loading...

Friday Blast #21

horia141 profile image Horia Coman Originally published at horia141.com on ・3 min read

Online migrations at scale (2017) - an article from Stripe engineering about how they performed a large-scale migration of their core data structures without any downtime. Now, not all migrations are this complex, but some are. And in this particular case, in any setup you’d have - large or small - you’d be stuck doing the same dance. I found the use of Scientist to be particularly interesting. This is a Ruby library which aids in refactoring data paths - they made sure that all of their new read paths produced the same results as their old read paths, and had extensive error reporting and monitoring around it.

Why configuration management and provisioning are different (2017) - an article from ThoughtWorks on the differences between the two. Provisioning deals with building the infrastructure for a project: booting machines, setting up networks, roles in IAM etc. So it is a global concern. Configuration management deals with making sure a single machine is configured: setting up the OS, firewall rules, patches etc. So it is a local concern. The author expresses concern over people confusing the two, and using configuration management for provisioning, which leads to bad results.

What you want is what you don’t: understanding trade-offs in distributed messaging (2015) - the author makes the point that strong guarantees from lower levels of your distributed infrastructure (ie message queues) are not that good. First, because they impose a big cost on everyone, even if they need it or not. Second, because things like exactly-once delivery are technically impossible to provide without help from the application, and any tool claiming otherwise is exposing you to problems down the road.

You cannot have exactly-once delivery (2015) - linked with the previous article, this is a deeper dive into why precisely “exactly-once” delivery of messages in a distributed system is impossible, and why we must settle for “at most once” or “at least once”.

The first few milliseconds of an HTTPS connection (2009) - a bit older but still relevant. There’s a lot happening in the TLS/SSL layer to allow HTTPS - much more involved than HTTP itself. For application programmers, it’s interesting to know that 3 extra roundtrips are added at the start of the connection, which adds some latency to each new connection.

Do you really want to be doing this when you’re 50? (2012) - part of a series of links about career advice. This one takes a more pessimistic approach, claiming that a job in programming is a high-stress one, which isn’t what you want when you’re 50. That’s a long way off though.

Do you really want to be making this much money when you’re 50 (2012) - an answer to the previous article, arguing that programming is in fact a much lower stress job than medicine/law etc., with a lot of upsides and with comparable pay for high-performers.

Options vs cash (2017) - startup options are basically worthless in the vast majority of time. Even when they’re not, they’re as good as working at a bigger company. Sometimes, like when you throw a coin and it lands on its edge, you hit it big. They’re also way complex - liquidation preferences, dilution, ISOs etc. seem like a lot of complexity.

Stock options: a balanced approach (2015) - another take on the options vs cash issue. Options are insurance in case the company goes big, rather than actual compensation. Which does make sense. Also, it’s a bit of a critique of Dan Luu’s assumptions - which most definitely don’t apply to everyone in IT for various reasons. From not having the skills to not being in the right place.

Posted on by:

horia141 profile

Horia Coman

@horia141

Leading the Bolt teams in Bucharest. We're working on cool products like food delivery, and interesting and challenging platforms like geo, A/B testing, user accounts and route tracking.

Discussion

markdown guide