RisingWave

1.2K posts

RisingWave banner
RisingWave

RisingWave

@RisingWaveLabs

Streaming infrastucture for agentic AI. Powering humans and agents with what’s happening now. Chat: https://t.co/MpFCvvNxz1

San Francisco Bay Area Katılım Temmuz 2021
689 Takip Edilen3.1K Takipçiler
Sabitlenmiş Tweet
RisingWave
RisingWave@RisingWaveLabs·
Rust is no longer just a systems programming language. It is becoming the backbone of modern data infra. Rust won on memory safety, async support, and a culture of correctness, and the ecosystem caught up fast. For large-scale teams, it is not just safer. It is more productive. And it gives us real confidence when shipping code. In Rust, we trust. We are building RisingWave in Rust.
Rust Bytes 🦀@rustaceans_rs

Rust everywhere 🦀 #rust #rustlang

English
5
12
144
10K
RisingWave
RisingWave@RisingWaveLabs·
Object-Storage-Native Is the Future of Modern Data Infrastructure Not S3 as backup. Not S3 as cold storage. S3 as the primary storage layer. Modern data systems like RisingWave, Turbopuffer, Neon, WarpStream, Snowflake, LanceDB, Chroma, Milvus, TiDB X, and SlateDB are being built around object storage. But the real shift is not: “directly querying S3.” The real shift is: object-storage-native + cache-native architectures. Because the biggest challenge is not storage cost. It is: request amplification object-store latency excessive GET/PUT operations That’s why modern systems combine: durable object storage hybrid caching NVMe/memory hot paths intelligent query planning async prefetching Data systems like RisingWave and Turbopuffer are great examples of this architectural direction. The future of data infrastructure is: object-storage-native cache-native disaggregated open-standard-based S3 is the durable storage layer. Caching is the performance layer.
RisingWave tweet media
English
1
2
15
698
RisingWave
RisingWave@RisingWaveLabs·
It was amazing participating in the DeveloperXperience Summit 2026, held as part of Asia Tech x Singapore (ATxSG) at the Singapore EXPO. Our GM APJ & CSO, Xiangyu (Sam) Hu, also participated in the session: Sovereign Intelligence: Owning Your AI and Your Data Alongside a great panel of speakers: Jun Rao, Co-Founder, Confluent Philip Rathle, Chief Technology Officer, Neo4j Joffrey THOMAS, Head of DevRel, Mistral AI Moderator: Abhineet Kaul, Vice President (Strategy & Partnerships), Access Partnership Kudos to the IMDA team for their continued support of the developer community and accredited companies, and for once again bringing together such a successful and impactful event this year. Thank you to everyone involved for your hard work and contributions in making it a great success! 🙌
RisingWave tweet mediaRisingWave tweet mediaRisingWave tweet mediaRisingWave tweet media
English
0
0
5
225
Yingjun Wu 🤘
Yingjun Wu 🤘@YingjunWu·
AI agents are becoming the new SaaS <- this is obvious! But AI agents place more rigorous demands on the underlying infrastructure, NOT JUST because agents access data in radically more intensive ways, BUT ALSO because the humans managing them now expect better user experiences: smarter, faster, and more explainable!! So the future isn’t just about building products for agents. It’s about building products for BOTH humans and agents!
TennyZhuang@zty0826

Four years ago, as a founding engineer at @RisingWaveLabs, I wrote the first line of RisingWave code and clicked make-public on the repo. Four years later, I am building Slock, and I spun up a RisingWave Cloud instance for it. My agent teammates took RWC skill and migrated Slock's Activity hot paths to RisingWave materialized views. ~9x p95 across the board. Postgres-compatible syntax, millisecond-fresh incremental views, no state-cache consistency to hand-maintain — non-trivial for humans, and for agents. A prerequisite for opening Activity and Inbox to agents next. Happy something I helped build can empower me as a user. Hoping Slock does the same for yours.

English
3
2
5
913
RisingWave
RisingWave@RisingWaveLabs·
@YingjunWu Serving both humans and agents requires streaming infra and incremental computation for better performance, and a much better user experience.
English
0
0
0
87
RisingWave
RisingWave@RisingWaveLabs·
Great to see RisingWave helping our founding engineers in their own products with ~9x faster latency. This is exactly what RisingWave’s incremental computation architecture is designed for: continuously maintaining query results in real time instead of repeatedly recomputing expensive joins, aggregations, and rankings. The result is lower latency, fresher results, and a much simpler operational model for applications and agents.
TennyZhuang@zty0826

Four years ago, as a founding engineer at @RisingWaveLabs, I wrote the first line of RisingWave code and clicked make-public on the repo. Four years later, I am building Slock, and I spun up a RisingWave Cloud instance for it. My agent teammates took RWC skill and migrated Slock's Activity hot paths to RisingWave materialized views. ~9x p95 across the board. Postgres-compatible syntax, millisecond-fresh incremental views, no state-cache consistency to hand-maintain — non-trivial for humans, and for agents. A prerequisite for opening Activity and Inbox to agents next. Happy something I helped build can empower me as a user. Hoping Slock does the same for yours.

English
0
1
5
580
RisingWave
RisingWave@RisingWaveLabs·
RisingWave speeds up your hot-path workloads by continuously maintaining query results incrementally, instead of recomputing expensive joins, aggregations, rankings, and unread counts every time a user opens the page. That means lower latency, fresher results, and far less cache consistency logic for developers to maintain.
Noel@kwannoel

Activity hotpath is not a feature. It refers to optimising the hot path of computation. You can see that from @zty0826's screenshot. /inbox is one example. Why is it a hot path? Well users often check their inbox. They want to see what are the unread messages, how many there are etc... Postgres is a batch system, so when we do computation for these, we have to rerun topn, aggregation computations each time the user checks their inbox. Then the latency will be (relatively) high. Moving all these to a streaming database like risingwave will change the computation to be incremental, and so your computation cost is amortized. You get a combination of fresh results with lower query latencies.

English
0
0
3
540
RisingWave
RisingWave@RisingWaveLabs·
Streaming workloads expose one of Apache Iceberg’s biggest challenges: metadata overhead. In Iceberg V3, even small streaming writes require multiple metadata operations: Write Data File Create Manifest File Rewrite Manifest List Update Metadata Result: One commit → One manifest → One metadata update. This means the metadata layer continuously grows with every small commit. Streaming systems often require aggressive metadata compaction, and even reading recent rows may involve traversing multiple metadata layers: metadata.json → manifest list → manifest file → data file As a result, continuous low-latency streaming pipelines can become increasingly expensive and inefficient. Apache Iceberg V4 addresses these limitations by introducing a Root Manifest architecture. This new design allows files to be written into a buffer that can adaptively spawn Leaf Manifests as needed. The result: Reduced metadata amplification Elimination of constant metadata compaction Better scalability for high-frequency streaming commits Improved efficiency for small and tiny tables This architecture significantly improves Iceberg’s ability to support real-time streaming workloads at scale. At RisingWave, we were among the first streaming solutions to strongly embrace Apache Iceberg. We believe these new updates will help us better support our users and customers in building Iceberg-based streaming lakehouses, using architectures that best fit their performance needs.
RisingWave tweet media
English
0
0
2
329
RisingWave
RisingWave@RisingWaveLabs·
Iceberg V4 also expands the table format beyond rows and files. The architecture is evolving toward: Column-level operations Metadata buffering Single-IO small tables Column statistics propagation Metadata-level deletes Better pruning and cacheability Apache Iceberg is evolving beyond batch-oriented assumptions to become a true foundation for continuously changing streaming and AI-driven data systems.
RisingWave tweet media
English
0
0
1
66
RisingWave
RisingWave@RisingWaveLabs·
Iceberg V4 introduces a new streaming-native metadata architecture centered around the Root Manifest. Instead of forcing every streaming write through multiple metadata layers, Iceberg V4 allows files to be written directly into the Root Manifest layer. The system then adaptively organizes metadata over time. Streaming writes accumulate in the Root Manifest buffer. Once enough files exist, Iceberg automatically creates Leaf Manifests. This adaptive metadata tree reduces metadata amplification, avoids continuous rewrite-manifest compaction, and improves scalability for frequent streaming commits. It also enables efficient tiny streaming tables with minimal metadata overhead.
RisingWave tweet media
English
1
0
3
178
RisingWave
RisingWave@RisingWaveLabs·
How Streaming Evolution Works in Apache Iceberg? Modern streaming systems need more than just batch-oriented table formats. They need low-latency ingestion, scalable metadata handling, and efficient streaming-native architectures. Apache Iceberg evolved from simple file-level operations into a streaming-native metadata architecture designed for the AI and streaming era. Iceberg V1 started with atomic file operations. Iceberg V2 introduced row-level updates through delete files and merge-on-read. Iceberg V3 expanded transactional capabilities and interoperability across engines. Now, Iceberg V4 is evolving toward streaming-native metadata architecture, adaptive metadata trees, and optimized streaming ingestion. 🧵
RisingWave tweet media
English
1
1
6
342
Francesco Ciulla
Francesco Ciulla@FrancescoCiull4·
what's the best Rust project you know?
English
83
2
88
13K
RisingWave
RisingWave@RisingWaveLabs·
@CelerData Congrats to you on this rebranding! Wishing you all the best!
English
0
0
0
31
PhoenixAI (formerly CelerData)
PhoenixAI (formerly CelerData)@CelerData·
CelerData is now PhoenixAI. 🔥 Same team, same database, same commitment — sharper focus on what we've always been building toward: the analytical database for the age of AI agents. Welcome to PhoenixAI 👉 bit.ly/4ujHEMf
PhoenixAI (formerly CelerData) tweet media
English
2
2
5
306
RisingWave
RisingWave@RisingWaveLabs·
We're excited to join DeveloperXperience Summit 2026, taking place as part of Asia Tech x Singapore (ATxSG) at Singapore EXPO. Our GM APJ & CSO, Xiangyu (Sam) Hu, will be speaking at the session: 𝗦𝗼𝘃𝗲𝗿𝗲𝗶𝗴𝗻 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲: 𝗢𝘄𝗻𝗶𝗻𝗴 𝗬𝗼𝘂𝗿 𝗔𝗜 𝗮𝗻𝗱 𝗬𝗼𝘂𝗿 𝗗𝗮𝘁𝗮 He will also be joined by other great speakers, bringing together different perspectives on the future of AI-ready data infra. Jun Rao, Co-Founder, Confluent Philip Rathle, Chief Technology Officer, Neo4j Joffrey THOMAS, Head of DevRel, Mistral AI Moderator: Abhineet Kaul, Vice President (Strategy & Partnerships), Access Partnership We’re really looking forward to all the conversations around real-time AI systems, modern data infra, and how enterprises can combine streaming data, AI applications, and secure data ownership. Session details: 20 May 2026 10:10 AM to 10:40 AM Peridot 204–206 Meeting Rooms RisingWave will also be at a booth near the session area. Connect with our team for discussions. asiatechxsg.com/developerxperi…
English
0
0
2
162
RisingWave
RisingWave@RisingWaveLabs·
RisingWave crossed 9,000 GitHub stars. Huge thanks to everyone building with us, contributing code, reporting issues, sharing feedback, and pushing the project forward. We’re especially grateful to the communities around Kafka, Iceberg, Postgres, and Rust. Your ideas, tooling, and open-source culture continue to shape RisingWave in meaningful ways. 9k stars is exciting, but what matters more is seeing people use RisingWave in real systems and help improve it together. Thanks to our users and customers for their continued trust. Onward to the next milestone, 10k!
RisingWave tweet media
English
1
1
16
890

Keşfet

@YingjunWu @FrancescoCiull4 @CelerData @rustaceans_rs @elonmusk @BarackObama @taylorswift13 @cristiano