We had an eventful July and have exciting news to share!
Your laptop is full of caches that manage themselves; you never manually tune them, set invalidation logic, or worry about the wrong result. Database caching has traditionally been one of the last places where caching is still a manual process. This month, we launched the Readyset Platform, making database caching invisible.
The Readyset Platform combines QueryPilot, rdst, and new components to take the automatic query caching you know Readyset for and deliver a self-tuning query acceleration layer. It all gets running in a few clicks, now available as a desktop app so you can test and prove your ideas with ease. With freshness that tunes itself, eviction that can't be thrashed, and built in debugging, it's the cache you can just leave on.
"Readyset has helped us reclaim capacity from our database infrastructure. Our optimization has moved from low config to no config. Readyset automatically identifies which queries are most costly, caches them, and tunes the freshness of each one. That ongoing optimization is now fully automated, and it has meaningfully slowed the growth of our database infrastructure."
Khanh Tran, VP of Engineering at Hussle
Try The Readyset Platform today.
- 75% Less CPU Load on Shallow Cache Workloads: Disabled the legacy
nom-sql parser when running Readyset with the now-default --cache-mode shallow, cutting CPU load on shallow cache workloads by roughly 75%.
- Improved Shallow Cache Restart Reliability: Dropped shallow caches no longer reappear after an adapter restart, hint-created shallow caches keep their sub-second
COALESCE windows across restarts, and shallow caches now serve traffic after an adapter restart. Fewer surprises after a restart, more consistent behavior end to end.
- New Observability for Shallow Caches: Added Prometheus gauges and histograms for adaptive refresh load, refresh scheduler queue depth,
COALESCE wait times and outcomes, and the shallow refresh worker pool. Also added a new readyset.shallow_cache_stats virtual relation exposing per-cache adaptive load, over-cap state, scheduler queue depth, and hit rate.
- Stronger Postgres Prepared Statement Handling: Fixed unnamed prepared statements from returning rows resolved against the wrong schema after changing
search_path, and stopped them from staying pinned to the upstream after a transaction (which was bypassing the cache for the rest of the connection's life).
- Wider Postgres Cache Coverage:
CREATE CACHE for shallow caches no longer fails against Postgres upstreams when the query references columns by ordinal position in ORDER BY or GROUP BY (for example, ORDER BY 2, 1), including the SELECT DISTINCT case that was previously rejected outright.
- Improved Shallow Cache Observability:
EXPLAIN LAST STATEMENT now reports the shallow cache's name on a hit, and identifies misses instead of making it look like a normal upstream bypass. SHOW PROXIED QUERIES also now lists the queries Readyset declined to cache automatically, along with why.
- Wider Automatic Cache Coverage: Fixed the
CREATE CACHE support check so several classes of supported queries are no longer falsely treated as unsupported. Automatic shallow-cache creation now recognizes MySQL aggregate and window functions, newer Postgres functions, and Postgres row constructors.
- And more!
We had two releases in July, check them out here and here.
Stop Buying Hardware to Fix Your Database | Readyset Platform Explained: A video walkthrough of how The Readyset Platform works in under 3 minutes!
The Real Cost of Read Replicas (and When They Stop Paying Off): Read replicas are the reflex fix for growing read traffic, but the cost scales linearly with no ceiling. This piece covers where replicas earn their keep, where they stop paying off, and how SQL-layer caching handles repetitive reads for a fraction of the spend.
TopK is now Generally Available: The TopK operator is now generally available in Readyset. Cache ORDER BY ... LIMIT queries like leaderboards and top-N feeds with a small constant of memory per key, and tune the buffer to your workload.
Readyset Internals: The left-right pattern: lock-free cache: How does Readyset serve cached reads without ever taking a lock? Explore the mechanics and real-world benchmarks of the left-right pattern and its 1.2-nanosecond atomic pointer swap.
How the AI Buildout Is Repricing the Rest of Your Infrastructure: When Apple and Xbox both say they can no longer absorb rising memory costs, it's time to start asking where that pressure is going next. Our latest research briefing follows the memory shortage from the AI buildout down to your database bill, and lays out which moves pay for themselves at current prices.
Live Webinar: Your Cache Should Tune Itself
Lower infrastructure costs. No cache maintenance. Faster development. See self-tuning caching for Postgres and MySQL in action. Tuesday Aug 18, 8 AM PT / 11 AM ET.
Register for free and join us live
Can't make it live? Register anyway and we'll send the recording, no strings attached.
Questions, bugs, ideas, or just want to say hi? Find us on Slack. Looking to contribute? Our newcomer issues are a good place to start.
Readyset Team
