Oxla

If you are using analytical databases, you are probably familiar with ClickBench: a benchmark created by ClickHouse based on data from Yandex. Unlike some other benchmarks, ClickBench's dataset does not contain generated data, making it more realistic by reflecting the most commonly occurring distributions: normal and exponential. However, its query set mainly focuses on aggregations over entire tables or over one or more columns of different types and cardinalities, with some filtering and occasionally selecting top results based on a metric (ORDER BY + LIMIT). While these queries are representative of real-time databases used for serving metrics, they do not include even a single query involving JOINs or subqueries. Data scientists or report generation tasks typically perform several JOIN operations, whereas GROUP BY operations usually run much faster than JOINs. Therefore we have concentrated so far on achieving amazing JOIN performance while also maintaining good GROUP BY performance. This changed in January. At the end of January, we noticed that a benchmarking result was submitted to ClickBench, and the results were not what we wanted to present to our users... Firstly, the person submitting the result had not created an index like the one used in the ClickHouse submission (in the form of order-by columns) or other submissions. Secondly, the data size was completely off. Most importantly, we had not yet started optimizing our query engine and planner to improve our performance on this kind of workload. Seeing those results on ClickBench was a great motivation for us to start working on it. After several quick improvements, we sent a new submission at the beginning of April using the newest publicly available version of Oxla at that time. This was approved and published last week. But this is not the end of the story: we began researching a new implementation of GROUP BY at the end of last year. It was not prioritized until now. In the picture below, you can see the initial results. The leftmost column represents ClickHouse, followed by the development version of Oxla, then the result of Oxla from April, and finally the Oxla version from January. As you can see, we are making rapid progress. Hopefully, we will be able to publish a new public version with this performance by the end of this month.

🌍 Reducing Greenhouse Emissions with Data Processing Efficiency 🌍 Data centers are the backbone of our digital world, but they come with a significant environmental cost. Every query processed in these centers consumes energy, leading to CO2 emissions that contribute to global warming. As businesses increasingly rely on data-driven decisions, it's crucial to consider the environmental footprint of our data operations. 🔍 Oxla vs. The Leading OLAP database ❄ : A Carbon Comparison 🔍 Here's a straightforward comparison of the CO2 emissions produced by the average query load times of the leading OLAP solution ❄ and Oxla: The leading solution ❄: With a sample query set, takes 62 seconds producing approximately 4.23 grams of CO2. Oxla: With the same query set, is optimized to run in just 12 seconds, producing only 0.82 grams of CO2. That's a reduction of 3.41 grams of CO2 per query. 💡 A Simple Step Towards Carbon Neutrality 💡 With the EU's Corporate Sustainability Reporting Directive (CSRD) and Energy Efficiency Directive (EED) propelling the shift towards greener computing, it's time to act. By switching to Oxla, companies can significantly reduce their carbon footprint. This straightforward change helps businesses make strides towards their carbon-neutral goals with minimal effort and a lot of added benefits. Choosing efficient data processing isn't just about saving time and money; it's about taking responsibility for our planet. Let's make every query count. #Databases #OLAP #GreenBusiness #GreenComputing