Spaces Direct Calculator PREVIEW

Choose Your Drives

Indicate what drives will be present in each node, excluding boot devices.


This results in 0 TB of physical capacity and 0 TB of cache per node. ?

If devices of multiple types (NVMe, SSD, or HDD) are present, the fastest are automatically used as cache. Each faster device binds dynamically to several slower devices, to accelerate I/O for consistent low latency. These devices do not contribute to the overall capacity, since everything stored in cache is also stored elsewhere.

You have chosen devices of 0 type(s). There will be no Software Storage Bus cache.

Choose Your Fault Tolerance

How many simultaneous hardware failures do you need to tolerate? ?

This includes drives, nodes, and their components such as network adapters.

Over its lifetime, Spaces Direct can tolerate any number of failures, because it restores to full resiliency after each one.

However, as with standard RAID, there are limits to how many failures can be tolerated at any one time. Choose 2 to ensure that all data remains safe and continuously accessible in the unlikely event that 2 drives fail simultaneously, or that 2 nodes go down simultaneously, or one 1 drive fails and 1 node goes down. Think distributed RAID-1 ("3-Way") or RAID-6.

Choose 1 to protect against any one failure at once. Think distributed RAID-1 ("2-Way") or RAID-5.

Do you want your deployment to be chassis or rack fault tolerant?

Choose Your Scale

Indicate the number of nodes in your cluster. ?

The minimum and maximum limits are based on your fault tolerance requirements, and what Microsoft supports.

Based on your chosen fault tolerance, the following deployments are possible. Choose one.

Choose Your Resiliency Mix

Choose which fraction of your data you'd like to store using Mirror, for faster access. ?

You can achieve resiliency using Mirroring, Parity Encoding, or both together. Mirroring provides faster access to data, but is not very space efficient. Parity Encoding incurs some computation time and cost to encode/decode data, but is far more space efficient.

We recommend using both in some proportion.

The proportions of your tiers are fixed by the capacity of your SSD and HDD devices.

To use Parity Encoding, you must have at least 4+ fault domains. Consider scaling your cluster. ?

Parity Encoding stripes data and parity symbols across fault domains to densely and efficiently store infrequently accessed data.

At large enough scale, it can achieve 80.0% efficiency without compromising fault tolerance. (Compare with just 33.3% efficiency for 3-Way Mirroring.) It requires at least 4 fault domains. Consider scaling your cluster to unlock Parity Encoding.





This is the usable capacity for each site-scoped pool. ?

In multi-site Stretch Clusters, separate storage pools are created for each site. One pool cannot span sites.

The figures provided here are for each site. We recommend using Storage Replica to synchronously replicate one site to the other. In this scenario, you achieve site fault tolerance for Disaster Recovery, but the second site does not ultimately contribute additional usable capacity.

Storage Efficiency: ?

Mirroring has efficiency. Using only Mirroring gives available capacity.

At this scale, Parity Encoding has efficiency. Using only Parity Encoding gives available capacity.

Combining these in the proportions you've chosen gives overall.

Excludes reserve.

Usable Resiliency Reserve

Reserve: ?

This deployment has total physical capacity, excluding cache.

We recommend setting aside for in-place recovery. This enables Spaces Direct to restore resiliency immediately after drives fail, even before they are replaced. This value is based on your chosen fault tolerance of and your largest drives, which are .

At certain scales, we may recommend more because repairs are constrained to happen on the local node. This can be non-obvious, but trust us -- it's true. The value above includes this consideration. This recommended reserve is not enforced; if you don't have enough, repairs simply have to wait for the replacement drive to complete.

This leaves total physical capacity after reserve.


This app is still in development. It is being tested on Chrome 51+ and Edge 25+.

For feature requests, bug reports, or general feedback email Cosmos Darwin.