Adam Prout
370 posts
Adam Prout
@a_prout
Postgres @AzureDBPostgres. Ex-MemSQL/SingleStoreDB cofounder. Database builder. One trick pony.
Katılım Haziran 2012
339 Takip Edilen964 Takipçiler
Representing team Postgres... I'll talk about some of the changes we've made to Azure (and to Postgres) to improve the performance/reliability/security of running PostgreSQL in the cloud.
CMU Database Group@CMUDB
Today's Postgres vs. World Seminar Speaker: Adam Prout (@a_prout) will present the architecture of the newly released Microsoft Azure HorizonDB. Zoom talk open to public at 4:30pm ET. YouTube video available after: db.cs.cmu.edu/events/pg-vs-w…
English
@eatonphil It's a good for the MySQL community. That said they're a VC backed startup that will (eventually) be under some pressure to deliver VC-level returns ($$ Billions with a B). I'm not sure how that will work out...
English
New database (fork) dropped. Great to see paths forward for MySQL as a community like this.
English
Adam Prout retweetledi
Spring 2026 @CMUDB Seminar Series: PostgreSQL vs. The World
db.cs.cmu.edu/seminars/sprin…
Starts Mon Feb 2nd @ 4:30pm EST. We will alternate between a speaker from either a @Postgres DBMS or a non-Postgres DBMS.
Open to the public over Zoom. All videos available on YouTube afterwards.
English
@andy_pavlo "Databases year in review" is a great read to start off the year for folks working on (or using) databases.
Good to see the Azure HorizonDB launch made the cut!
Andy Pavlo (@andypavlo.bsky.social)@andy_pavlo
Here is my latest article on the world of databases: cs.cmu.edu/~pavlo/blog/20… All the hot topics from the last year: • More Postgres action! • MCP for everyone! • MongoDB gets litigious with FerretDB! • File formats! • Market movements! • The richest person in the world!
English
feedback welcome.
We're interested in improving PostgreSQL's support for analytical queries at Microsoft. Is pg_duckdb enough?
English
The pg_duckdb extension is now available to use with Azure Postgres. It's a quick way to get faster analytical query results from Postgres via running queries with duckdb's very good vectorized query engine.
#community-4470412-queries" target="_blank" rel="nofollow noopener">techcommunity.microsoft.com/blog/ADforPost…
English
@AlexMillerDB Its closest to Socrates. WAL is quorum committed to specialized WAL storage nodes (Service fabric does leader election).
English
@a_prout I missed the livestream :(
Did y’all align yourselves with one of the existing architectures? How is the WAL replicated, and how are page servers replicated?
English
We’re shipping something new for PostgreSQL on Azure. HorizonDB pairs upstream compatibility with disaggregated storage designed for Postgres. We push most replication and durability work into the storage layer leaving more CPU for PostgreSQL to run queries and transactions.
English
Congrats team @SingleStoreDB
Thats two @nikitabase founded database companies acquired in the past ~3 months for XXX millions each (Neon -> Databricks, MemSQL/Singlestore -> private equity)?
How often has that happened before?
Raj Verma@RajVermaCEO
The journey continues! Excited to announce @capital_vector's growth buyout of @SingleStoreDB. I look forward to partnering with Amish Mehta & team to innovate, serve customers, and build a multi-generational company. SingleStore Announces Growth Buyout Led by Vector Capital - businesswire.com/news/home/2025…
English
@Pruxis @penberg @JasonThorsness rollback as in after an app deployment (AWS Aurora blue/green deployment vs Planetscale Branching) ?
Agree with you that the big clouds don't put enough energy into developer experience.
English
@penberg @a_prout @JasonThorsness As a consumer of dbs idc about the cost of the provider
All i care about is uptime, security, ease of use (how fast can I backup, rollback,…)
Aurora has been very reliable, but ease of use is just not there.
if I’ll ever need to rollback im fkd. Hence im prolly going to ps
English
Great Post. I almost wrote something similar myself after reading all the FUD spit out by PlanetScale about Metal. There are trade-offs between shared-nothing and shared-storage, but shared-storage is the dominate arch for cloud OLTP (AWS Aurora, SQL Hyperscale, Neon, etc.).
Neon Postgres@neondatabase
Separation of storage and compute ≠ slow reads from extra hops On Neon, the compute layer can use LFC (Local File Cache) to serve hot reads without a trip to storage layer.
English
@Pruxis @JasonThorsness I can't speak to the pricing models, only the technical trade-offs involved in the designs.
English
@a_prout @JasonThorsness How come its literally half the price for the same setup on aurora
English
@sunbains @eatonphil yeah, for writes both designs should have similar perf. They both need to send data over the network to make it durable and that should be the dominate cost
English
@eatonphil @a_prout The write performance is determined by the log, a WAL in single write and whatever log if using a consensus protocol, these logs can be on bare metal, EBS etc.
It's not all or nothing.
English
@JasonThorsness Where is the line between FUD and purposefully miss leading folks on the trade-offs?
Metal is classic shared-nothing - a ~40 year old design with well-known issues with cost + elasticity.
English
@a_prout Would you go so far as calling it FUD? Low latency to all data not just working set seems intrinsic to how it works.
English
@philippemnoel @samlambert @nikitabase Definitely some sour grapes going after the big Neon acquisition (not just from Planetscale too).
English
@samlambert @nikitabase Ugh, why? This is so uncalled for.
English
You can fight the trend of vibe coding saying it's not for "real engineers" or not for "production workloads".
No one will remember this in a few year as the industry fully transitions to the new way of doing things.
English
Adam Prout retweetledi
It’s strange how quickly software companies go from making engineers accessible to customers to hiding them behind sales / marketing / evangelists /account management / customer support / customer success / executives / developer relations / lawyers / user research / etc. It’s got to be hard to build something great when you can never talk to the customer.
English
@kellabyte @sriramsubram I know the planetscale folks are pushing hard on Metal, but Aurora/AlloyDB/SQL Hyperscale are all tiered systems already. They cache hot data on fast local NVMe disks on the same machine as the SQL compute ("SSD Cache" in the diagram here):
#hyperscale-architecture-overview" target="_blank" rel="nofollow noopener">learn.microsoft.com/en-us/azure/az…
English
@sriramsubram Agreed but I think we are solving the problems in the wrong ways. We keep going to replication at the DB layer or replication at some kind of network block layer below the DB engine.
In reality what we need is tiered system like a CPU to get benefits of both.
English
Every time we separate compute from storage, we bring it back together it seems. And then we do it again a decade later.
English
@samlambert "SSD Cache" in that diagram (Alloy/Aurora are similar).
English
@samlambert I think you missed my point. They are all caching on local NVMe (no network hop).
#hyperscale-architecture-overview" target="_blank" rel="nofollow noopener">learn.microsoft.com/en-us/azure/az…
English
the true redpill for the DB providers that are doing OLTP on separated storage and compute is that even if you believe in the architecture (i dont) it is copied it from Amazon Aurora. Aurora has 4000 incredible engineers. why not buy it from them? they dont have daily outages
English