Comments URL: https://news.ycombinator.com/item?id=47271790

Points: 1

# Comments: 2

# Background (The challenge and the subject system)

My goal was to improve performance/cost ratio for my Kubernetes cluster. For performance, the focus was on increasing throughput.

The operations in the subject system were primarily CPU-bound, we had a good amount of spare memory available at our disposal. Horizontal scaling was not possible architecturally. If you want to dive deeper, here's the code for key components of the system (and architecture in readme) - https://github.com/rudderlabs/rudder-server, https://github.com/rudderlabs/rudder-transformer, https://github.com/rudderlabs/rudderstack-helm.

For now, all you need to understand is that the Network IO was the key concern in scaling as the system's primary job was to make API calls to various destination integrations. Throughput was more important than latency.

# Solution

Increasing CPU when needed. Kuberenetes Vertical Pod Autoscaler (VPA) was the key tool that helped me drive this optimization. VPA automatically adjusts the CPU and memory requests and limits for containers within pods.

# What I liked about VPA

* I like that VPA right-sizes from live usage and—on clusters with in-place pod resize—can update requests without recreating pods, which lets me be aggressive on both scale-up and scale-down improving bin-packing and cutting cost.

* Another thing I like about VPA is that I can run multiple recommenders and choose one per workload via spec.recommenders, so different usage patterns (frugal, spiky, memory-heavy) get different percentiles/decay without per-Deployment knobs.

# My challenge with VPA

One challenge I had with VPA is limited per-workload tuning (beyond picking the recommender and setting minAllowed/maxAllowed/controlledValues), aggressive request changes can cause feedback loops or node churn; bursty tails make safe scale-down tricky; and some pods (init-heavy etc) still need carve-outs.

That's all for today. Happy to hear your thoughts, questions, and probably your own experience with VPA. I did learn that k8s team is working on some of these feedback and we might see it get solved in 2026.

Comments URL: https://news.ycombinator.com/item?id=46510959

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=46463769

Points: 2

# Comments: 0

Revolut's 2024 revenue grew 72% to $4 billion with profit before tax increasing 149% to $1.4 billion Disruption Banking. Unlike most fintech darlings, they're actually profitable and growing fast across 65 million customers.

But here's the kicker: this is a secondary sale—meaning existing shareholders are cashing out. When insiders are selling at these valuations, it's worth asking who's buying the peak. Especially when traditional banks are still struggling to innovate and Revolut is building a global bank from scratch.

The company is solid, but $75 billion for a neobank? That's more than many established global banks. The fintech premium is alive and well, even as the AI bubble inflates around it. What do you think?

Comments URL: https://news.ycombinator.com/item?id=46042430

Points: 3

# Comments: 3

[1] Ref: https://news.ycombinator.com/item?id=44659678

Comments URL: https://news.ycombinator.com/item?id=45723907

Points: 1

# Comments: 0

No need to make an exhaustive list, sharing even one can help the discussion.

Comments URL: https://news.ycombinator.com/item?id=45564089

Points: 1

# Comments: 0

Thanks for building this. I am trying to set it up but facing this issu

> `torch` (v2.3.1) only has wheels for the following platforms: `manylinux1_x86_64`, `manylinux2014_aarch64`, `macosx_11_0_arm64`, `win_amd64`

Comments URL: https://news.ycombinator.com/item?id=45470953

Points: 1

# Comments: 0