[Dev Catch Up #15] - GPUs-vs-CPUs, LLMs and programming in 2024, and more.
Bringing devs up to speed on the latest dev news from the trends including , and a bunch of exciting developments and articles.
Welcome to the 15th edition of DevShorts, Dev Catch Up!
I write about developer stories and open source, partly from my work and experience interacting with people all over the globe.
Some recent issues from Dev Catch up:
Join 1000+ developers to hear stories from Open source and technology.
Here’s this week’s catch up!
Must Read
In this era of AI and LLMs, we have seen an extensive rise in the use of GPUs and CPUs. Both of them help in processing AI models of different kinds. GPUs help in faster processing of models and GPU-powered libraries can be a solution to your problem with speed for using CPU based python libraries like NumPy, Pandas, or Scikit-learn. A benchmark showed CuDF, a GPU-based replacement of Pandas to be 15x or 80x faster. While this speedup is impressive and can be accurate by adding some context, simply stating it will be misleading. This article from Pythonspeed tells you in detail how there are misconceptions on GPU vs CPU benchmarks and gives you a better understanding on the comparisons.
With AI and ML making its presence felt in recent days, technologies around AI are also taking a leap forward. We have already talked about vector databases in the past which are particularly useful for applications that deal with AI and ML use cases which require geometric or spatial data. In recent times, a number of organisations are focusing on vector databases and it pops up a question that whether this is the peak for vector databases. Doug Turnbull explains this thought beautifully in this article where he dives deep into the topic and gives you a clear understanding of the argument.
2024 has just started and we are already experiencing a wave of tech innovations. Last year was a special one for Artificial intelligence and this year will also expect to get a wave of innovations on AI and LLMs. Programming with AI and LLMs have come like a blessing for seasoned programmers. Know more about it from this article titled “LLMs and programming in the first days of 2024” published by Antirez, where he discussed in detail on the experience of writing high-level code with the help of LLMs.
GitHub actions is one of the most popular CI/CD platforms that helps in automating your workflows. It gets triggered when you want to make a change on your existing repository by means of a pull request. Actions can now be turned into a time-sharing supercomputer with the help of actions-batch meaning submitting a job now will give you the results at some point in the future. Actions-batch is an unofficial API that does this efficiently. With actions-batch, you can include secrets, fetch the output of the build, use self hosted or hosted runners. Learn more about it from the article published by its developer Allex Ellis himself, where he has given several examples and described how actions-batch can help in different scenarios.
OpenAI and ChatGPT always stay as a hot topic in the news of the tech world. So, any changes in the activities of the AI-enabled chatbot doesn’t stay away from the hawkeyes of the developer community. Recently, some changes were pushed to the GPT-4 model by OpenAI, which drastically changed ChatGPT-4’s browsing capabilities for the past weeks. The recent changes to the model’s interface now prevents the chatbot to directly cite quotes from webpages and limits viewing of full contents, thus significantly altering the model’s web browsing tools. Have a look at this article published by dmicz devblog that uncovers the underlying mechanism of GPT-4 web browsing, the hidden changes, and their implications.
Now, we will head over to some of the news and articles that will be at a place of interest for developers and the tech community out there.
Good to know
The monolithic vs microservice architecture tug of war always has a winner based on circumstances. People at Doordash have recently changed to a microservices architecture and have reduced the time required for development, testing, and deployment. In the meanwhile, it has improved scalability and resiliency for end users and consumers. With the increase of back-end and microservices, the cross availability zones data transfer costs also got a shoot up. These data transfers are highly critical because it helps Doordash manage a highly available service but the surge of pricing prompted the Doordash team to use a service mesh to reduce the cost without sacrificing the service quality. Learn more about this change from this article published by the Doordash team where they talked more about this service in detail and give a detailed analysis on the process.
The open-source tool of the week is coming from the Apache foundation. With over 2.5k forks and almost 5k likes, they are one of the most popular repositories on GitHub in recent days. This tool manages data storage of large analytical datasets on distributed file systems such as cloud stores, HDFS, etc. It can atomically publish data with rollback support and save savepoints for data recovery. Also, it can isolate snapshot between writer and queries and manage the file sizes and layouts using statistics and support multiple types of queries like snapshot, incremental, and read-optimised queries. Check out Hudi from their GitHub page here and try it out.
This article also provides insight into the OpenTelemetry data model and how it differs from the Prometheus data model. It is a guide to using the OpenTelemetry Kubelet Stats Receiver to collect node, pod, container, and volume metrics from the API server, sending them for further processing.
Debugging is an art in software engineering, and effective problem-solving in incident response goes beyond just fixing errors; it requires a deep understanding of the system. To respond to the challenge, the piece delves deeper into the concept of problem detection and dynamic fault management, exploring the layers that transform debugging from a task into an art form. Early problem detection is a nuanced process that goes beyond simple error spotting, and it involves a blend of technical acumen, situational awareness, and experience. The ability to detect problems early can be honed through experience and by ensuring detailed production readiness documentation and knowledge sharing sessions with system creators. Read more about Debugging and RCA in this interesting piece.
Lastly, we will take a look at some of the trending scoops that hold a special mention for the community.
Notable FYIs
Developing a programming language is never easy. It is a complex process, and there are a lot of challenges associated with it. Here is a well-explained article from the creator of the Inko programming language, Yorick Peterse, detailing the do’s and don’ts of creating a programming language.
You can serve videos with the help of features that are used for serving content based on user preferences and accessibility features. With assorted media queries, you can present alternate versions of a video. This article from Adrian Roselli details assorted media queries on HTML by embedding the video with its alternate version and showing all the alternate versions.
The art of keeping emails hidden away from spam bots but accessible to readers is called email obfuscation. There are different ways to do that. Here is a detailed article from Spencer Mortensen where he explained the different techniques of obfuscation and showed live examples to test the effectiveness of the techniques in blocking spam.
Every cloud provider has faced at least one regional outage. In recent months, the big three of the cloud-providing field i.e. GCP, AWS, and Azure had all faced regional outages. Get a detailed analysis of the causes and the tackling of the three outages through this comparison article from The Pragmatic Engineer by Gergely Orosz.
It's always exciting for devs of any level to start a new project and this adds extra enthusiasm to those who are a newbie in tech. So, if you are thinking of starting a new project, this article from Firecamp will give you a glimpse of 11 free and fun APIs that you can use in your new gig.
That’s it from us with this edition. We hope you are going away with a ton of new information. Lastly, share this newsletter with your colleagues and pals if you find it valuable. A subscription or like to the newsletter will be awesome if you are reading for the first time.