This document provides an overview ofRedis deploymentsand migrations to Google Cloud, including the options and tradeoffs fordeploying Redis on different services, based on your requirements.

Redis is an in-memory data structure store that you can use as a database,cache, message broker, and more. Google Cloud fully supports Redis,including the following:

• Fully managed options provided by Memorystore and Redis Ltd.

• Self managed options using the following products:

  • Compute Engine

  • Google Kubernetes Engine

  • OpenShift

The best way to deploy Redis on Google Cloud depends on your specificneeds and requirements. The recommendations provided in this guide are based ongeneral best practices and considerations. It's important to thoroughly analyzeyour Redis workload and consult official documentation or seek professionaladvice for specific use cases or requirements.

Architectures

You can deploy Redis using one of the following architectures:

Architecture	Description	Use case	Deployment options	High availability	Read throughput	Write throughput
Standard (standalone)	A single Redis node, with no read replicas and no high availability.	Cases where all data fits on one node, write and read throughput can be served by one node, and high availability is not required.	Supported on Memorystore (fully managed) and Redis open source software (OSS) (self-managed). Self-managing requires a more complex setup. Memorystore is a good option to get started quickly.	No	Single node	Single node
HA and/or read replicas	A single Redis node for write operations, with additional nodes to provide high availability and optionally share the read load, for example, using Sentinel.	Cases where write throughput can still be served by one node, but either read throughput cannot be served by one node, or high availability is required.	Supported on Memorystore (fully managed) and Redis OSS (self managed). Redis Cluster architectures offer automated scaling, high availability, and data sharding, which are ideal for large-scale, distributed applications. To understand the tradeoffs and required maintenance efforts in manual scaling, clustering and sharding, review Zero-downtime scaling in Memorystore for Redis Cluster. Self-managing requires a more complex setup. Memorystore is a good option to get started quickly.	Multi-AZ	Multi node	Single node
Cluster (without proxies)	Several nodes split the data write operations with separate data shards. High availability and read replicas can be optionally added.	Cases where write throughput cannot be served by one node, and high availability or read replication is optionally required.		Multi-AZ	Multi node	Multi node
Cluster (with proxies)	Several nodes split the data write operations with separate data shards. High availability and read replicas can be optionally added. Proxies are deployed on each primary node.	Cases where write throughput cannot be served by one node, and high availability or read replication is optionally required, and where it is too costly or inconvenient for client applications to be refactored to use the Redis Cluster API, or the use of proxies has other benefits.	Supported by Redis Enterprise Cloud (fully managed) or Redis Enterprise Software (self-managed). Self-managing with Redis OSS requires a more complex setup. Redis Enterprise Cloud is a good option to get started quickly.	Multi-AZ or multi-region (Redis Enterprise only)	Multi node	Single node

Deployment options

Google Cloud offers the following Redis deployment options:

• Fully-managed Memorystore for Redis by Google Cloud: A fullymanaged, highly available, and durable Redis service managed by Google that iscost effective and fast to set up, operate, and scale. Memorystoresupports bothRedis Clusterandstandalone Rediswith optional high availability.
• Self or fully-managed Redis Enterprise by Redis Ltd.: A highly availableand durable Redis cluster licensed by Redis Ltd. and with two managementoptions: managed by Redis Ltd. ("Redis Enterprise Cloud"), or self-managed("Redis Enterprise Software") with Redis Ltd.'s support. You can procure RedisEnterprise directly fromRedis Ltd., or throughGoogle Cloud Marketplace.Redis Ltd. supports deployments on Compute Engine , Google Kubernetes Engine andOpenShift.
• Self-managed Redis Open source software (OSS): A self-managed Rediscluster or standalone Redis with optional high availability, deployable onCompute Engine, Google Kubernetes Engine, or OpenShift.

Note: On April 20, 2024,Redis Ltd. announced its adoption of Dual Source-Available Licensing.Starting with Redis 7.4, Redis will no longer be distributed under thethree-clause Berkeley Software Distribution (BSD).The Linux Foundation launched the Open Source Valkey Community- a Redis Open Source fork.

Choose a Redis deployment option

This section describes how to choose a Redis deployment option that is bestsuited for your workload. Figure 1 provides a visual overview of the decisionpoints:

Choose a Redis management model

You can choose between one of the following management models:

• Fully managed deployment. You offload deployment and management operationsto the service provider. Choose this model when you need to focus on buildingyour app and offload management tasks.

• Self managed deployment. You are responsible for deployment and managementoperations. Choose this model if any of the following are true:

  • You have an existing operational economy of scale, and taking on managing andoperating Redis makes economic sense in your organization.

  • You have a strategic preference for IaaS-only dependency.

  • You require advanced optimizations.

Evaluate deployment options

After you choose your management model, evaluate the deployment options thatare available to you.

Fully Managed Options

For fully managed deployments, you can use Memorystore or Redis EnterpriseCloud.

Memorystore

Choose Memorystore if any of the following are true:

• You have a preference for consolidating support of managed software withGoogle Cloud.
• You have a preference for optimizing for integration with Google Cloudconstructs such as Identity and Access Management, APIs, org policies, quota, or Cloud Asset Inventory.
• You require specific features only available in Memorystore(e.g, scale back in).

For more information about Memorystore, see theMemorystore product documentation.

Redis Enterprise Cloud

Choose Redis Enterprise Cloud if any of the following are true:

• You require specific features only available in Redis Enterprise Cloud(for example, cross-region active-active multi-primary writes with its99.999% SLA, RedisSearch use case).
• You require cluster scaling for an application that does not support theRedis Cluster API.

For more information about Redis Enterprise Cloud, see theRedis Cloud documentation.

Self managed options

For self-managed deployments, you can choose between Redis Enterprise and RedisOpen Source Software.

Redis Enterprise

Choose self-managed Redis Enterprise if any of the following are true:

• Your application requires its unique features, such as automatic re-shardingfor scaling out, Redis on flash, or Redis Enterprise Operator for Kubernetes.
• Your operations team does not have the required skill set to handle complexRedis issues internally without qualified third party support.
• You prefer the enterprise support provided by Redis Ltd., and the associatedlicensing costs are manageable by your organization.

For more information about Redis Enterprise Software, see theRedis Enterprise Software documentation.

Redis Open Source Software

Choose self-managed Redis Open Source Software if any of the following are true:

• You require or have a preference for full customization not otherwise possible.
• Your operations team has the required skill set to handle complex Redis issuesinternally without qualified third party support.
• You want to avoid licensing costs.
• You have extensive in house Redis and Linux kernel tuning resources, or youruse case does not require tuning.

When deploying self-managed Redis Open Source Software, choose a deploymenttarget based on your platform strategy choice. Redis Open Source Software isdeployable on Compute Engine, Google Kubernetes Engine, or OpenShift. GKEAutopilot can reduce deployment and management efforts, but may be morelimited in ways such as being harder to scale in.

For more information about Redis Open Source Software, seeRedis.io.

Feature comparison

The following table summarizes the key differences between all deploymentoptions:

Deployment characteristics	Deployment options
	Memorystore for Redis andRedis Cluster	Redis Enterprise Cloud	Redis Enterprise software	Redis open source software
Managed by	Fully managed by Google	Fully managed by Redis Ltd.	Self managed	Self managed
Supported by	Google	Redis Ltd.	Redis Ltd.	Self supported
Billed by	Google	Redis Ltd. or Google	Infrastructure is billed by Google. Redis Ltd. license and support is billed by Redis Ltd. or Google.	Google
Cost elements	All costs included. Includes: infrastructure, licensing, support, and management costs. For more information, seeMemorystore pricing.	All costs included. Includes: infrastructure, licensing, support, and management costs. For more information, seeRedis Enterprise Cloud Pricing.	Software license and support costs are included. Infrastructure usage is billed separately by Google Cloud. Management costs, including deployment, tuning, personnel and downtime are absorbed by the customer. For more information, seeRedis Enterprise Software Pricing.	No service or licensing fees. Infrastructure usage billed by Google Cloud. Management costs, including deployment, tuning, personnel and downtime are absorbed by the customer.
SLA	Redis standard: 99.9% uptime; Less than or equal to 43.2 min/mo downtime Redis Cluster: 99.99% uptime with multi-AZ HA; Less than or equal to 4.38 min/mo downtime. For more information, see theMemorystore Service Level Agreement.	Redis standard: 99.9% uptime; Less than or equal to 43.2 min/mo downtime Redis Cluster: 99.99% uptime with multi-AZ HA; Less than or equal to 4.38 min/mo downtime. Multi-region active-active HA: 99.999% uptime; Less than or equal to 26.3 min/mo downtime. For more information, see theRedis Cloud Service Level Agreement.	Not applicable. You are responsible for uptime.	Not applicable. You are responsible for uptime.
Free tier	No	Yes	30 day free trial	Not applicable
Data tiering	No	Auto tiering	Auto tiering	No
Multi cloud	No	Yes	Manually	Possible, but involves high effort
Multiregion active-active	No	Yes	Manually	Possible, but involves high effort
Modules	Vector Search OtherRedis modules support varies, depending on licenses	RediSearch REST JSON RedisTimeSeries RedisBloom VectorDatabase OtherRedis modules	RediSearch REST JSON RedisTimeSeries RedisBloom VectorDatabase OtherRedis modules	VectorDatabase OtherRedis modules support varies, depending on licenses
Compliance	Built-in support for the different compliance regimes. SeeCompliance offerings for more information.	Built-in support for the different compliance regimes. SeeRedis Trust Center for more information.	Built-in support for the different compliance regimes. SeeRedis Trust Center for more information.	Manual compliance management is required. SeeCompliance offerings for more information.
Scaling cluster writes	Scales in and out	Scales in and out	Scales out. Scaling in requires manual effort.	Self managed, requires manual effort.
Auto rebalancing	Yes	Yes	Self managed, requires manual effort	Self managed, requires manual effort
Adding high availability	Seamless, no redeployment required	Seamless, no redeployment required	No redeployment is required, but requires manual effort	Requires substantial manual effort - redeployment may be required depending on your original architecture
Adding read replicas	Seamless, no redeployment required	Seamless, no redeployment required	Requires substantial manual effort - redeployment may be required depending on your original architecture	Self managed, requires manual effort
Moving to a data-sharded Redis Cluster when outgrowing write throughput	Requires redeployment, but tooling is provided to ease effort. Clients need to be refactored to support the Redis Cluster API.	Seamless, no redeployment required	Seamless, no redeployment required	Self managed, requires manual effort