Jump to Content

Regional Persistent Disks: Delivering maximum resilience for your mission-critical workloads

April 23, 2024
Michael Ng

Product Manager

David Seidman

Group Product Manager

Try Gemini 1.5 models

Google's most advanced multimodal models in Vertex AI

Try it

Are you looking for a high-availability solution for your mission-critical workloads that delivers zero-RPO, is production-ready, and is built natively for Google Compute Engine? Look no further than Regional Persistent Disk. In this blog we dive into how Regional PD, with its cross-zone synchronous replication capabilities can help you deliver resilience, management simplicity, and continuous protection for your most important and demanding applications.

Regional PD: a strong fit for your mission-critical workloads 

Mission-critical workloads are essential to the day-to-day operations of any modern business today. They are the systems that are required to keep the business running, and any disruption to these workloads can have a significant impact on the business. For this reason, businesses need highly available infrastructure in place, including highly available storage that can:

  1. Seamlessly replicate data with performance optimized for these workloads

  2. Be highly resilient to a range of failures affecting availability 

  3. Be very simple to set up and manage

Regional PD hits the mark for all of these aspects and is a good fit for many of our Google Cloud customer’s mission critical high availability storage needs.

Highly performant for your mission critical workloads

"Regional Persistent Disk makes it easy for us to ensure high availability for our production fraud-detection system by embedding synchronous replication into our primary storage that is reliable and performant. We especially like the performance we get that satisfies the needs of our application especially since it is coupled with replication that safeguards our data in case of disaster." - Danilo Bustos, Chief Analytics Architect at Equifax

Regional PD delivers I/O performance with write performance up to 1000MB/s and 80K IOPS per volume, and read performance up to 1200 MB/s and 100K IOPS, leading Google Cloud customers to choose it for workloads including MySQL, Postgres, SQL Server, Jupyterlab, Kafka, Druid, Redis, Solr, EventStoreDB, Elasticsearch, MongoDB and many others.

Highly available and failure-resilient

Regional PD provides continuous synchronous replication between zones, with RPO=0 if one zone experiences an outage. In the case of a zonal outage, Regional PD prioritizes availability and automatically redirects I/O to the remaining available storage replica. Later, when the unavailable zone recovers, Regional PD self-heals and brings the unavailable replica back online to continue RPO=0 replication. If a VM needs to be recovered in the second zone, it can be attached to the Regional PD to maintain workload availability. 

Full replication for your mission-critical workloads implies replicating all data that is needed to ensure your compute instance is highly available. This goes beyond just replication of data disks, but also boot disks of the VM instances. Using Regional PD as the boot disk for VMs protects users from any storage errors that could prevent the VM from booting, or from unavailability of data from VMs that leverage the boot disk for data like Windows VMs.

Simple setup and management

Regional PD is extremely easy to configure for mission-critical workloads. Simply add a new Regional PD disk when creating a new VM or to any existing VM and the disk is added in minutes. You can add Regional PDs using the Google Cloud console or gcloud, Terraform and REST APIs. Once configured, your workload can begin serving reads and writes to the Regional PD disk just like to any other block storage — except that all writes are automatically replicated to another replica on a secondary zone using synchronization replication.


Figure 1: Adding a Regional PD to a VM is simple. Simply create a Regional Disk and add it when creating a new VM or to any existing VM

Regional PD in highly available Google Cloud services

Besides using Regional PD in Compute Engine workload deployments, a range of Google Cloud services also use Regional PD, including GKE Stateful High Availability (HA) Controller, Cloud SQL HA configuration, and Cloud Shell.

"I must commend your team for creating a product that is user-friendly and has been really helpful while we manage our stateful Kubernetes applications. The fact that Regional PD helps replicate data automatically helps remove this responsibility from us, saving us time. The simplicity of adding and managing Regional PD for our workloads was very valuable as it was apparent that it was well integrated with the rest of the Google Cloud ecosystem. Regional PD is reliable, secure, and has allowed me to attach and relocate our user data with ease."  - Eugene Chupkin, Sr. Manager, Devops, The Kroger Co

Google Cloud SQL is Google Cloud’s fully managed database service for MySQL, PostgreSQL, and SQL Server. When enabled, the Cloud SQL High Availability configuration automatically provisions a Cloud SQL database instance with Regional PD as the primary storage — automatically replicating writes to a secondary zone as soon as the instance is created. Failover, when required, is a simple one-step command that reroutes all compute, networking and client applications to the alternate instance in the secondary zone. 

Cloud Shell is Google Cloud's terminal window that provides command line access to VMs in your projects. Every Cloud Shell terminal session provides users with 5GB of Regional PD disk storage for their terminal’s home directory. This helps ensure that the home directory is highly available if there are errors in the zone that’s housing the home directory storage. If a zonal error occurs, the Cloud Shell service automatically uses the additional Regional PD replica, so there’s no interruption of home directory data availability.

Concluding thoughts

In this blog we’ve seen how Regional PD is purpose-built to help provide maximum availability for your mission-critical workloads in Google Cloud, delivering performance, resilience, and management simplicity. Additionally we discussed how customers have deployed Regional PD in large-scale production environments and how it provides storage availability in several foundational Google Cloud services. Try it today for your deployments by checking out our Regional PD documentation pages.

Posted in