WTF is a DTU?! (And Other Bizarre Azure Units That Actually Make Sense… Kinda)

Share on:

Let’s be honest, using Azure for the first time is a bit like trying to learn Klingon and metric conversions at the same time. Honestly it sometimes feels like Microsoft just spins a giant acronym wheel and goes with whatever sounds vaguely techy. DTUs, RUs, PUs, SUs, FUs... no wait that is not a thing 😅

Fear not we are about to break down some of Azure’s most mysterious measuring units with real explanations and bad jokes. By the end of this post, you’ll be fluent in Azure-ese.

DTU – "Database Totally Unclear"

Actual meaning: Database Transaction Unit  

Used in: Azure SQL Database (DTU-based model)

A DTU is Microsoft’s "trust me bro" combo of CPU, memory, and I/O bundled into a single performance metric. You don’t know exactly what’s in it — like a tech smoothie — but it gets the job done.

Think of it as: A pre-mixed performance cocktail. Good luck reverse-engineering the recipe.

RU/s – "Ridiculously Unclear (but Real)"

Actual meaning: Request Units per second  

Used in: Azure Cosmos DB

This is the currency of Cosmos DB. Every read, write, or query costs a certain number of RUs. A simple 1KB document read = 1 RU. But if you try something spicy like a cross-partition query... better bring your RU wallet.

Think of it as: A tiny toll booth. Every operation pays its dues to pass.

PU – "Please Understand"

Actual meaning: Processing Unit  

Used in: Azure Event Hubs (Dedicated Tier)

When you're tired of sharing and want your own playground, you go Dedicated. PUs give you a guaranteed slice of CPU and memory to handle massive event streams.

Think of it as: Renting a private concert hall for your data — no more fighting over bandwidth with strangers.

TU – "Throughput Undefined"

Actual meaning: Throughput Unit  

Used in: Azure Event Hubs (Standard/Premium)

Each TU gives you the right to 1 MB/s ingress and 2 MB/s egress — or 1000 events/sec. Not bad, right?

Think of it as: Reserving lanes at the cloud’s bowling alley for your events.

SU – "Streaming Unknowns"

Actual meaning: Streaming Unit  

Used in: Azure Stream Analytics

SUs = the number of chefs Azure assigns to cook your real-time data meals. The more complicated your job, the more chefs (ahem, compute) you’ll need. Think of it as: Shots of espresso for your stream processing job.

CU – "Compute-ish Units"

Actual meaning: Compute Units

Used in:

  • Azure App Service Plans (web app compute)

  • Azure Front Door & Traffic Manager (routing)

  • Azure Cognitive Search (indexing/query/storage)

CU means slightly different things depending on where you are, but it usually refers to how much compute juice you’ve got.

Think of it as: That friend who changes their job title every week but still shows up and works hard.

ACU – "Azure’s Confusing Unit"

Actual meaning: Benchmark for VM performance  

Used in: Azure Virtual Machines ACUs help you compare CPU performance across VMs. Microsoft gave the Standard_A1 VM an ACU of 100. Everything else is stronger or weaker by comparison.

Think of it as: Horsepower for your VMs. More ACUs = more get-up-and-go.

vCore – "Very Core-y"

Actual meaning: Virtual Core  

Used in: Azure SQL, Synapse, and more

Finally, some sanity! 1 vCore = 1 logical CPU thread. What you see is what you get. Easy.

Think of it as: The golden retriever of Azure units. Friendly. Predictable. Always good.

DSVM – "Data Science Voodoo Machine"

Actual meaning: Data Science Virtual Machine  

Used in: Azure ML, AI, and data science projects

Think of this as your all-in-one lab — TensorFlow, Jupyter, R, Python, etc. — it’s all installed and ready to go. Think of it as: A science lab in a box. Bring your data, and get cooking.

AU – "Analytics Uhh…"

Actual meaning: Analytics Units  

Used in: Time Series Insights

AUs control how much event data you can ingest and how fast you can query it.

Think of it as: How many detective hats you can wear at once to investigate your IoT data.

MU – "Mind Units"

Actual meaning: Memory reserved in Service Fabric  

Used in: Azure Service Fabric

Each MU is essentially a slice of memory. You define how many MUs each service needs, and Azure makes sure there’s room.

Think of it as: Beds in a dorm. Assign too few, and your services are sleeping on the floor.

Page Units – "Pricey Page Turners"

Used in: Azure Page Blobs

Each Page Blob is split into 512-byte chunks, and you’re billed based on how many you provision — even if they’re empty.

Think of it as: Renting lockers and getting charged whether you use them or not.

Log Analytics Units - Log-a-lots

Used in: Azure Monitor / Log Analytics

Azure charges for how much data you ingest (GB/day) and retain. You can also reserve capacity to save 💵

Think of it as: Paying by the decibel for how loudly you scream into your logs.

Wrap-Up: Azure’s Unit-verse

Now you’ve got the lowdown on Azure’s funhouse of measuring units. Some are logical. Some are... less so. But all of them help you plan, scale, and (ideally) not go broke in the cloud.

TL;DR

Unit Used For
DTU Azure SQL performance (blended)
RU/s Cosmos DB operation costs
PU Event Hubs Dedicated compute
TU Event Hubs Standard throughput
SU Stream Analytics compute
CU Varies: App Service, Search, Front Door
ACU VM CPU benchmarks
vCore Logical CPUs
DSVM Data science-ready VMs
AU Time Series Insights capacity
MU Service Fabric memory
Page Units Blob storage billing
Log Analytics Units Azure Monitor data usage