(estimate_azure_costs)=

# Estimate Azure costs

:::{warning}
Technical support for Turing TREs represents a fixed cost across all projects.
Currently, these costs are subsidised by the Turing, but this may change in future.
:::

<!--
- for short-running projects: **£3,380** as a one-off charge when the project is set up (which includes a 2 week "sandbox")
- for long-running projects: **£550** for each month of operation
-->

## Azure infrastructure

The basic infrastructure of our TRE contains the following:

- Remote desktop gateway
- Networking and routing components
- Secure file storage
- Gitea server for version control
- HedgeDoc server for collaborative document writing

The cost of this infrastructure is around **£1,100 per month**.
On top of this, you will also need some virtual machines to analyse the secure data which will cost **£250 per month or more**.

We have guidance on {ref}`how to estimate Azure costs <estimate_azure_costs>`, which includes some recommended setups for common project scenarios:

- No special requirements and {ref}`a small project team <estimate_azure_costs_small_project>`.
- Needing a PostgreSQL or Microsoft SQL {ref}`database server <estimate_azure_costs_database>`.
- Needing {ref}`extra CPU processing power <estimate_azure_costs_more_cpu>`.
- Needing {ref}`higher amounts of memory <estimate_azure_costs_more_ram>`.
- Needing {ref}`to use GPUs <estimate_azure_costs_more_gpu>`.
- Having a {ref}`large number of users <estimate_azure_costs_more_users>`.
- Needing to work with a {ref}`large amount of input data <estimate_azure_costs_more_storage>`.

## Cost examples

Some example projects with associated costings are included below.
You may want to use one of these as a template for your own project costing.

:::{note}
Our recommended setup is to use one `Standard_D8s_v5` virtual machine (8 CPUs and 32 GB RAM).
There is a list of available virtual machine sizes and costs [here](https://cloudprice.net/?currency=GBP&timeoption=month&sortField=name).
:::

Please [get in touch with us](mailto:trustedresearch@turing.ac.uk) if you have any questions.

(estimate_azure_costs_small_project)=

## Small project

For a project with no special requirements and a small project team, the following should be sufficient.

| Item               | Cost (£ / month) |
| ------------------ | ---------------: |
| SRE infrastructure |             1100 |
| Standard_D8s_v5 VM |              250 |
| **Total**          |             1350 |

(estimate_azure_costs_database)=

## Project needing a database

For a project requiring a PostgreSQL or SQL database, the following should be factored in.

| Item                       | Cost (£ / month) |
| -------------------------- | ---------------: |
| SRE infrastructure         |             1100 |
| Standard_D8s_v5 VM         |              250 |
| PostgreSQL or SQL database |               50 |
| **Total**                  |             1400 |

Note database costs can vary.

(estimate_azure_costs_more_cpu)=

## Project needing more processing power

For a project that needs to do a lot of parallel or batch processing, the standard virtual machine will be too small.
See [here](https://learn.microsoft.com/en-us/azure/virtual-machines/sizes-compute) to read more about compute-optimised virtual machine sizes.
Using, for example, a `Standard_F16s` (16 CPUs and 32 GB RAM) will give the following costs:

| Item               | Cost (£ / month) |
| ------------------ | ---------------: |
| SRE infrastructure |             1100 |
| Standard_F16s VM   |              530 |
| **Total**          |             1040 |

(estimate_azure_costs_more_ram)=

## Project needing more memory

For a project that needs to do in-memory caching or manage large datasets in memory, the standard virtual machine will be too small.
See [here](https://learn.microsoft.com/en-us/azure/virtual-machines/sizes-memory) to read more about memory-optimised virtual machine sizes.
Using, for example, a `Standard_E8s_v5` (8 CPUs and 64 GB RAM) will give the following costs:

| Item               | Cost (£ / month) |
| ------------------ | ---------------: |
| SRE infrastructure |             1100 |
| Standard_E8s VM    |              350 |
| **Total**          |             1450 |

(estimate_azure_costs_more_gpu)=

## Project needing a GPU

For a project that wants to do computer vision or deep learning, a GPU might be needed.
See [here](https://learn.microsoft.com/en-us/azure/virtual-machines/sizes-gpu) to read more about GPU-optimised virtual machine sizes.
Using, for example, a `Standard_NC8as_T4_v3` (8 CPUs, 1 GPU and 56 GB RAM) will give the following costs:

| Item               | Cost (£ / month) |
| ------------------ | ---------------: |
| SRE infrastructure |             1100 |
| Standard_NC8as VM  |              520 |
| **Total**          |             1620 |

(estimate_azure_costs_more_users)=

## Project with many participants

For a project with many participants it may be convenient to deploy several different virtual machines.
For example, a project with ten people that wants to share each virtual machine between two people would have the following costs:

| Item               | Cost (£ / month) |
| ------------------ | ---------------: |
| SRE infrastructure |             1100 |
| Standard_D8s_v5 VM |              250 |
| Standard_D8s_v5 VM |              250 |
| Standard_D8s_v5 VM |              250 |
| Standard_D8s_v5 VM |              250 |
| Standard_D8s_v5 VM |              250 |
| **Total**          |             2350 |

(estimate_azure_costs_more_storage)=

## Project needing lots of data storage

For a project that deals with large datasets, the standard 1 TB of data storage might be insufficient.
Using, for example, 10 TB of storage instead will have the following costs (using a typical cost of £148 per month for each TB of storage - based on [Azure Storage Pricing Calculators](https://azure.microsoft.com/en-gb/pricing/details/storage/blobs/)).

| Item                       | Cost (£ / month) |
| -------------------------- | ---------------: |
| SRE infrastructure         |             1100 |
| Additional 9 TB of storage |             1330 |
| **Total**                  |             2430 |