Topic
10 posts
Browse by topic
A single server provides a forgiving environment. Everything lives in one place. The application, file storage, session data, and cron jobs all run on the same machine and can communicate with each other without much thought. Then, when the product gains traction and one server is no longer enough, the team adds a second and…
A founder with AWS startup credits and a product ready for launch is in a good position. They have twelve months of free computing power, genuine market momentum and the problem of how to actually run the thing on a platform designed for a much larger scale. By month six, some of them will have…
Self-hosting has many advantages. You have full control, costs are predictable, and you are not dependent on a platform you don’t own. For the right situation, it’s the right solution. However, most businesses that self-host didn’t make that choice deliberately. They chose it years ago because it was the cheapest option, and they’ve been stuck…
Most first-time cloud setups go wrong in the same way. Someone opens an AWS or DigitalOcean account, starts clicking around, and two weeks later has a running server and an unexpected monthly bill. They also have an architecture that already needs rethinking. I’ve helped untangle enough of these situations to know that the decisions made…
I am asked about Kubernetes more often than I expected, given that most of my small and medium-sized enterprise (SME) clients don’t need it. The pattern is usually the same. Someone’s developer reads an article, a competitor mentions it, or they feel behind after reading a job posting. Then, they call me asking how quickly…
Update, November 2026 UpCloud now offers spot pricing on its GPU servers. For the L4 plan used in this guide, the spot rate is only slightly below the on-demand price, so the saving is small, and spot servers can be interrupted. The prices in this guide are on-demand. Check UpCloud’s pricing page for current rates….
You have probably noticed the pattern by now. First, you experiment with GPT-4 for internal tools. Then, a few customer-facing features appear. Finally, someone forwards you the monthly OpenAI bill, and your eyes widen.At what point does “just using the API” become a problem worth solving?Not every company should self-host LLMs. For many companies, the…
Open-source AI inference without OpenAI API costs. For developers who need control. If you’re running LLM-powered features at any meaningful volume, OpenAI API costs add up fast. At $15 per million output tokens for GPT-4o, a modest production workload can easily hit hundreds of dollars a month – before you even consider the data privacy…
Many years ago, I worked as part of a development team developing and maintaining a large Israeli internet portal called Nana. Development was not where it is today, most of the mechanisms that are widely available for developers today, were written from scratch by our team, but the work was about the same. Infrastructure was…