Cloudian
Hardware GuideHelp
Back to Quick Sizer

Help & Documentation

Learn how to use Quick Sizer and understand the sizing calculations.

Open Quick Sizer

Getting Started

Quick Sizer calculates the net writable storage available in a Cloudian HyperStore cluster. Enter your hardware configuration and choose a data protection policy — the results update instantly as you type.

The result shown is the capacity HyperStore will let you actually write to, after accounting for filesystem formatting, your chosen protection policy, and the built-in stop-write safety threshold.

No login is required. When you are happy with your configuration, use the Save & Get Shareable Link button to generate a permanent URL you can share with your team.

Step-by-Step Guide

1Hardware

Enter the physical specifications of your cluster:

  • Total nodes — the total number of storage nodes across all data centers.
  • Drives / node — how many data drives are in each node.
  • Drive size — the capacity of a single drive, in TB or TiB (see TB vs TiB).

The raw capacity hint below the fields updates live so you can verify the total before applying any policy overhead.

2Data Protection

Choose how many data centers participate, then select a protection policy.

  • Data centers — click 1, 2, or 3. The policy list below filters automatically to only show policies valid for that topology.
  • Protection policy — also filters to only show policies your node count can actually support. Policies that require more nodes than you have entered are hidden.
If a policy you expect to see is missing, increase the node count in Step 1. Each policy has a minimum node requirement.
3Policy Configuration

Fine-tune the protection parameters:

  • Replication policies — set the number of full copies per data center.
  • Erasure Coding policies — pick a preset k+m scheme from the dropdown, or choose Custom k+m… to enter your own values. The dropdown only shows schemes that fit your node count.

Badges below the selector show the selected scheme's fragment count and storage efficiency at a glance.

4Notes & Saving

Optionally add notes — customer name, planned expansion date, assumptions — then click Save & Get Shareable Link. A dialog will appear with a unique URL, a 6-digit share code (for read-only access), and a 6-digit edit code (to update later). See Saving & Sharing for details.

How Capacity Is Calculated

Every result is produced by the same four-step formula applied to your raw hardware numbers:

Raw Capacity

Nodes × Drives/node × Drive size

−3.13%filesystem overhead (formatting & reserved blocks)

After Filesystem

Raw × 96.87%

× RF/EC efficiency (the fraction of capacity devoted to your data vs. protection overhead)

After RF/EC

After filesystem × policy efficiency

× 90% stop-write threshold (HyperStore stops accepting writes at 90% disk fill to protect performance)

Net Usable (Writable)

After RF/EC × 90% — this is the number shown in the results panel

Worked Example

9 nodes · 12 drives/node · 16 TB drives · EC Distributed 3DC (5+4):

StepOperationResult
Raw9 × 12 × 16 TB → TiB1,571.6 TiB
After filesystem× (1 − 3.13%)1,522.4 TiB
After EC (5+4)× 5/9 = 55.6% efficiency845.8 TiB
Net usable× 90% stop-write761.2 TiB

Protection Policies

Cloudian HyperStore supports five protection policies across one to three data centers. Policies are filtered automatically to only show options compatible with your selected DC count and node count.

Replication — 1 Data Center1 DCMin 3 nodes (RF3)

HyperStore stores N full copies of every object, spread across N different nodes within a single data center.

Efficiency

1/N (e.g. 33% for RF3)

Fault tolerance

Up to N−1 node failures

Best for

Simplicity, low-latency reads

Replication — 2 Data Centers2 DCsMin 3 nodes (RF3)

Copies are split across two sites. You specify how many copies go in DC1 and how many in DC2. The cluster can survive the complete loss of either site as long as copies remain in the other.

Efficiency

1/(DC1 + DC2 copies)

Fault tolerance

Full DC loss (if ≥1 copy in other DC)

Best for

Simple DR, 2-site setups

Erasure Coding — 1 Data Center1 DCMin k+m nodes

An object is split into k data fragments plus m parity fragments, stored across k+m different nodes. Any k fragments can reconstruct the full object. More efficient than replication for large datasets.

Efficiency

k/(k+m) — e.g. 67% for 4+2

Fault tolerance

Any m node failures

Best for

Large capacity, single site

EC Replicated — 2 Data Centers2 DCsMin 2×(k+m) nodes

A complete k+m stripe is independently maintained in each data center — the full EC scheme runs in DC1, and an identical copy runs in DC2. This gives full DC-level redundancy at the cost of doubling the storage overhead compared to single-site EC.

Efficiency

k / (2×(k+m))

Fault tolerance

Full DC loss + m node failures

Best for

Mission-critical, 2-site EC

EC Distributed — 3 Data Centers3 DCsMin 6 nodes

The k+m fragments are spread evenly across three sites — no duplication. Losing an entire DC means losing roughly ⅓ of fragments, which stays below the parity count m, so data remains readable. This gives the same efficiency as single-site EC while tolerating a full DC outage.

Efficiency

k/(k+m) — same as 1DC EC

Fault tolerance

Full DC loss

Best for

Geo-distributed, efficient DR

Only the 5+4 and 7+5 schemes are supported for EC Distributed 3DC. These are the only standard Cloudian presets where (k+m) is divisible by 3 and losing one DC's share of fragments does not exceed the parity count.

Quick Comparison

PolicyDCsEfficiencyTolerates
Replication11/NN−1 node failures
Replication21/(c₁+c₂)Full DC loss
EC1k/(k+m)Any m node failures
EC Replicated2k/(2(k+m))Full DC + m node failures
EC Distributed3k/(k+m)Full DC loss

Erasure Coding Schemes

An EC scheme is written as k+m, where:

  • k = number of data fragments (the object is split into k pieces)
  • m = number of parity fragments (m extra pieces computed from the data for redundancy)
  • Together k+m fragments are stored, one per node. Any k of them are sufficient to reconstruct the full object.
  • Storage efficiency = k / (k+m). Higher k relative to m means better efficiency but less fault tolerance.
Cloudian HyperStore requires a minimum of k ≥ 3 and m ≥ 2. The smallest supported scheme is 3+2 (60% efficiency, tolerates 2 node failures).

Preset Schemes

SchemeEfficiencyMin nodesFault tolerance3DC distributed
3+260%52 nodes—
4+267%62 nodes—
5+271%72 nodes—
5+456%94 nodes✓ Yes
6+275%82 nodes—
6+367%93 nodes—
7+558%125 nodes✓ Yes
8+280%102 nodes—
8+467%124 nodes—
9+375%123 nodes—
10+283%122 nodes—
10+471%144 nodes—
12+475%164 nodes—

Rows highlighted in green are valid for EC Distributed 3DC. Custom k+m values can be entered in the calculator for advanced use cases.

Saving & Sharing

Saving a configuration generates a permanent link plus two 6-digit codes. These codes are shown once only — copy them before closing the dialog.

Shareable Link
A unique URL for this configuration. Opening it asks for a 6-digit code — share this URL with anyone you want to give access to.
Share Code (view)
A 6-digit read-only code. Recipients who enter this code can view the configuration and results, but cannot edit it.
Edit Code (private)
A 6-digit edit code. Keep this private — anyone with this code can update the configuration in place at the same URL.
Save your codes. Once you close the dialog, the share code and edit code cannot be retrieved. If you lose the edit code, you will need to create a new configuration.

Accessing a Saved Configuration

Anyone who opens the link is prompted for a 6-digit code. Entering the share code shows a read-only view. Entering the edit code opens the full form pre-filled with the saved values — changes save in place to the same link.

TB vs TiB

Drive manufacturers rate capacity in TB (terabytes), while operating systems often report capacity in TiB (tebibytes). They are not the same:

UnitDefinitionBytes
1 TB1012 — decimal, as labeled by drive vendors1,000,000,000,000
1 TiB240 — binary, as reported by most OSes1,099,511,627,776

1 TB ≈ 0.909 TiB. A "16 TB" drive actually holds about 14.55 TiB as reported by Linux or Windows.

In Quick Sizer, the Drive size unit selector sets how your input is interpreted. The TB / TiB toggle in the results panel independently controls how the output numbers are displayed — you can enter drives in TB and view results in TiB, or vice versa.
Open Quick Sizer