Fast Compilation or Fast Execution: Just Have Both!

April 2, 2025

Learn the basics of code generation, which is one of the secrets behind CedarDB's performance. CedarDB creates custom machine code for every query. This keeps data in CPU registers as long as possible and minimizes unnecessary data transfers.

To B or not to B: B-Trees with Optimistic Lock Coupling

March 6, 2025

B-Trees stand the test of time. In this article, we explore why we still use a 55 year old data structure: It is still super efficient on modern hardware when we use contention free optimistic lock coupling!

Why Trees Without Branches Grow Faster: The Case for Reducing Branches in Code

January 29, 2025

In some of our blog posts, we explained what steps we take to reduce the number of branching instructions in our critical paths. However, we only ever claimed that this is much better and faster, and always omitted explaining why. So today we will fix this and take a deep dive into the gritty details of instruction execution on modern CPUs.

Helping Christmas Elves Count Presents (or: Vectorized Overflow Checking)

December 24, 2024

In a previous post, we explained the importance of overflow checks when summing numbers, and mentioned that the usual approaches are not easily vectorized. Read here how to get 4x the performance when adding integers by using specialized vector instructions.

The History of the Decline and Fall of In-Memory Database Systems

December 3, 2024

A decade ago, there was a sudden surge of high-performance in-memory systems dominating the world of interactive analytics. Today, almost everyone has gone back to using persistent storage. Does this mean that building these in-memory systems was a mistake?

Offset Considered Harmful or: The Surprising Complexity of Pagination in SQL

November 19, 2024

Have you ever wondered why you sometimes see duplicate results when clicking on the second page of a website? In this blog post, we explore techniques for result pagination, how they impact the work necessary to compute the results, and why using the SQL offset keyword for it is not a good idea.

How to Correctly Sum Up Numbers

October 30, 2024

When you learn programming, one of the first things every book and course teaches is how to add two numbers. So, developers working with large data probably don't have to think too much about adding numbers, right? Turns out it's not quite so simple!

Why You Shouldn't Forget to Optimize the Data Layout

October 8, 2024

The underlying data layout of your program can either help or hurt your algorithms. We explore how to improve the runtime of your program by analyzing the impact of the data layout on low-level properties such as cache line optimizations and SIMD execution, as well as other memory-level optimizations such as compression.

The Hidden Cost of Data Movement

September 24, 2024

Moving data between disk, memory, caches, and CPU registers is one of the most critical paths when processing large amounts of data. In this post, we explore the often-overlooked costs of these data movements and show how they can be reduced for algorithms in general and specifically in a database system.

Why I Prefer Exceptions to Error Values

September 10, 2024

Exceptions are often a better way to handle errors than returning them as values. We argue that traditional exceptions provide better user and developer experience, and show that they even result in faster execution.