Azure SQL Database: A Field Guide to the DTU, or How DBAs Learned to Stop Worrying and Just Double Everything
DBAs are good like that for your bottom line.
Let me tell you about the Database Transaction Unit (DTU), or as I like to call it, Microsoft’s attempt to make database pricing simple by inventing an entirely new unit of measurement that nobody understands. It’s calculated by throwing dice in a full moon, whilst chanting some voodoo-like utterance.
You know what was wrong with measuring databases in CPU cores, memory, and IOPS? Apparently, it was too transparent and made sense. Too easy. Someone at Microsoft had the brilliant idea to blend all those metrics into a magical number called a DTU, then tell everyone “don’t worry about what’s inside, just pick a number between 5 and 4,000.” It’s like ordering at a restaurant where the menu just says “Food Units” instead of listing ingredients. “I’ll have 50 Food Units, please.” “Sir, would you like to know if that’s chicken or—” “FIFTY. FOOD. UNITS and make it snappy my man.”
The Mythical DTU: Nobody Knows What It Is, Everyone Pays For It
A DTU is allegedly a blended measure of CPU, memory, reads, and writes. The exact ratio? Microsoft’s secret recipe, like KFC’s eleven herbs and spices, except instead of delicious chicken (depends where you go and I use the word delicious liberally in this context), you get database throttling at unexpected moments with loads of MSG. The official explanation is that DTUs represent the relative performance of a database. Excellent, that’s a great start, thanks Microsoft. The higher DTUs mean more power. Lower DTUs mean less power. It’s beautifully circular logic, a DTU is worth exactly one DTU’s worth of performance, and if you don’t like that answer, maybe you should’ve paid attention in database philosophy class. A DTU doesn’t exist, or does it?
