draped up technology brother. irl soyjack wife guy.
Nov 29, 2023 • 11 tweets • 4 min read
For every 10 likes this gets, I'll ask ChatGPT to make this Kubernetes Administrator a bit more weary.
even more weary!
Nov 17, 2021 • 10 tweets • 3 min read
Over the year I've been at @planetscaledata, developers have consistently asked us the question: "when are you going to support PostgreSQL?"
The answer: We aren't. That's it. But why? 🧵 1/
There are two sides to it: the market and our team. And what we believe the market is signaling is that PostgreSQL vs MySQL is effectively a petty skirmish at this point. There are huge issues than these communities have not positioned themselves to solve. 2/
Aug 25, 2021 • 11 tweets • 2 min read
In the modern context of horizontally-partitioned SQL databases which work fairly well, scaling by reading from replicas is positively antiquated. Let me explain... 1/
A common, tried-and-true web scalability tool is to replicate your database to other replicas which allow read queries, and send a subset of your read queries to those nodes. This requires a bit of care. 2/
Jul 8, 2021 • 25 tweets • 5 min read
Why is their no relatively easy-to-use, decomposed library for distributed fault-tolerant consensus (i.e. "libpaxos")? Why do we have to use etcd or ZooKeeper or comply with a very needy Raft library? It's Time For Some Game Theory... 🧵
You may have heard of Paxos. It is hard to implement correctly in practice, but it's based on a very basic concept: voting. If the majority of nodes vote for a given value, then that is the value. Minority be damned! But voting isn't enough; Paxos is just the conceptual kernel.
Nov 20, 2020 • 7 tweets • 1 min read
I've always had a hard time taking the whole "agile" movement seriously given that, in my experience, velocity and agility have SO MUCH MORE to do with the software's architecture than how work is planned. [1/7]
"Waterfall" exists to hedge against the risk of making a mistake. We know this isn't very effective though as often the biggest risks are unearthed during the construction process itself. [2/7]
Oct 12, 2020 • 12 tweets • 3 min read
There are two buried leads in the Segment/Twilio news:
1) This is the biggest YC acquisition by a large margin. Cruise was next at $1B.
2) The company was setup so that everyone on the team shared in this win, including alumni. Let's expand on this...
[1/11]
2/ Point #2 was long-term thinking that started many years ago when they eliminated the post-employment exercise time limit (typically 90 days) from their employee option agreements. This is *KEY*
Queues are bad, but software developers love them. You'd think they would magically fix any overload or failure problem. But they don't, and bring with them a bunch of their own problems.
First off, queues turn your system into a liar. Convert something to an async operation and your system will always return a success response. But there's no guarantee that the request will actually ever be processed successfully.
Apr 26, 2018 • 14 tweets • 2 min read
"Why won't people listen to me???" cries the engineer who wants more influence, thinking that it's just a matter of getting the right role or title or reporting to the right person. I struggled quite a bit with this for a number of years.
This simplistic thinking is behind promotions at those stodgy, bureaucratic organizations who we all love to hate. Ever have your thoughts or work baselessly steamrolled by someone senior or with "Architect" or "Principal" or "Fellow" in their title? Exactly.
