@platformatic
Account based inItaly
About this account
- Account based in
- Italy
- Connected via
- Italy App Store
Account-level information from X, not a live location or the device used for a specific post.
The platform bridging the gap between Node.js developers, operators, and wider business. So you can focus on building π
Joined April 2022
- Tweets1.2K
- Following115
- Followers3.3K
- Likes1.4K
Pinned Tweet
π₯ New paper is out!
State Without a Landlord. A durable workflow's journal already has the shape of a peer-to-peer log: one writer, append-only, deterministic replay.
We propose sharing it so running workflows can move, resume and be verified without a single owner.
ALT State Without a Landlord: Durable Workflows Without a Single Owner
ποΈ New banter coming up soon!
The Sandbox Was Fine. JavaScript Wasn't!
A discussion between @lucamaraschi & @matteocollina about why teams keep making this mistake and what a better design would look like.
Wed. 30th. 8:00 AM (PT) | 5:00 PM (CEST)
streamyard.com/watch/YXFA43dβ¦
π A new article is out now!
Sizing your worker pool by CPU count alone is a common trap.
On real production hosts, 'available' CPUs are rarely idle. Adding more workers often reduces throughput instead of increasing it.
ALT More Threads on an Already Busy Machine
We see this in benchmarks and live systems. Dive into @matteocollina 's analysis:
blog.platformatic.dev/more-tβ¦
ποΈ New Episode!
Workflows function as an append-only log, and so does HyperCore.
@lucamaraschi explored that line of inquiry and, as a result, produced a 50-page paper in which agent state is replicated peer-to-peer.
ALT State Without a Landlord
We will be discussing our latest release: blog.platformatic.dev/state-β¦
π White Paper
arxiv.org/abs/2609.17645
Ten requests could be ten cache hits or ten heavy renders. Same count, totally different load.
The ICC Planner skips request counts and learns from Event Loop Utilization and heap signals instead the stuff that actually reflects what your app is doing.
blog.platformatic.dev/introdβ¦
Ten requests could be ten cache hits. Or ten heavy renders that max out every core.
Same number. Completely different load.
Most capacity planning starts with request counts anyway, because they're easy to measure, not because they're right.
The signal that actually matters is what your process is doing under the hood, not how many times it got called.
blog.platformatic.dev/introdβ¦
No central repository needed.
No cloud provider in the middle, and all of it runs entirely within your own network.
@lucamaraschi & @matteocollina will explain the origin of the idea, how the various components fit together.
Why should enterprises question the fact that their agentic workloads are leaving the building?
Register now: streamyard.com/watch/7chhP2vβ¦
π Something new is here!
secure-eval-worker, for running @JavaScript in @nodejs with only the permissions you grant.
Moving code to a worker thread doesn't isolate it; it still has filesystem, network and process access unless you strip it.
ALT Secure-eval-worker: Least-Privilege JavaScript Execution in Node.js
This drops permissions before guest code compiles.
blog.platformatic.dev/introdβ¦
@platformatic/memcached, a @nodejs client built on memcached's modern meta protocol.
Full pipelining on one connection, opaque-token verification, zero deps.
350k+ SETs/sec, 369k+ GETs/sec, ~3x the fastest client we tested.
blog.platformatic.dev/introdβ¦
Speeding up your autoscaler doesn't speed up @kubernetesio.
Scheduling, startup, readiness checks, routing, that all still takes real time, no matter how fast you detect the spike. Reacting faster has a ceiling.
Knowing what's coming doesn't.
Stop tuning your autoscaler for speed. Lower thresholds, tighter polling, faster triggers.
None of it moves the needle if @kubernetesio scheduling and pod startup are your real bottlenecks.
You can't outpace the platform's own limits.
If your traffic spike happens every evening, that's not a detection issue, but a predictable pattern your system should anticipate.
Chasing speed won't solve a memory constraint.
Learn more about our @platformatic Autosclaer Planner: blog.platformatic.dev/introdβ¦
π π§ππ° πππ§πππ« ππ©π’π¬π¨ππ!
What if you could execute a package without ever unzipping it?
What if npm had no central server, just a peer-to-peer mesh with cryptographic integrity baked in?
No slides, no answers, no agenda.
Just @lucamaraschi and @matteocollina following the current wave of @npmjs and @github outages to their logical conclusion: π¦ππ²ππ π’π'π¬ ππ’π¦π ππ¨ π’π¦ππ π’π§π π ππ¨π¦π©π₯ππππ₯π² ππ’ππππ«ππ§π π°π¨π«π₯π.
π Register now
Wednesday, September 16th, 8:00 am (PT) / 5:00 pm (CEST)
streamyard.com/watch/FqjGHqcβ¦
"We're not done yet. Once we complete the last three or four improvements, we think we can get our Node.js system to 500β600 MB/s per second per container."
Fabrizio Fenoglio, @supabase
Currently at 300. Target: 500β600.
No new hardware. Same architecture, pushed further.
ALT Platformatic x Supabase