Prepare and test your skills
Prepare and test your skills
Block storage performance and capacity management is the practice of matching disk type, size, and placement to workload needs for speed, durability, and cost. Persistent Disk (opens in a new tab), Hyperdisk, and Local SSD offer different combinations of replication, durability, and tunable performance. Placement as zonal or regional controls whether data stays available during a zone outage.
Persistent Disk is durable network storage that is independent from a virtual machine (VM). A disk can be detached from one VM and attached to another, and data remains after the VM is deleted. A zonal disk stores data in a single zone, while Regional Persistent Disk (opens in a new tab) replicates every write synchronously across two zones in the same region. Both types include built-in redundancy that protects against equipment failure and maintenance events.
Regional allocation provides high availability without a restore step. If the primary zone fails, a backup VM in the surviving zone can force-attach the Regional Persistent Disk and continue serving data. Zonal disks cannot fail over this way and must be recovered from a snapshot or backup in another zone or region. Choose regional allocation when continuous access and near-zero recovery point objective (RPO) are required, and choose zonal when restore-based recovery is acceptable.
A regional Persistent Disk in us-central1 keeps synchronous replicas in Zone A and Zone B. The primary VM in Zone A attaches to the local replica; if Zone A fails, the backup VM in surviving Zone B force-attaches the replica there with no restore step.
| Option | Replication scope | Behavior during zone outage |
|---|---|---|
| Zonal Persistent Disk | Data stored in one zone | Disk is unavailable with its zone; restore from snapshot or backup to recover |
| Regional Persistent Disk | Writes replicated synchronously between two zones in one region | Backup VM can force-attach the disk in the surviving zone |
| Hyperdisk Balanced High Availability | Data replicated synchronously between two zones in one region | Volume remains available for high-availability workloads during a single-zone outage |
Exam tip: Regional Persistent Disk protects against a zone outage by synchronous replication between two zones, while snapshots protect against data loss and region outage by storing a copy that can be copied across regions.
Google Cloud Hyperdisk (opens in a new tab) provides higher performance, flexibility, and efficiency than Persistent Disk for Compute Engine workloads. With Hyperdisk Balanced, input/output operations per second (IOPS) and throughput can be provisioned separately and changed dynamically. One volume can therefore be tuned for transaction-heavy, throughput-heavy, or mixed workloads. Hyperdisk Balanced High Availability adds synchronous replication between two zones in the same region for zone-level resilience.
Hyperdisk Storage Pools share pre-provisioned capacity, throughput, and IOPS across multiple Hyperdisk volumes instead of provisioning each disk alone. The storage pool logically groups physical storage and uses efficiency features such as thin provisioning and deduplication. The pool attempts to add capacity automatically to keep utilization at or below 80% of provisioned capacity, so space remains available as workloads grow. Use storage pools when many volumes have uneven demand and sharing headroom lowers cost while preserving performance.
Local SSDs are physically attached solid-state drives that provide high throughput for a single VM. Data on a Local SSD lasts only until the instance is stopped or deleted, so the drive is not durable storage. The VM depends on its host for Local SSD data, and stopping the VM breaks that dependency and loses the data.
Choose Local SSDs for temporary data, caches, and data that can be reconstructed by other means. Choose Persistent Disk or Hyperdisk for data that must survive VM deletion, host failure, or zone maintenance. This separation keeps high-speed scratch space fast while system and application data stays on replicated block storage.
Snapshots (opens in a new tab) capture the point-in-time state of Persistent Disk volumes for durability and recovery. Snapshots are incremental by default to use less space and can be created in minutes even when disks are attached to running instances. Snapshots are stored redundantly with automatic checksums to verify integrity, and the Cloud Storage location for snapshots is configurable.
Synchronous replication inside one region handles high availability, while cross-region copies handle disaster recovery:
Synchronous zonal replication provides zero to near-zero recovery time objective (RTO) and RPO inside one region. Cross-region snapshot copies and backups use asynchronous replication and require restore steps, which increase RTO and RPO but protect against region outages.
Managed database storage scaling and provisioning is the allocation of capacity, performance, and availability so changing data volumes do not cause slowdowns or outages. Allocation has two parts: deciding which storage services are needed, and choosing the required features and design options inside each service. For enterprise workloads, the goal is optimal performance, cost-efficiency, and high availability at the same time.
Block storage for databases in Google Cloud centers on three options that divide work by durability and performance control. Hyperdisk (opens in a new tab) is Google's recommended durable block storage and is required for the latest machine series, with performance and capacity that scale independently. Persistent Disk provides durable block storage for IOPS-intensive or latency-sensitive applications and databases, while Local SSD provides ephemeral flash storage for fast scratch, cache, and flash-optimized databases.
| Storage option | Suitable database workloads | Performance and capacity control |
|---|---|---|
| Hyperdisk | IOPS-intensive or latency-sensitive applications, databases, shared read-only storage, rapid durable VM backups, scale-out analytics | Provision IOPS and throughput separately and dynamically; tune with Hyperdisk Balanced for a wide variety of workloads |
| Persistent Disk | IOPS-intensive or latency-sensitive applications, databases, shared read-only storage, rapid durable VM backups | Choose disk type, size, and IOPS for Extreme Persistent Disk |
| Local SSD | Flash-optimized databases, hot-caching for analytics, scratch disk | Use as ephemeral fast scratch disk or cache, not durable primary storage |
Exam tip: Choose Hyperdisk when database capacity and performance must scale independently with durability, and choose Local SSD only when the data is temporary cache or scratch that can be lost.
Storage provisioning for managed database services means setting capacity, provisioned performance, and data-protection options that keep the system stable as demand changes. With Hyperdisk Balanced, throughput and IOPS are provisioned separately from capacity, so an administrator can raise IOPS for transaction spikes or raise throughput for scans without over-provisioning disk size. If provisioned capacity is exceeded, requests can fail, so business-critical workloads reserve throughput ahead of demand.
Provisioning decisions for durable block storage include peer settings that are configured together:
These settings depend on each other because the disk type determines which performance controls are available, and the region and replication choice determines robustness against zonal outages. Snapshot schedules and encryption apply to the provisioned volumes, so protection follows the same capacity that serves live traffic.
Availability for managed data in Google Cloud comes from replicating storage across zones and regions and pairing it with backup and recovery. Spanner (opens in a new tab) is designed for high data availability and global scalability, and its multi-region configurations replicate data in multiple zones across multiple regions to stay available even during a region outage. AlloyDB for PostgreSQL (opens in a new tab) clusters provide high availability by default with automatic failover, where the primary instance has redundant nodes located in two different zones within a region. Cloud Storage (opens in a new tab) buckets provide the same pattern for objects, with regional buckets replicating synchronously across multiple zones, and dual-region or multi-region buckets replicating asynchronously across regions.
Spanner pairs replication with disaster-recovery controls that protect growing datasets from deletion and corruption:
Continuity also requires tradeoffs between cost and robustness. An AlloyDB for PostgreSQL primary instance can be changed to a basic instance to reduce cost, but a basic instance is not robust against zone outages and has longer downtime during maintenance. For file-based database dependencies, files stored in a Filestore Regional instance are replicated synchronously across three zones in the region, and the instance can be replicated to a different region for robustness against region outages.
Cloud Storage (opens in a new tab) is a low-cost, no-limit object store for diverse data types that can be accessed from within and outside Google Cloud. Data allocation in Cloud Storage depends on two choices that work together: the bucket location type controls where data is replicated for availability and residency, and the storage class controls cost based on how often data is accessed. Object lifecycle management automates movement across tiers so frequently accessed data stays in low-latency storage and aging data shifts to lower-cost tiers.
Cloud Storage location types define the geographic scope and replication behavior for a bucket, with three options: regional, dual-region, and multi-region. Data in regional buckets is replicated synchronously across multiple zones within one region, so the bucket survives a zone outage but not a region outage. 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 objects replicate across the region pair with a recovery point objective (RPO) of 15 minutes.
The location type directly affects data residency, availability, and cost. Cloud Storage stores or replicates data within the region that is specified, which helps meet data residency and compliance requirements. A regional bucket provides low-cost backup storage that is redundant across zones, while a dual-region or multi-region bucket protects backups against region outage because the data can be used to restore in another region. Choose regional when users and compliance scope are contained in one region and zone-level resilience is enough, and choose dual-region or multi-region when the workload needs higher availability across regions.
| Location type | Replication scope | Replication behavior |
|---|---|---|
| Regional bucket | Multiple zones within one region | Data replicated synchronously across zones in the region |
| Dual-region bucket | Two specific geographic regions | Metadata written synchronously across regions, data replicated asynchronously; turbo replication option provides 15-minute RPO |
| Multi-region bucket | Two or more geographic locations | Metadata written synchronously across regions, data replicated asynchronously across regions |
Application and database backups show this tradeoff in practice. Backups stored in a regional Cloud Storage bucket are preserved if a zone outage occurs, while backups stored in a dual-region bucket are preserved if a zone or region outage occurs. Cross-region replication increases durability and disaster recovery options but adds storage and transfer cost compared to staying in one region.
Cloud Storage storage classes match storage cost to data-retention and access-frequency requirements. Standard storage is for high-frequency access, Nearline and Coldline storage are for low-frequency access, and Archive storage is for lowest-frequency access. The bucket owner selects the starting class based on how long data must be retained and how often it will be read, and lifecycle rules then transition objects as access cools.
Infrequent-access classes reduce storage cost compared to Standard storage but add retrieval costs and minimum storage durations:
Nearline, Coldline, and Archive are optimized for infrequent access, so the pricing model charges for minimum storage durations and adds costs for retrieving data or metadata earlier than the minimum duration for the class. Choose a colder class only when the data will truly remain untouched long enough to offset early-delete and retrieval charges.
Object lifecycle management in Cloud Storage is the automated mechanism that moves data across storage tiers for efficient resource utilization. The lifecycle policy applies to objects in a bucket and transitions them from Standard to Nearline, Coldline, or Archive as retention age and access frequency change, so the bucket does not pay Standard rates for data that is rarely read. This automation links the storage class decision to the data lifecycle: hot ingestion and active use stay in Standard, then aging backups and archival data flow to colder tiers without manual moves.
Cost optimization also depends on protecting retained data against loss. The Soft Delete feature preserves deleted and overwritten objects for a specified period, and Object holds prevent deletion or updates to objects. Together with dual-region or multi-region replication and turbo replication for faster cross-region copies, lifecycle management lets one bucket balance residency, availability, and cost across the full life of the data.
Exam tip: Regional buckets replicate data synchronously across zones, while dual-region and multi-region buckets write metadata synchronously but replicate data asynchronously — turbo replication on dual-region buckets tightens that async copy to a 15-minute RPO.
Zonal Persistent Disk stores data in a single zone and becomes unavailable with its zone, so recovery requires restoring from a snapshot or backup in another zone or region. Regional Persistent Disk replicates every write synchronously across two zones in the same region, so a backup virtual machine (VM) in the surviving zone can force-attach the disk and continue serving data without a restore step.
Local SSDs should be used for temporary data, caches, scratch space, and data that can be reconstructed by other means because they are physically attached drives whose data is lost when the virtual machine (VM) is stopped or deleted. Persistent Disk or Hyperdisk should be used for data that must survive virtual machine (VM) deletion, host failure, or zone maintenance because they provide durable replicated block storage.
Google Cloud Hyperdisk Balanced lets input/output operations per second (IOPS) and throughput be provisioned separately from capacity and changed dynamically, so one volume can be tuned for transaction-heavy, throughput-heavy, or mixed workloads. If provisioned capacity is exceeded requests can fail, so business-critical workloads reserve throughput ahead of demand.
Data in regional buckets is replicated synchronously across multiple zones within one region, so the bucket survives a zone outage but not a region outage. Data in dual-region buckets is stored in two specific geographic regions where metadata is written synchronously across regions and object data is replicated asynchronously, with turbo replication providing a 15-minute recovery point objective (RPO).
A retail organization runs its primary transactional database on Google Cloud using Cloud SQL for MySQL. During seasonal promotions, the application experiences sudden, unpredictable surges in transaction volume and data growth. In the past, unexpected data spikes caused on-premises database crashes due to disk space exhaustion.
The cloud architect needs to design a storage provisioning and capacity management strategy that satisfies the following requirements:
Which configuration strategy should the architect implement?
Last updated