Building a Redis Successor Cache Infrastructure with Valkey — License Migration and Performance Architecture in Custom Development for 2026

In Beyond Speed Limits: Exploring the Performance Power of Valkey (2026-06-08), InfoQ highlighted how Valkey, which emerged as the open-source successor to Redis, is delivering significant performance gains over conventional Redis through features such as multithreaded I/O. Valkey is a fork based on Redis 7.2.4 released under the BSD 3-Clause license, backed by major cloud providers including AWS, Google, and Oracle under the Linux Foundation. Because it is wire-protocol compatible with Redis, its key feature is that teams can migrate while keeping existing client code and data structures almost entirely intact.

In custom system development and infrastructure projects for clients, anxiety quickly spread: "With Redis licensing shifting to RSALv2 / SSPL, and further to AGPLv3, is it safe to keep using our current caching infrastructure?" and "Managed service policies are changing and forcing migrations, but we cannot predict compatibility." This is not simply a challenge of "can we use the latest caching engine?" but rather incorporating into the design "can we hand over a caching foundation where licensing risks are isolated, compatibility is verified, and performance is guaranteed?" Connecting with the memory and cost optimization policies discussed in Optimizing Cloud Costs by Reducing Backend Memory (GH Media) and the approach to numerically guaranteeing performance covered in Guaranteeing Performance with k6 Load Testing (GH Media), this article outlines "Caching Infrastructure Migration and Performance Design Support" as a custom development package.

Why consider Valkey from Redis "right now"

In 2024, Redis Inc. transitioned from the permissive BSD 3-Clause license it had long maintained to more restrictive source-available licenses (RSALv2 / SSPLv1). Furthermore, in 2025, AGPLv3 was added as an option for Redis 8.0 and later. Because AGPLv3 includes copyleft provisions, many organizations prohibit its adoption from legal and commercial risk perspectives, and usage terms changed as managed service providers required direct commercial agreements with Redis Inc.

Valkey was created in response to this trend. Forking from the last BSD-licensed release of Redis (7.2.4), it preserves the BSD 3-Clause license. Operating under the Linux Foundation without a single controlling corporation, participants include Amazon, Google, Oracle, Huawei, and Ericsson.

In short, "continuing to use Redis" and "operating a caching foundation with low licensing risk" have become two completely different things. In custom development as well, "isolating licensing risks, verifying compatibility, and delivering with guaranteed performance" has become a fundamental prerequisite of quality.

What changes with Valkey

Performance gains through multithreaded I/O

Valkey 8.0 introduced a new multithreaded I/O architecture that enables offloading tasks such as reading, parsing, and writing responses for commands to dedicated I/O threads. Official benchmarks report that throughput increased by approximately 230% compared to the 7.2 release line (scaling from around 360,000 to 1,190,000 RPS), with average latency improving substantially. Valkey 8.1 further offloaded TLS negotiation to I/O threads, driving additional gains in accepting new connections.

The core benefit lies in alleviating the single-threaded bottleneck of conventional Redis by delegating I/O processing to separate threads. However, as detailed later, the thread count must be tuned appropriately according to CPU cores and workloads; it is not simply a matter of "more threads always equals faster performance."

Managed support across the community and major cloud platforms

Major cloud providers have integrated Valkey into their managed service portfolios. AWS offers Amazon ElastiCache for Valkey / MemoryDB for Valkey, Google Cloud provides Memorystore for Valkey, and Oracle delivers OCI Cache with Valkey. AWS announced pricing for ElastiCache for Valkey that is approximately 33% lower for serverless and about 20% lower for node-based configurations compared to Redis OSS. The ability to simultaneously mitigate licensing risks and optimize costs is accelerating migration evaluations.

Compatibility and risks to verify during migration

Because Valkey is protocol-compatible with Redis 7.2, in most cases systems run simply by switching the client connection target. An example connection string is shown below.

# 既存の Redis クライアントはそのまま Valkey に接続できる
redis-cli -h my-valkey-endpoint.example.com -p 6379 PING
# => PONG

# valkey-cli も利用可能（Valkey 同梱）
valkey-cli -h my-valkey-endpoint.example.com -p 6379 INFO server | grep -i version

When enabling multithreaded I/O, specify the number of I/O threads in the configuration file. Verification aligned with core counts and workload profiles is required.

# valkey.conf — I/O スレッドの有効化（例）
# CPU コア数より小さい値から始め、負荷テストで最適値を探る
io-threads 4

# 永続化方針は Redis から引き継ぐ前に必ず再確認する
appendonly yes
appendfsync everysec

When checking compatibility, teams must thoroughly identify the following points: version discrepancies (which Redis release you are migrating from), module incompatibilities (certain modules such as RediSearch or RedisJSON may have separate implementations or alternative names in Valkey), differences in managed service specifications (parameter groups and failover behaviors across providers), and the fact that once migrated to Valkey, rolling back to Redis OSS may not be possible in some cases. Before talking about performance, the golden rule is always to verify in an actual environment whether existing assets run without modification.

Five phases of "caching infrastructure migration and performance design support" for custom development

Phase 1: Inventory and assessment (1 week)

• Inventorying current Redis versions, modules, and utilized features
• Confirming license classifications (OSS vs. managed contract) and associated risks
• Benchmarking client libraries, connection topologies, and data volumes
• Deliverables: Caching asset inventory sheet + migration feasibility and compatibility report

Phase 2: Design (1 week)

• Selecting migration targets (self-hosted Valkey / ElastiCache / Memorystore / OCI Cache)
• Architecting multithreaded I/O, persistence, and failover policies
• Defining rollback procedures and identifying irreversible milestones
• Deliverables: Migration design specification + performance architecture document

Phase 3: Migration and implementation (1 to 3 weeks)

• Building compatibility testing environments and eliminating discrepancies
• Tuning I/O thread counts and system parameters
• Implementing phased migration (parallel operation and cutover runbooks)
• Deliverables: Migration runbook + tuned configuration files

Phase 4: Verification and handover (1 week)

• Verifying throughput and latency via load testing
• Validating failover and behavior during failure scenarios
• Deliverables: Performance verification report + maintenance manual

Phase 5: Continuous maintenance (ongoing)

• Tracking upstream versions and applying security updates
• Continuous performance monitoring and periodic tuning reviews
• Architectural support for new caching use cases

Implementation standards set for custom development

Which projects need this and which do not

Six clauses to include in custom development contracts

Client ROI estimate (assuming managed Redis operations)

Even an initial assessment (starting from 200,000 yen) provides immediate value by visualizing how much licensing, compatibility, and performance risk is lurking in your current caching infrastructure. Overlooked license classifications almost always surface abruptly during contract renewals or audits.

Five common pitfalls to avoid

Pitfall 1: Migrating without checking version compatibility

Switching without understanding source Redis versions or utilized features leads to unexpected runtime behavior. Always inventory the current version and active features before migrating.

Pitfall 2: Overlooking module incompatibilities

Modules like RediSearch or RedisJSON may have different implementations or naming conventions in Valkey. Identify dependent modules and evaluate viable alternatives.

Pitfall 3: Ignoring managed service discrepancies

Parameter groups and failover behaviors differ across cloud providers. Verify the destination specifications individually.

Pitfall 4: Skipping failover validation

Testing only happy paths risks data loss or failed failovers during incidents. Always execute validation including fault injection. For production monitoring perspectives, Pinterest's CPU Zombies and SRE Performance Auditing (GH Media) provides valuable guidance.

Pitfall 5: Placing blind faith in multithreaded I/O performance

Recklessly increasing io-threads can backfire depending on core counts and workload profiles. Measure and identify optimal values through load testing.

90-day action plan

Conclusion — Moving from "continuing to use Redis" to "delivering with risks isolated"

Valkey reaching production maturity shifts caching platform strategy away from "being tossed around by licensing changes" toward "insulating against risks with a BSD fork while boosting performance under preserved compatibility." For custom system development and infrastructure teams, providing licensing isolation, verified compatibility, and guaranteed performance through "Caching Infrastructure Migration and Performance Design Support" is our flagship service for delivering dependable, long-lasting systems. If you are reviewing your data layer design down to tenant isolation, please also read Multi-Tenant Design with PostgreSQL RLS (GH Media).

If you are concerned about Redis license changes, want an assessment on whether migrating to Valkey is right for you, or need quantitative performance guarantees post-migration, please feel free to reach out via our contact form.

Sources

• Beyond Speed Limits: Exploring the Performance Power of Valkey（InfoQ 2026-06-08）
• Unlock 1 Million RPS: Experience Triple the Speed with Valkey (Valkey Official Blog)
• Valkey 8.1: Continuing to Deliver Enhanced Performance and Reliability (Valkey Official Blog)
• What is Valkey? A comparison with Redis（Redis）
• Optimizing Cloud Costs by Reducing Backend Memory Usage (GH Media)
• Pinterest's CPU Zombies and SRE Performance Auditing (GH Media)
• Performance Assurance with k6 Load Testing (GH Media)
• Multi-Tenant Design with PostgreSQL RLS (GH Media)