best when structure is loose and strict transactions are unnecessary |
Exam tip: Choose Cloud SQL for full-SQL OLTP, Spanner for global-scale relational transactions, Firestore when flexible entities still need atomic transactions and strong consistency, and Bigtable when the workload does not require ACID or highly structured data.
Assess Analytical and Lifecycle Requirements means matching data volume, access frequency, and retention to the right storage for analytics and long-term cost. BigQuery (opens in a new tab) is the choice for large-scale analytics over massive datasets, while Cloud Storage classes, locations, and lifecycle controls provide cost-effective retention for data that is accessed less often or must meet durability and residency goals.
BigQuery is a fully managed analytic data warehouse that separates storage from compute so each can scale independently on demand. You do not provision storage resources or reserve units, you pay only for the storage you use, and compute and storage are billed separately. When a query runs, the query engine distributes work in parallel across multiple workers over a petabit network and executes completely in memory for very fast movement of data to worker nodes.
A BigQuery query follows an ordered flow from distribution to return:
BigQuery storage is designed for 99.999999999% annual durability by replicating data across multiple availability zones to protect against machine-level and zonal failures. Data is automatically encrypted before it is written to disk with Google-managed or customer-provided keys, and it is stored in an efficient columnar format optimized for analytic workloads. BigQuery supports high-throughput streaming ingestion and high-throughput reads, but it is designed for analytics and is not designed for OLTP workloads, which belong in Cloud SQL, Spanner, or Firestore.
How access patterns determine where data lives is the decision between storing data in BigQuery and querying it where it lives. Data stored in BigQuery is primarily table data that you are billed for, and different table types support different lifecycle needs:
Data flows into BigQuery through batch loads from local files or Cloud Storage in formats that include Avro, Parquet, ORC, CSV, JSON, Datastore, and Firestore, through streaming with the Storage Write API, or automatically through the BigQuery Data Transfer Service. For analysis, BigQuery supports ANSI-standard SQL with joins, nested and repeated fields, analytic and aggregation functions, multi-statement queries, and spatial functions for geospatial analytics, plus a Pandas-compatible Python API, views for sharing analysis, BigQuery ML for machine learning, and BI Engine with Looker, Looker Studio, Google Sheets, Tableau, and Power BI. When data does not need to move, federated queries and external tables let BigQuery query data where it lives, including Cloud Storage, Bigtable, Spanner, Cloud SQL, AWS S3, Azure Blob Storage, and Google Drive.
How partitioning, clustering, and pricing control analytic cost is the practice of scanning less data per query and paying for compute in the right model. Partitioning and clustering organize large tables so predicate filters skip unneeded data, which reduces bytes scanned and improves performance. BigQuery does not support hierarchical partitioning, EXCHANGE PARTITION, SPLIT PARTITION, or converting a nonpartitioned table to a partitioned one, and no indexes are needed for batch analytics.
BigQuery implements three partitioning approaches that divide tables by time or integer ranges:
Clustering organizes data within partitions or tables by columns that are commonly filtered and aggregated together. Use clustering when tables are filtered on string columns or on multiple columns accessed together, since partitioning alone does not cover those patterns. Unmodified data stored for more than 90 days automatically drops to half-price storage, and key-value labels on datasets, tables, and views support cost tracking and internal chargebacks.
| Pricing option | How you pay | Choose when |
|---|---|---|
| On-demand pricing | Pay for bytes processed by each query | Query volume is variable or unpredictable and you want to pay only for queries you run |
| Capacity-based pricing | Purchase slot reservations with a minimum of 100 slots; projects default to a maximum of 2,000 slots | Workloads have predictable, concurrent reporting and extract-load-transform queries and need predictable cost and capacity with autoscaling |
Exam tip: BigQuery supports partitioning only by dates and integers; for strings, use clustering instead.
How storage classes and location types support lifecycle strategy is the way Cloud Storage matches retention, access frequency, availability, and residency to bucket configuration. For the bucket used to stage or retain data, choose a storage class based on data-retention and access-frequency requirements so infrequently accessed or long-retained objects cost less than frequently read data. Cloud Storage stores or replicates data only within the region you specify, which helps meet data residency requirements.
Bucket location controls how Cloud Storage replicates data for durability and availability:
This separation of durable retention from active analysis keeps analytics performant while lowering lifecycle cost. Keep hot analytic tables in BigQuery with partitioning, clustering, clones, snapshots, time travel, and dataset cloning for recovery, and retain source files, exports in Avro or Parquet, and long-term backups in Cloud Storage where class and location align with how often the data will be read and where it must reside.
Cloud Storage provides object storage in buckets for unstructured data, Filestore provides shared file storage that multiple virtual machines can mount at the same time, and Persistent Disk and Google Cloud Hyperdisk provide block volumes attached to a single Compute Engine virtual machine. Choose Cloud Storage when low-cost multi-location durability for objects matters most, Filestore when many machines must read and write the same files, and Persistent Disk or Hyperdisk when a virtual machine needs durable tunable block performance.
Choose Cloud SQL when the workload needs a managed relational database with full SQL support for standard online transaction processing (OLTP) using MySQL, PostgreSQL, or SQL Server. Choose Cloud Spanner when the workload needs relational transactions and SQL beyond a single zone or region because Spanner combines relational structure with global distribution.
Choose Firestore in Datastore mode when the workload needs flexible schemaless entities with atomic transactions and strong consistency without joins. Choose Cloud Bigtable when the data is not highly structured or when strict transactional consistency is not required because Bigtable is built for high-throughput reads and writes on large operational datasets.
BigQuery is a fully managed analytic data warehouse that separates storage from compute so each can scale independently and provides 99.999999999 percent annual durability by replicating data across multiple availability zones. Partitioning divides tables by time or integer ranges and clustering organizes data within partitions by commonly filtered columns so predicate filters skip unneeded data, which reduces bytes scanned and improves performance. BigQuery is not designed for online transaction processing (OLTP) workloads, which belong in Cloud SQL, Spanner, or Firestore.
An enterprise is planning the migration of three distinct enterprise workloads to Google Cloud:
ext4 journaling file system.Which combination of Google Cloud storage services should you recommend?
Workload 1: Cloud Storage; Workload 2: Filestore; Workload 3: Persistent Disk
Workload 1: Cloud Storage; Workload 2: Cloud Storage; Workload 3: Filestore
Workload 1: Persistent Disk; Workload 2: Cloud Storage; Workload 3: Filestore
Workload 1: Filestore; Workload 2: Persistent Disk; Workload 3: Cloud Storage
Last updated