Wiktor Krzyczkowski

New release: Oxla 1.68 is out! This update brings a few minor additions and plenty of bug fixes. Find out all the details in the changelog 🔗📝...więcej...więcej...więcej...więcej...więcej...więcej

In its first pricing update since launch, Power BI has increased prices by 40% for Pro and 20% for PPU (Premium Per User) licenses. For organizations deeply invested in the MS ecosystem, this may be a good moment to revisit certain assumptions. While the value of cloud Power BI lies in its tight integration with MS365, the underlying data infrastructure can be swapped out for performance and cost-efficiency gains. At Oxla, we recently ran performance tests for a potential customer comparing our self-hosted data warehouse with a number of solutions, including SQL Server Analysis Services. The results: ⤷ Oxla's geometric average query runtime was 5× faster than SSAS ⤷ Query execution was 7.5× more cost-efficient on Oxla compared to SSAS on the same infra ⤷ SSAS storage was 77× more expensive than Oxla I guess that means replacing SSAS with Oxla could actually lower total costs, even with Power BI’s price increase. Especially for teams that require on-prem or hybrid deployments, Oxla can be a robust DWH that goes beyond traditional SQL Server constraints like rigid cube-based modeling or slow processing of large dimension tables. DM me if you’re looking to modernize your analytics stack without the expense of a cloud warehouse....więcej...więcej...więcej...więcej...więcej...więcej

So you’d like to self-host a modern open-source OLAP engine on-prem. That’s the beauty of OSS, right? Well, it’s possible, and plenty of teams are doing just that. But if you’re building an on-prem data warehouse, you’ll have to wrestle with implementation and maintenance complexity. That’s because these systems are cloud-native by design, built for the scale required to power fully-managed cloud services for thousands of customers. And with that comes cloud infra complexity baggage. But there’s another way. Instead of struggling with cloud-first architectures, on-prem data engineers and infrastructure teams need a modern, ultra-efficient OLAP database that’s easy to deploy and maintain. That’s exactly what we built at Oxla. In a recent customer POC, the infra engineer expected the setup to take about two weeks. Imagine his surprise when he deployed Oxla in just 17 minutes on a call with our support engineer. If you need to modernize your data warehouse but are skeptical about cloud migration, let’s talk!...więcej...więcej...więcej...więcej...więcej...więcej

Recently, I posted about how cloud data warehouse cost optimization is just damage control for an overpriced model. Today, I’d like to share this great writeup by Quesma’s Antoni Olendzki exploring database pricing trends and the various gotchas in cloud DB costs. Highly recommend reading the full article (link below) but here are Antoni’s key takeaways: ⤷ 𝐄𝐱𝐩𝐞𝐜𝐭 𝐩𝐫𝐢𝐜𝐞 𝐡𝐢𝐤𝐞𝐬 𝐚𝐬 𝐯𝐞𝐧𝐝𝐨𝐫𝐬 𝐦𝐚𝐭𝐮𝐫𝐞: What’s cheap today may not be tomorrow. ⤷ 𝐀𝐮𝐭𝐨𝐬𝐜𝐚𝐥𝐢𝐧𝐠 𝐩𝐫𝐢𝐜𝐢𝐧𝐠 𝐢𝐬 𝐩𝐨𝐰𝐞𝐫𝐟𝐮𝐥 𝐛𝐮𝐭 𝐝𝐚𝐧𝐠𝐞𝐫𝐨𝐮𝐬: Set safeguards to avoid surprise bills. ⤷ 𝐂𝐨𝐦𝐩𝐮𝐭𝐞 𝐢𝐬 𝐭𝐡𝐞 𝐫𝐞𝐚𝐥 𝐜𝐨𝐬𝐭: Storage is cheap, but processing data is where vendors make their money. ⤷ 𝐃𝐚𝐭𝐚 𝐭𝐫𝐚𝐧𝐬𝐟𝐞𝐫 𝐟𝐞𝐞𝐬 𝐚𝐝𝐝 𝐮𝐩 𝐪𝐮𝐢𝐜𝐤𝐥𝐲: Minimize unnecessary data movement to keep costs under control. ⤷ 𝐏𝐢𝐜𝐤 𝐭𝐡𝐞 𝐫𝐢𝐠𝐡𝐭 𝐯𝐞𝐧𝐝𝐨𝐫 𝐟𝐨𝐫 𝐲𝐨𝐮𝐫 𝐰𝐨𝐫𝐤𝐥𝐨𝐚𝐝: Instead of throwing away data, optimize how and where you store and process it. Couldn’t agree more, but I’d add there’s still another approach: self-hosting on-prem with per-core pricing. For analytics, refurbished servers in a local data center (co-located if needed) running the right software can get you far without surprise cloud bills. Of course, you’ll need the right personnel to pull this off, but at a certain scale and for certain workloads, I’d argue it’s worth the hassle compared to cloud gymnastics. Probably the biggest dilemma here is how predictable your workload is. Auto-scaling is powerful indeed, but what if overprovisioning is a viable alternative? It’s true performance has been fetishized by vendor database engineers, while data engineers are more concerned with quality of life aspects. Still, I believe there’s room to improve OLAP efficiency AND ease of use, so that: ⤷ Self-hosted data warehouses don’t need an exuberant server fleet. ⤷ It’s easy to deploy and maintain them on-prem. ⤷ As a result, overprovisioning could be more optimal than auto-scaling for both certain “spiky” workloads and unconstricted exploratory analysis. Am I crazy to suggest this? What do you think?...więcej...więcej...więcej...więcej...więcej...więcej

𝐉𝐮𝐬𝐭 𝐬𝐡𝐢𝐩𝐩𝐞𝐝 𝐚 𝐦𝐚𝐣𝐨𝐫 𝐫𝐞𝐥𝐞𝐚𝐬𝐞: 𝐎𝐱𝐥𝐚 𝟏.𝟔𝟔! 🟢 This update brings a ton of features our customers asked for: ⤷ Support for 𝙼𝙸𝙽, 𝙼𝙰𝚇, 𝙲𝚄𝙼𝙴_𝙳𝙸𝚂𝚃, 𝙽𝚃𝙸𝙻𝙴, 𝙱𝙾𝙾𝙻_𝙾𝚁, and 𝙱𝙾𝙾𝙻_𝙰𝙽𝙳 (in 𝚁𝙾𝚆𝚂 mode) 𝐰𝐢𝐧𝐝𝐨𝐰 𝐟𝐮𝐧𝐜𝐭𝐢𝐨𝐧𝐬 ⤷ Support for 𝚁𝙰𝙽𝙶𝙴 mode in 𝐰𝐢𝐧𝐝𝐨𝐰 𝐟𝐮𝐧𝐜𝐭𝐢𝐨𝐧𝐬 ⤷ Support for 𝙼𝙾𝙳𝙴, 𝙿𝙴𝚁𝙲𝙴𝙽𝚃𝙸𝙻𝙴_𝙲𝙾𝙽𝚃, and 𝙿𝙴𝚁𝙲𝙴𝙽𝚃𝙸𝙻𝙴_𝙳𝙸𝚂𝙲 𝐨𝐫𝐝𝐞𝐫𝐞𝐝-𝐬𝐞𝐭 𝐚𝐠𝐠𝐫𝐞𝐠𝐚𝐭𝐞 𝐟𝐮𝐧𝐜𝐭𝐢𝐨𝐧𝐬 ⤷ Support for 𝐬𝐭𝐚𝐭𝐢𝐬𝐭𝐢𝐜𝐚𝐥 𝐚𝐠𝐠𝐫𝐞𝐠𝐚𝐭𝐢𝐨𝐧 𝐟𝐮𝐧𝐜𝐭𝐢𝐨𝐧𝐬: 𝙲𝙾𝚁𝚁, 𝙲𝙾𝚅𝙰𝚁_𝙿𝙾𝙿, 𝙲𝙾𝚅𝙰𝚁_𝚂𝙰𝙼𝙿, 𝚁𝙴𝙶𝚁_𝙰𝚅𝙶𝚇, 𝚁𝙴𝙶𝚁_𝙰𝚅𝙶𝚈, 𝚁𝙴𝙶𝚁_𝙲𝙾𝚄𝙽𝚃, 𝚁𝙴𝙶𝚁_𝙸𝙽𝚃𝙴𝚁𝙲𝙴𝙿𝚃, 𝚁𝙴𝙶𝚁_𝚁𝟸, 𝚁𝙴𝙶𝚁_𝚂𝙻𝙾𝙿𝙴, 𝚁𝙴𝙶𝚁_𝚂𝚇𝚇, 𝚁𝙴𝙶𝚁_𝚂𝚇𝚈, 𝚁𝙴𝙶𝚁_𝚂𝚈𝚈, 𝚂𝚃𝙳𝙳𝙴𝚅, 𝚂𝚃𝙳𝙳𝙴𝚅_𝙿𝙾𝙿, 𝚂𝚃𝙳𝙳𝙴𝚅_𝚂𝙰𝙼𝙿, 𝚅𝙰𝚁𝙸𝙰𝙽𝙲𝙴, 𝚅𝙰𝚁_𝙿𝙾𝙿, 𝚅𝙰𝚁_𝚂𝙰𝙼𝙿 Our team also fixed quite a few bugs 😇 Check out the full changelog 🔗📝...więcej...więcej...więcej...więcej...więcej...więcej

No one ever got fired for buying IBM/Oracle/Snowflake/Databricks/choose your poison. But are the big, established platforms always the right tool for the job? Sometimes risk aversion leads to short-sighted decisions imposed by business leaders, against the better judgment of technical stakeholders. And in 2025, there are too many vendors to choose from to default to the obvious ones. Unless you’d like to pay per feature. Pay more for support every year. Never get the features you request. Get more features you don’t need. Pay high margins on top of overpriced compute. _________________________________________ Or you can choose to work with next-generation solutions. With fresh codebases and novel approaches to widely accepted problems. With more engineers than salespeople. Who swiftly ship the features you ask for because they’re eager to serve you as a paying customer. Because they’re focused on helping you solve your specific problems. Rather than chasing the biggest possible market to justify big tech exuberance....więcej...więcej...więcej...więcej...więcej...więcej

Why is Oxla so efficient at handling complex queries over terabyte-scale datasets? Looking under the hood, there are quite a few innovations optimizing data transfer between the CPU and RAM to achieve better resource utilization compared to other modern OLAP solutions. One such innovation, multimap query optimization, has been recently patented. This boils down to a novel approach to using hash tables. Oxla optimizes query performance by partially sorting records and efficiently storing key-value mappings. By leveraging a multimap structure with contiguous memory storage, this approach enhances cache efficiency and reduces lookup overhead. Additionally, when a key has only a single associated value, Oxla optimizes query execution by storing data values directly instead of using pointers, reducing lookup overhead. For data engineers, this means faster queries, lower infrastructure costs, and a more efficient way to handle datasets at terabyte scale and beyond. Swipe through the PDF for more geeky details 🤓...więcej...więcej...więcej...więcej...więcej...więcej

Oxla Patent No. 2 📄 We recently obtained the second patent for our OLAP DBMS innovations! This patent covers a novel approach to using hash tables for processing queries: ⤷ Oxla optimizes query performance by partially sorting records and efficiently storing key-value mappings. ⤷ By leveraging a multimap structure with contiguous memory storage, this approach enhances cache efficiency and reduces lookup overhead. ⤷ Additionally, when a key has only a single associated value, Oxla optimizes query execution by storing data values directly instead of using pointers, reducing lookup overhead. For data engineers, this means faster queries, lower infrastructure costs, and a more efficient way to handle datasets at terabyte scale and beyond....więcej...więcej...więcej...więcej...więcej...więcej

What if your favorite open-source database suddenly changed its license? Everyone loves open source. So much so that, for years, database vendors have defaulted to open-sourcing their work as a go-to-market play, monetizing through the cloud. But open-source licensing comes with a risk: someone else might build a competing cloud service on their technology. That’s why we’re seeing a trend of databases moving to source-available licensing: ⤷ MongoDB was one of the first to do this in 2018 when it switched to the Server Side Public License (SSPL). ⤷ That same year, Timescale made a similar move, switching to the source-available Timescale License (TSL). ⤷ Last year, Redis moved from a BSD-3 license to a dual-license model, adopting both the Redis Source Available License (RSAL) and MongoDB’s SSPL to restrict how cloud providers can monetize its software. ⤷ Also last year, ScyllaDB announced it would no longer maintain its fully open-source version and shifted its enterprise offering from closed-source to source-available. While a free version still exists, it now comes with additional restrictions. In general, source-available licenses restrict free use to prevent public cloud vendors like AWS from spinning up cloud services based on the technology. To be honest, I find it incredibly ironic that the same public cloud providers enabling OSS monetization also pose a threat to this business model. While I greatly admire open-source contributors, I don’t think it’s fair that their labor often goes uncompensated, and I’m skeptical about the long-term viability of business models based on open source. What do you think?...więcej...więcej...więcej...więcej...więcej...więcej