Valkey 9.2 is now available, bringing the latest feature update to the open-source, in-memory key-value database that has become the most prominent fork of Redis, according to Phoronix.

The headline additions are zstd compression support and reduced memory overhead per stored structure — two changes that land directly on the concerns teams weigh when comparing Valkey against Redis for production caching and session workloads.

Compression is a workload decision

The most consequential detail for latency-sensitive operators is what zstd support actually asks of a deployment: compressing and decompressing values costs CPU on the hot path, and whether that trade is worthwhile depends entirely on the shape of the data and the latency budget around it.

That distinction matters more than it might appear. Compression features can catch operators who benchmark tail latency before anything else off guard, and they complicate capacity planning for workloads with large values and tight p99 targets. The useful framing for Valkey 9.2 is that compression raises a per-workload question rather than a blanket upgrade decision: memory-frugal configurations for cost-controlled caches on one side, uncompressed configurations for workloads where every microsecond on the read path counts on the other.

Why the memory-overhead numbers matter to cloud bills

The second headline improvement — lower memory overhead — addresses a different cost lever. In-memory databases are typically planned against instance memory ceilings, and the fixed per-key and per-structure overhead that the engine carries on top of the data itself is a real part of that footprint. Across a fleet holding millions of small keys, that structural overhead accumulates into a meaningful share of RAM that is billed but never seen by the application.

Reducing that overhead means more usable data fits within the same instance size, or the same data fits within a smaller one. For teams running managed or self-hosted fleets where memory is the dominant line item, that is a direct cloud-cost win — and it is achievable without changing a single line of application code.

Taken together, the two changes pull in the same direction: each byte of stored data now costs less in raw memory, and values that remain large can be compressed further where the CPU trade-off makes sense.

A practical evaluation point for teams weighing alternatives

For infrastructure and DevOps teams that maintain Redis deployments — particularly those self-hosting or managing fleets where memory is the primary scaling constraint — Valkey 9.2 gives the evaluation a concrete, measurable basis. Both of this release's headline features translate into either lower infrastructure spend or greater operational flexibility, and neither requires application changes to trial: compression can be tested against a staging workload with representative data sizes, while the memory-overhead improvements can be validated by sizing a shadow instance before any migration decision.

The practical takeaway for anyone with an upgrade window: Valkey 9.2 is a feature release worth benchmarking against your own workload rather than a mandatory cutover. Run the numbers on a representative dataset with compression on and off, compare footprint and latency against your current Redis or Valkey version, and let your own p99 and instance-cost figures drive the decision.

Phoronix covered the release as the latest feature update to the in-memory, key-value database.


Valkey 9.2 現已發布,為這款開源的記憶體 key-value 資料庫帶來最新功能更新。據 Phoronix 報導,該資料庫已成為 Redis 最主要的 fork。

本次重點新增功能包括 zstd 壓縮支援,以及降低每個已儲存結構的記憶體開銷——這兩項改動正好切中團隊在比較 Valkey 與 Redis 用於生產環境快取及 session 工作負載時所關注的問題。

壓縮取決於工作負載

對延遲敏感的維運人員而言,最關鍵的細節在於 zstd 支援實際上對部署有何要求:在 hot path 上壓縮和解壓 values 會消耗 CPU,而這項取捨是否值得,完全取決於資料的形態及其周邊的延遲預算。

這個區別的重要性可能超出表面所見。壓縮功能往往令那些以 tail latency 基準測試為首要任務的維運人員措手不及,亦會令數值龐大且 p99 目標嚴格的工作負載在容量規劃上更加複雜。理解 Valkey 9.2 的有用框架是:壓縮提出的是每個工作負載各自的問題,而非一概而論的升級決定——一邊是面向成本受控快取的節省記憶體配置,另一邊則是面向讀路徑上每一微秒都至關重要的工作負載之不壓縮配置。

為何記憶體開銷數字影響雲端帳單

第二項重點改進——降低記憶體開銷——針對的是另一個成本槓桿。記憶體資料庫通常按實例記憶體上限來規劃容量,而引擎在資料本身之上所承載的每 key、每結構的固定開銷,正是這部分佔用的實際組成。在一個擁有數百萬個小 key 的 fleet 中,這類結構性開銷會累積成可觀的 RAM 份額:它會被計費,卻從來不會被應用程式所使用。

降低該開銷,意味著更多可用資料可納入同一實例規格,或同樣的資料可裝入更小的實例。對於管理託管或自建 fleet、且記憶體為主要開支項目的團隊來說,這是直接的雲端成本得益——而且無需改動任何一行應用程式代碼即可實現。

綜合而言,兩項改動方向一致:每 byte 儲存資料所消耗的原始記憶體現在更低,而仍然龐大的 values 則可在 CPU 取捨合理的情況下進一步壓縮。

團隊評估替代方案的實際參考點

對於維護 Redis 部署的基建及 DevOps 團隊——特別是自行託管或管理 fleet、且記憶體為主要擴充限制的團隊——Valkey 9.2 為評估提供了具體且可量化的依據。本版本兩項重點功能均可轉化為更低的基建開支或更大的維運靈活性,而且兩者在試用時都無需更改應用程式:壓縮可在預發布環境以具代表性的資料大小對 staging 工作負載進行測試,而記憶體開銷改進則可在任何遷移決定之前,透過配置一個 shadow 實例來驗證。

對於任何有升級時段的使用者,實際結論是:Valkey 9.2 是一個值得以自身工作負載進行基準測試的功能版本,而非必須執行的切換。請在啟用和停用壓縮的情況下,分別以具代表性的資料集執行測試,將佔用及延遲與現有 Redis 或 Valkey 版本作比較,讓自己的 p99 及實例成本數字主導決定。

Phoronix 就此版本作為該記憶體 key-value 資料庫的最新功能更新作出報導。

新聞來源 / Original News Source