The scorecard at the end of my raggio post kept bugging me. raggio won on memory, recall and vector tail latency, but Weaviate took most of the rest, and not by …
Performance
- Home /
- Categories /
- Performance
Performance is the thread through almost everything I write. The SQL Server posts measure wait stats, parallelism and loading strategies; the raggio series applies the same habit to a RAG vector database. Every claim comes with a number.
Sooner or later every RAG system sends you the same bill, and it’s written in gigabytes of RAM. The model wants memory for its weights and its KV cache. …
UPDATE: This is now somewhat deprecated, because you can install the Book as an extension, without having to type anything, as detailed here. Good news …
Ahh, functions, the greatest tool at a programmer’s disposal, they make code reusable and easy to read, they’ve been essential since the first …
Oh boy, this is a spinoff of my previous post on “How To Use Grafana (On Docker) To Monitor Your Sql Server (Eventually On Docker Too) – Feat. Influxdb …
Note: I'll be focusing on Azure Cloud and SQL Server, but these considerations are valid for any Cloud/DB Vendor. Whenever people talk about “The …
FYI: You can get the Notebook for this article on my github and experiment yourself (opens with Azure Data Studio). Preamble Everybody knows that using loops in …
A conversation had me wondering about performance implications in SQL Server RLS feature, specifically on how to mitigate the performance impact; TL; DR; ? …
Columstore indexes are a “new” neat data structure that I like, even if technically they’ve been around for years, only recently they’ve …
Are you one of the people that says “we have a CPU related issue” whenever it sees SOS_SCHEDULER_YIELD popping up? Let me explain why you’re …
A thing that can happen once in a while in a DW is the need to massively update a column in a table, let’s find out. I’ll be using the same table as …
Most of the support request I get involved with can be summarized with the following keywords: “slow” + “datawarehouse” + …
I was reading Paul White’s post about bitmaps in SQL Server and, other than realizing that I basically know nothing about anything (like a Socrate’s …
As any other *Server, SQL Server is born with concurrency in mind, a lot of users/applications can connect concurrently and do their job, as designed, perfect, …
I work with a vendor application which needs to do all sorts of data crunching in some nightly batches, and, depending on the size of data, this process may …