--- title: What five P99 CONF talks reveal about database trade-offs description: Five P99 CONF 2026 talks show engineers at different companies trading read speed for write cost, or memory for latency, depending on workload. Almog Gavra... canonical: https://daily.dev/posts/what-five-p99-conf-talks-reveal-about-database-trade-offs-ugm1hn9ir twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: What five P99 CONF talks reveal about database trade-offs | daily.dev og:description: Five P99 CONF 2026 talks show engineers at different companies trading read speed for write cost, or memory for latency, depending on workload. Almog Gavra... og:url: https://daily.dev/posts/what-five-p99-conf-talks-reveal-about-database-trade-offs-ugm1hn9ir og:image: https://api.daily.dev/og/posts/UgM1HN9Ir.png og:image:alt: What five P99 CONF talks reveal about database trade-offs og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # What five P99 CONF talks reveal about database trade-offs **[The New Stack](https://daily.dev/sources/newstack)** · 8 min read · 0 upvotes · 0 comments ## Summary Five P99 CONF 2026 talks show engineers at different companies trading read speed for write cost, or memory for latency, depending on workload. Almog Gavra frames LSM trees as a dial between read and write amplification, and built SlateDB, an object-store-native LSM library. ScyllaDB's Logstor engine hit 1.1 million reads per second with sub-millisecond p99.9 latency by trading more RAM and write work for faster reads. AWS's Madelyn Olson got Valkey to roughly 1.2 million requests per second, about eight times its prior ceiling, by activating idle I/O threads. TigerBeetle built its own object storage client in Zig, reporting uploads about 20 times faster than official SDKs. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How much faster is ScyllaDB's Logstor engine compared to its existing LSM engine for read-heavy workloads? In a preliminary read-dominant test, ScyllaDB's Logstor storage engine sustained 1.1 million reads per second with p99.9 latency below one millisecond and p99.99 latency around 1.5 milliseconds. Logstor uses an in-memory index with a hash and flash location per key, letting a read complete with one disk I/O instead of an LSM search across SSTables and Bloom filters. The trade-off is a larger RAM index and heavier writes, since even small updates rewrite the full value. _Evaluating storage engine trade-offs for a read-heavy workload? daily.dev surfaces benchmarks like this as they appear._ ### How did AWS improve Valkey's throughput ceiling for single-threaded command execution? AWS engineer Madelyn Olson reported raising Valkey's throughput ceiling to about 1.2 million requests per second, roughly eight times the prior limit, in an unpipelined GET benchmark with about three million keys and 650 clients. The fix used lockless queues so existing I/O threads handled network I/O and command parsing while the main thread kept executing commands single-threaded. Valkey 9.0 also added atomic slot migration for faster horizontal scaling. _Scaling Redis-compatible caches? daily.dev tracks Valkey performance changes like these as they land._ ### Why did TigerBeetle build its own object storage client instead of using the AWS or GCP SDKs? TigerBeetle built a custom multi-cloud object storage client in Zig because official SDKs conflicted with its single-core transaction processing design, citing the AWS C++ SDK's blocking getaddrinfo call as one mismatch along with SDK thread pools and runtime allocation. The custom client, under 40,000 lines with no external dependencies, uses io_uring and zero-copy sends, and reported uploads about 20 times faster than official SDKs in an 8 MB PutObject test. _Weighing a custom storage client against vendor SDKs? daily.dev follows engineering trade-offs like TigerBeetle's._ ## Similar posts on daily.dev - [HPTS 2026 \(Day 2 and Day 3\)](https://daily.dev/posts/hpts-2026-day-2-and-day-3--xashwu5jv) · Metadata · 0 upvotes · 0 comments - [ScyllaDB cut Sprig’s read latency 4X after Redis and ClickHouse hit a wall](https://daily.dev/posts/scylladb-cut-sprig-s-read-latency-4x-after-redis-and-clickhouse-hit-a-wall-ya5cnipcg) · The New Stack · 0 upvotes · 0 comments - [10X more data, same 4 seconds: single-query scaling in Redpanda SQL on 1TB](https://daily.dev/posts/10x-more-data-same-4-seconds-single-query-scaling-in-redpanda-sql-on-1tb-qj6rt0fu2) · Redpanda · 0 upvotes · 0 comments --- Tags: [#database](https://daily.dev/tags/database), [#scylladb](https://daily.dev/tags/scylladb) [View this post on daily.dev](https://daily.dev/posts/what-five-p99-conf-talks-reveal-about-database-trade-offs-ugm1hn9ir) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"What five P99 CONF talks reveal about database trade-offs","url":"https://daily.dev/posts/what-five-p99-conf-talks-reveal-about-database-trade-offs-ugm1hn9ir","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/what-five-p99-conf-talks-reveal-about-database-trade-offs-ugm1hn9ir"},"datePublished":"2026-10-30T13:04:27.810Z","dateModified":"2026-10-30T13:05:00.993Z","description":"Five P99 CONF 2026 talks show engineers at different companies trading read speed for write cost, or memory for latency, depending on workload. Almog Gavra...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/61e5d5e11319b04f14684e090b9a457b?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/61e5d5e11319b04f14684e090b9a457b?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"The New Stack","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"The New Stack","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/newstack","url":"https://daily.dev/sources/newstack"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/what-five-p99-conf-talks-reveal-about-database-trade-offs-ugm1hn9ir","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"database,scylladb","timeRequired":"PT8M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"The New Stack","item":"https://daily.dev/sources/newstack"},{"@type":"ListItem","position":3,"name":"What five P99 CONF talks reveal about database trade-offs"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/what-five-p99-conf-talks-reveal-about-database-trade-offs-ugm1hn9ir#faq","mainEntity":[{"@type":"Question","name":"How much faster is ScyllaDB's Logstor engine compared to its existing LSM engine for read-heavy workloads?","acceptedAnswer":{"@type":"Answer","text":"In a preliminary read-dominant test, ScyllaDB's Logstor storage engine sustained 1.1 million reads per second with p99.9 latency below one millisecond and p99.99 latency around 1.5 milliseconds. Logstor uses an in-memory index with a hash and flash location per key, letting a read complete with one disk I/O instead of an LSM search across SSTables and Bloom filters. The trade-off is a larger RAM index and heavier writes, since even small updates rewrite the full value. Evaluating storage engine trade-offs for a read-heavy workload? daily.dev surfaces benchmarks like this as they appear."}},{"@type":"Question","name":"How did AWS improve Valkey's throughput ceiling for single-threaded command execution?","acceptedAnswer":{"@type":"Answer","text":"AWS engineer Madelyn Olson reported raising Valkey's throughput ceiling to about 1.2 million requests per second, roughly eight times the prior limit, in an unpipelined GET benchmark with about three million keys and 650 clients. The fix used lockless queues so existing I/O threads handled network I/O and command parsing while the main thread kept executing commands single-threaded. Valkey 9.0 also added atomic slot migration for faster horizontal scaling. Scaling Redis-compatible caches? daily.dev tracks Valkey performance changes like these as they land."}},{"@type":"Question","name":"Why did TigerBeetle build its own object storage client instead of using the AWS or GCP SDKs?","acceptedAnswer":{"@type":"Answer","text":"TigerBeetle built a custom multi-cloud object storage client in Zig because official SDKs conflicted with its single-core transaction processing design, citing the AWS C++ SDK's blocking getaddrinfo call as one mismatch along with SDK thread pools and runtime allocation. The custom client, under 40,000 lines with no external dependencies, uses io_uring and zero-copy sends, and reported uploads about 20 times faster than official SDKs in an 8 MB PutObject test. Weighing a custom storage client against vendor SDKs? daily.dev follows engineering trade-offs like TigerBeetle's."}}]} ```