Skip to content

Private Cloud datastores from data centres in Germany: 4 classes of shared storage used by all hosts in a cluster — connected via NFS 4.1, iSCSI or NVMe over TCP, each with an IOPS limit, latency commitment and the condition under which both apply. Billing is per provisioned gigabyte and month with a minimum commitment per class; the calculator below shows the monthly amount per class and quantity.

This page describes the shared storage of the Private Cloud. Local NVMe in the host are part of the host class and are not billed separately.

Datastore classes
4
NVMe, capacity, archive
IOPS per datastore
up to 120,000
4 K random, 70/30
Latency p99
0.4 ms
from four queues
Rate from
€0.0068$0.0079
per GB and month, net

Your datastore in three decisions

Class, quantity, snapshots — after each decision, the monthly amount is updated next to it, calculated with the same code as in the configurator. You see the minimum purchase as the difference between the requested and calculated quantity, not as a footnote.

Always included

  • Platform operationsHypervisor, BIOS and firmware, fabric switches, storage controllers
  • Redundancy in the setyou order usable capacity, not raw capacity
  • Dedicated storage fabricStorage and guest traffic on separate links
  • Backup windowfixed per class, during the night
  • no setup feefor the datastore

Class

Class

Requested quantity

€294.91$342.10net / month

512 GB2048 TB

requested 4 TBbilled 4 TBRate €0.0720$0.0835 per GB

Snapshots

The same quantity in all four classes

Below the minimum commitment, the seemingly more expensive class is the cheaper one.

The break-even points are listed under prices
Net per month€294.91$342.10Summary

Four classes, one storage network

All four classes are attached to the same storage network (the fabric) and are managed in the same vCenter or Cloud Director. What differs is the medium, protocol and the commitment we provide — not the operational path.

For running guests

Operational datastores for databases, application servers and everything that depends on write latency. Both classes are on NVMe; they differ in the protocol, the latency commitment and the condition under which it applies.

Row of drive carriers at the front of a storage enclosure; one carrier is half pulled out, the drive inside visible, a green status light on each carrier.

Datastore NVMe Performance

SAP-approved

NVMe over TCP for databases that depend on write latency.

  • NVMe/TCP
  • NFS 4.1
IOPS per datastore
up to 120,000
Latency p99
0.4 ms
Throughput
6,000 MB/s
Minimum commitment
2 TB

Redundancy: Triply replicated across two fire zones of the location

LimitThe IOPS guarantee applies per datastore, not per VM. A single guest only reaches it with at least four parallel queues; with one queue, around 28,000 IOPS remain.

€0.1090$0.1264 per GB and month

from €223.23$258.95 per month

View in calculator — Datastore NVMe Performance

Datastore NVMe Standard

The standard case: NVMe backend via NFS or iSCSI, for mixed guest workloads.

  • NFS 4.1
  • iSCSI
IOPS per datastore
up to 45,000
Latency p99
0.9 ms
Throughput
2,400 MB/s
Minimum commitment
1 TB

Redundancy: Triply replicated across two fire zones of the location

LimitFrom 24 concurrently active VMs on a datastore, the guaranteed latency drops from 0.9 ms to 1.4 ms. Larger environments are therefore split across multiple datastores.

€0.0720$0.0835 per GB and month

from €73.73$85.53 per month

View in calculator — Datastore NVMe Standard

For capacity and storage

Plenty of space at a low rate — for file services, quiet volumes, templates and exports. Both classes have lower rates, but higher minimum commitments than the NVMe classes: if you need little, you pay more here than the rate promises.

From above into the open drawer of a disk shelf: rows of upright hard drives, every slot occupied; below it a closed identical shelf.

Datastore Capacity

NVMe cache in front of HDD capacity — for file services and quiet volumes.

  • iSCSI
  • NFS 4.1
IOPS per datastore
up to 9,000
Latency p99
4.0 ms
Throughput
1,100 MB/s
Minimum commitment
8 TB

Redundancy: Erasure coding 8+2 in the location, cache triply mirrored

LimitThe NVMe cache covers 8% of the datastore size. Working sets beyond this pass through to the HDD tier; the 4 ms latency guarantee then no longer applies.

€0.0290$0.0336 per GB and month

from €237.57$275.58 per month

View in calculator — Datastore Capacity

Datastore Archive

Cold capacity for powered-off guests, templates and exports.

  • NFS 4.1
IOPS per datastore
up to 900
Latency p99
25.0 ms
Throughput
350 MB/s
Minimum commitment
20 TB

Redundancy: Erasure coding 8+2, second copy in another EU region

LimitNot approved as an operational datastore. Running VMs must not reside here — the class accommodates powered-off guests, templates and OVA exports.

€0.0068$0.0079 per GB and month

from €139.26$161.54 per month

View in calculator — Datastore Archive

The commitments apply per datastore, not per virtual machine and not per host. If you want to push a single database to the limit, you get a dedicated datastore — otherwise it shares the queue with everything else that resides on the same LUN (the logical drive behind the datastore).

Three protocols, one fabric

The fabric is the storage network that connects all hosts to the datastores. Which protocol a class speaks over it is not a matter of taste: it follows from the medium behind it and the latency we can guarantee on it.

HostHostHostFabric ASwitchFabric BSwitchDatastore
Path via Fabric APath via Fabric BWith iSCSI, two paths run via separate fabric switches, while with NFS 4.1, session trunking distributes a connection across multiple links — if one path fails, the other takes over without the guest noticing.
Rear of a storage controller in a rack: two identical modules, each with four connected fibre optic cables, routed separately to the left and right, two power supply units with connected cables.

NVMe over TCP

The drive's command set runs unchanged over the storage network — without translation to SCSI and without the single queue that iSCSI has per target. Each virtual machine gets multiple queues on the same namespace.

This is exactly what the guaranteed latency of the performance class depends on. With a single queue, the gain over iSCSI remains small; the guarantee applies from four parallel queues.

1 of 4 classes · NVMe Performance

iSCSI

Block over TCP, one target per datastore. The hypervisor creates its cluster file system on it and manages the locks itself — this is the path that older guest operating systems and cluster software follow without modification.

It is connected via two paths on separate fabric switches. If one path fails, the second takes over without the target having to log in again; a path change is not visible in the guest.

2 of 4 classes · NVMe Standard, Capacity

NFS 4.1

File instead of block: the datastore is an export, every virtual disk a file within it. Thinly provisioned without intervention, and a single guest can be backed up or copied without going through the cluster file system.

Session trunking distributes a connection across multiple lines. Locks run via the server instead of the host — which is why NFS 4.1 is the protocol that all four classes speak.

4 of 4 classes · NVMe Performance, NVMe Standard, Capacity, Archive

The upper limit is guaranteed, not the best-case figure

The values in the table are upper limits per datastore, measured with 4 KB random access in a 70 to 30 ratio of read and write. If you measure differently, you get different numbers — that is why the measurement profile is included.

Performance profile of the four datastore classes. IOPS and throughput apply per datastore and not per virtual machine. The latency is the 99th percentile (p99) over a 24-hour measurement window — 99 out of 100 measured accesses are below this.

Performance profile of the four datastore classes
ClassProtocolsIOPS, 4 K randomThroughputLatency p99Redundancy
Datastore NVMe PerformanceNVMe/TCP, NFS 4.1up to 120,0006,000 MB/s0.4 msTriply replicated across two fire zones of the location
Datastore NVMe StandardNFS 4.1, iSCSIup to 45,0002,400 MB/s0.9 msTriply replicated across two fire zones of the location
Datastore CapacityiSCSI, NFS 4.1up to 9,0001,100 MB/s4.0 msErasure coding 8+2 in the location, cache triply mirrored
Datastore ArchiveNFS 4.1up to 900350 MB/s25.0 msErasure coding 8+2, second copy in another EU region

One queue versus four

Datastore NVMe Performance, per datastore, 4K random, 70% read / 30% write.

Four parallel queues
up to 120,000 IOPS
One queue
around 28,000 IOPS

A number like 120,000 IOPS comes from many simultaneously open requests, not from one fast one. A guest with a single queue waits for the acknowledgement after each request; the round-trip time over the storage network then limits it, not the medium.

The same relationship applies in the other direction when many guests are on the same datastore. From 24 concurrently active VMs on a datastore, the guaranteed latency drops from 0.9 ms to 1.4 ms. Larger environments are therefore split across multiple datastores.

Measurement profile

Block size
4 KB
Access pattern
random, 70% read / 30% write
Queues per guest
at least 4 for the IOPS commitment
Latency measurement window
24 hours, 99th percentile
Unit
per datastore, not per VM
Range across all classes
0.4 to 25.0 ms

If an environment permanently exceeds the values, we split it across multiple datastores. This is the usual way in operations — booking a higher class does not help if the queue and not the medium is the bottleneck.

Every class has a minimum commitment. If you request less, you still pay it — right from the first invoice. The rates range from €0.1090$0.1264 to €0.0068$0.0079 per GB and month, the minimum amounts from €73.73$85.53 (NVMe Standard) to €237.57$275.58 (Capacity). The order of the two ranges is not the same.

Rates, minimum commitment and upper limit of the four datastore classes. All amounts net, billed per started month. The minimum amount is the rate multiplied by the minimum commitment — it applies even if you use less.

Rates, minimum commitment and upper limit of the four datastore classes
ClassRate per GB per monthSnapshots per GB per monthMinimum commitmentMinimum amount per monthUpper limit
Datastore NVMe Performance€0.1090$0.1264€0.0210$0.0244+ 19%2 TB€223.23$258.95256 TB
Datastore NVMe Standard€0.0720$0.0835€0.0140$0.0162+ 19%1 TB€73.73$85.53256 TB
Datastore Capacity€0.0290$0.0336€0.0060$0.0070+ 21%8 TB€237.57$275.58512 TB
Datastore Archive€0.0068$0.0079€0.0009$0.0010+ 13%20 TB€139.26$161.542048 TB

All amounts net. The lowest minimum amount corresponds to €87.74$101.78 incl. 19% VAT.

The example calculation for 4 TB is in the storage calculator: the same amount in all four classes, requested and calculated amount side by side — and any other amount you set there.

Different amount? In the calculator above

Break-even points

From which requested amount the class with the lower rate costs less despite a higher minimum commitment.

Capacity vs NVMe Performance
from 2,180 GB
Capacity vs NVMe Standard
from 3,300 GB
Archive vs NVMe Standard
from 1,935 GB

Below these points, the seemingly more expensive class is the cheaper one. The rate of €0.0290$0.0336 per GB is of no use as long as 8 TB are billed.

Snapshots cost extra per class and are between 13% and 21% of the storage rate. They can be disabled, but not retroactively: what was not recorded cannot be recovered.

Two different things, two different windows

The snapshot is located on the same datastore and is used to roll back after a change. The backup leaves it and is used for recovery. Retention is between 14 and 90 days — different for each class, because the classes have different tasks.

Snapshot retention, backup window and snapshot rate per class. The rate is applied to the same calculated amount as the storage rate — including the minimum commitment.

Snapshot retention, backup window and snapshot rate per class
ClassSnapshot retentionSnapshot per GB per monthBackup window
Datastore NVMe Performance14 days€0.0210$0.0244Daily 22:00–04:00 CET, application-consistent via VSS or fsfreeze
Datastore NVMe Standard14 days€0.0140$0.0162Daily 22:00–05:00 CET, application-consistent
Datastore Capacity30 days€0.0060$0.0070Daily 21:00–06:00 CET, crash-consistent
Datastore Archive90 days€0.0009$0.0010Weekly Sat 20:00–08:00 CET
Snapshot — remains in the datastoreDatastoretoday14 days30 days90 daysBackup — leaves the datastoreDatastoreCopysecond region

What happens in the backup window

  1. TriggerThe run starts in the agreed windowThe windows are during the night and differ per class. For Datastore NVMe Performance: Daily 22:00–04:00 CET, application-consistent via VSS or fsfreeze. There is no postponement — outside the window, the fabric bandwidth is committed to guest traffic.
  2. QuiescingApplication consistency via VSS or fsfreezeBefore copying, the guest is briefly quiesced: on Windows via the Volume Shadow Copy Service, on Linux via fsfreeze on the mounted file systems. Without this step, the copy is crash-consistent — it starts, but a database inside must roll back its log.
  3. CopyingOnly the changed blocks leave the datastoreThe first run transfers the used capacity, each subsequent one the difference to the previous run. Therefore, the change rate determines the duration of the window, not the total size of the datastore.
  4. ExpirationOld versions are removed after expiryRetention is defined per class and ranges from 14 to 90 days. Longer retention is not a snapshot setting, but a backup to a second region.

Further off-site

Backups to a second region, replication with a guaranteed RPO (Recovery Point Objective: the maximum time period lost in an emergency) and rehearsed disaster recovery are a separate tier. It depends on the same calculated capacity as the datastore.

Disaster recovery: RPO and RTO per tier

Three fully populated storage racks side by side in a bright aisle; each rack unit holds a chassis with a row of drives, all status lights green.

Behind every datastore are chassis like these, connected to all hosts in the cluster via the fabric. The redundancy is built in: you order usable capacity, not the raw capacity of the drives.

Storage row

Not every class is attached to every stack

Which datastore class you can book depends on the virtualisation stack above it — the software layer comprising the hypervisor and management — and the location. Both are fixed before the first guest starts. Reattaching them later requires a migration.

Availability of datastore classes per virtualisation stack. A tick means the class can be booked under this stack. The columns list stacks, not certifications — all four classes can be integrated under the SAP stack, but only Datastore NVMe Performance is certified.

Availability of datastore classes per virtualisation stack
ClassKVM PlatformManaged vSphereHCISAP Certified
Datastore NVMe Performance
Datastore NVMe Standard
Datastore Capacity
Datastore Archive

Locations per class

A class can be booked where the appropriate storage fabric is located. The archive is available in the fewest locations because it writes its second copy to another EU region and needs a counterpart for this.

Datastore NVMe Performance2 locations
  • Frankfurt am Main fra1
  • Düsseldorf dus1
Datastore NVMe Standard2 locations
  • Frankfurt am Main fra1
  • Düsseldorf dus1
Datastore Capacity2 locations
  • Frankfurt am Main fra1
  • Düsseldorf dus1
Datastore Archive1 location
  • Frankfurt am Main fra1

Cities listed twice are two separate data centres with their own fire zones and their own power supply — the prerequisite for triple replication within a location to be meaningful at all. Locations and fire zones

Who patches what

Platform operations are included in the rate per gigabyte and cannot be deselected. They end at a strict boundary: we work below the guest system, you work above it.

ENTRONYX CLOUD patches

  • Hypervisor
  • BIOS and firmware
  • Fabric switches
  • Storage controllers

Firmware maintenance requires a host evacuation. Without a second host in the cluster, this means downtime of up to 45 minutes per maintenance window.

You patch

  • Guest operating system
  • Middleware
  • Application
  • Application data

The file system in the guest, the application above it and the data within it remain in your hands. If you want to hand this over too, you can book a higher operations level.

A maintenance window for the guests only arises where the cluster itself is touched — not at the storage. The smallest purchase of the fastest class is 2 TB; below this, the minimum quantity is charged, not the used quantity.

What is usually asked before ordering

Ten frequently asked questions before placing an order. The answers use the same catalogue values as the storage calculator and the tables above.

Is your question not listed? The storage calculator calculates every class and quantity, including the minimum order quantity.

What is a datastore in the Private Cloud, and how much does it cost per month?

A datastore is the shared storage used by all hosts in a cluster — under vSphere, this is the prerequisite for HA restart and vMotion. ENTRONYX CLOUD operates it in data centres in Germany in 4 classes with an IOPS limit and latency commitment; billing is per provisioned gigabyte and month, with a minimum commitment per class.

Rate from, per GB and month, net€0.0068$0.0079

Calculate monthly amount in the storage calculator

How quickly does support respond if a host fails?

An outage (P1) receives an initial response in 30 minutes, around the clock, in the Platform operations, which is included in every managed host; from the “Platform and guest operations” level it is 15 minutes. Initial response means: a human has accepted the report, classified it and reported back the processing status — we do not guarantee a resolution time. Reports are made via the ticket in the account; a change without an outage requires the lead time of the operations level, in the Platform operations 48 hours.

Who operates and patches the storage

What is the term, and how do I cancel?

Without a term agreement, the contract runs for one month and renews for another; terms of 12 or 24 months are paid in advance and you receive a 5% or 8% term discount respectively. You can cancel monthly contracts with one day's notice to the end of the term, contracts for 12 or 24 months with 30 calendar days' notice — otherwise they renew for twelve months. After the contract ends, your data remains accessible for 30 calendar days; after that, it is irretrievably deleted.

How is a datastore billed?

According to the provisioned amount in gigabytes, at the rate of the class per started month — not according to the used amount. Each class has a minimum purchase from 1 TB to 20 TB; if you request less, you still pay it, from the first invoice. Snapshots cost extra, there is no setup fee for the datastore.

Lowest monthly amount, Datastore NVMe Standard for minimum purchase, net€73.73$85.53

Calculate monthly amount in the storage calculator

Can I order a datastore individually?

No, a datastore is ordered together with the Private Cloud cluster; you select the term and discount there in the configurator. Datastore Archive is not booked via the configurator: The class supplements an existing cluster and is requested via the contact form.

Configure Private Cloud

What is guaranteed for IOPS and latency?

An upper limit per datastore is guaranteed for each class, not per virtual machine — in Datastore NVMe Performance up to 120,000 IOPS at 0.4 ms latency p99. Every guarantee applies under a condition stated with the class: Datastore NVMe Performance only reaches the IOPS limit with several parallel queues, for example; with a single one, around 28,000 remain. A higher class does not help if the queue and not the medium is the bottleneck.

Measurement profile and values per class

Via which protocols is a datastore connected?

Via NVMe/TCP, iSCSI or NFS 4.1, depending on the class — the protocol follows from the underlying medium and the latency guaranteed on it. NVMe/TCP speaks only Datastore NVMe Performance; iSCSI speak Datastore NVMe Standard and Datastore Capacity; NFS 4.1 speak all four classes. Storage and guest traffic run on separate links of the storage fabric.

Three protocols, one fabric

How large can a datastore be?

Depending on the class, up to 256 TB, 512 TB or 2048 TB per datastore. Larger environments are cut into several datastores, as are environments that permanently exceed the performance values. Which class you can book follows from the virtualisation stack and location; moving both later means a migration.

Minimum purchase and upper limit per class

How long are snapshots kept, and what do they cost?

Depending on the class 14 to 90 days; they cost an additional 13% to 21% of the storage rate, calculated on the same billed amount. Snapshots are optional, but do not work retroactively: what was not recorded cannot be recovered. They reside on the same datastore and therefore do not replace a backup.

Retention and backup window per class

Where is the data located?

In data centres in Germany: Frankfurt am Main (fra1) and Düsseldorf (dus1). Not every class is available everywhere: Datastore Archive is only available in Frankfurt am Main (fra1) and writes its second copy to another EU region. The redundancy is included in the rate — you order usable capacity, not raw capacity.

Locations per class

Class selected? The cluster is added.

A datastore is ordered with the cluster: the configurator guides you from the platform via the hosts to the storage and shows the monthly invoice before you order — including the minimum purchase. All amounts net, without setup fee.

Classes
4
Rate from
€0.0068$0.0079 / GB
from
€73.73$85.53 / month
IOPS up to
120,000