Approved are 4 size classes — HANA S, HANA M, HANA L, HANA XL — and 4 of 5 host classes. Approved means: A test report according to SAP HWCCT, the Hardware Configuration Check Tool from SAP, is available for exactly this configuration. There is none for any other combination — and without a report, the memory commitment does not apply.
Nodes from €2,313.00$2,683.08 net per month, host and licence, excluding datastore — the shared storage cluster where the volumes reside. Licensing is per host, not per core. Approval is possible at 3 locations: anywhere the certified datastore is available alongside the host class.
A node is only considered accepted when an HWCCT report is available for exactly this configuration: host, memory, datastore and volume layout — the distribution of storage across the three HANA file systems — in the combination in which they are delivered. You receive the report upon handover and again after every change.
ENTRONYX SAP Certified
per host and monthCertified HANA nodes, licensed per host — without overcommitment.
The maintenance window is the previously agreed period during which work may be performed on a node; outside of this, it is not touched. The licence is tied to the host, not the core count. A node with 128 cores costs the same in terms of licensing as one with 32 — unlike under vSphere and HCI, which bill per core.
The delivered configuration is always approved, not a single metric. Therefore, the justification is per class and not as a general clock limit.
3 Gbit uplink without a separate storage fabric. Shared NFS or iSCSI datastores are not permitted on this class so that replication and guest traffic do not share the same line. Without a shared datastore, there is no automatic HA restart.
The class designates the physical configuration of the host. This determines the volume cut, the allocatable database size and the number of vCPUs, i.e. the virtual processor cores — without any leeway, because the SAP stack does not allow compute or memory overprovisioning.
S/4HANA development, quality assurance, small secondary systems
€2,313.00$2,683.08
per month, net
Host including platform operations and licence. Datastore is billed separately.
Productive S/4HANA in SMEs, Intel certification with AMX
€2,383.00$2,764.28
per month, net
Host including platform operations and licence. Datastore is billed separately.
S/4HANA with connected BW, mixed OLTP and OLAP load
€3,188.65$3,698.83
per month, net
Host including platform operations and licence. Datastore is billed separately.
BW/4HANA with large columnar storage, consolidated corporate systems
€7,083.18$8,216.49
per month, net
Host including platform operations and licence. Datastore is billed separately.
| Class | Host | Processor | Cores | Base clock | RAM | allocatable | vCPU | per month |
|---|---|---|---|---|---|---|---|---|
| HANA S | PC-M64 | EPYC 9454P | 48 | 2.75 GHz | 256 GB | 232 GB | 48 | €2,313.00$2,683.08 |
| HANA M | PC-M64i | Xeon 6741P | 48 | 2.50 GHz | 256 GB | 232 GB | 48 | €2,383.00$2,764.28 |
| HANA L | PC-L128 | EPYC 9554 | 64 | 3.10 GHz | 512 GB | 472 GB | 64 | €3,188.65$3,698.83 |
| HANA XL | PC-X256 | 2× EPYC 9965 | 384 | 2.25 GHz | 1 TB | 960 GB | 384 | €7,083.18$8,216.49 |
The hypervisor — the virtualisation layer of the host — occupies a fixed base plus a share of the configuration. With PC-M64, 232 GB remain for the guest out of 256 GB. The SAP volume rules still calculate based on the physical configuration — this is how SAP defines them. The guaranteed database size of 128 GB is significantly lower, because HANA needs the second half of the memory for merges, intermediate results and temporary tables.
Certification possible in 3
The price per chip shows HANA S including location factor, net and without datastore. Although SAP-capable host classes are available in Berlin, the certified datastore NVMe Performance is not — certification is not possible there.
SAP dictates the layout, not us. The three file systems follow directly from the memory of the class — and thus also the amount that must be ordered on the certified datastore.
| Class | RAM | /hana/data | /hana/log | /hana/shared | Total | Database up to |
|---|---|---|---|---|---|---|
| HANA S | 256 GB | 256 GB1 × 256 | 128 GBmin(128; 512) | 256 GBmin(256; 1,024) | 640 GB | 128 GB |
| HANA M | 256 GB | 256 GB1 × 256 | 128 GBmin(128; 512) | 256 GBmin(256; 1,024) | 640 GB | 128 GB |
| HANA L | 512 GB | 512 GB1 × 512 | 256 GBmin(256; 512) | 512 GBmin(512; 1,024) | 1,280 GB | 256 GB |
| HANA XL | 1,024 GB | 1,024 GB1 × 1,024 | 512 GBmin(512; 512) | 1,024 GBmin(1,024; 1,024) | 2,560 GB | 512 GB |
The certified datastore NVMe Performance has a minimum commitment of 2,048 GB at €0.1090$0.1264 per GB and month. 3 of the 4 classes are below this with their volume layout and therefore pay the same minimum amount — from HANA S to HANA L the storage price does not change. Only HANA XL exceeds it with 2,560 GB.
| Class | Total volume | billed | per month | with snapshots |
|---|---|---|---|---|
| HANA S | 640 GB | 2,048 GB | €223.23$258.95 | €266.24$308.84 |
| HANA M | 640 GB | 2,048 GB | €223.23$258.95 | €266.24$308.84 |
| HANA L | 1,280 GB | 2,048 GB | €223.23$258.95 | €266.24$308.84 |
| HANA XL | 2,560 GB | 2,560 GB | €279.04$323.69 | €332.80$386.05 |
The SAP stack lowers the vCPU limit to 1:1: one virtual CPU per physical core, not per thread. This removes a significant portion of the placement reserve on the same bare metal that the same host class would be allowed to carry under any other stack.
| Feature | PC-M64HANA S | PC-M64iHANA M | PC-L128HANA L | PC-X256HANA XL |
|---|---|---|---|---|
| Physical cores | 48 | 48 | 64 | 384 |
| ThreadsSMT — two threads per core — is active, but does not count as a vCPU | 96 | 96 | 128 | 768 |
| Base clock | 2.75 GHz | 2.50 GHz | 3.10 GHz | 2.25 GHz |
| Platform vCPU limitwithout SAP stack, according to host class | 288 · 6:1 | 288 · 6:1 | 256 · 4:1 | 1,536 · 4:1 |
| vCPU under certification1:1, capped by ENTRONYX SAP Certified | 48 · 1:1 | 48 · 1:1 | 64 · 1:1 | 384 · 1:1 |
| Difference | − 240 · 83.3% | − 240 · 83.3% | − 192 · 75.0% | − 1,152 · 75.0% |
| Physical memory | 256 GB | 256 GB | 512 GB | 1 TB |
| of which allocatableafter reserve for the hypervisor, the virtualisation layer | 232 GB | 232 GB | 472 GB | 960 GB |
| Memory overcommit | 0% | 0% | 0% | 0% |
SAP checks four metrics in the HWCCT run. The limit is what we contractually guarantee; the measured value comes from the acceptance run on the certified datastore and merely proves it. If the configuration or datastore changes, it is measured again.
| Metric | guaranteed | measured |
|---|---|---|
| Log write latency, 4 KB | ≤ 1000 µs | 310 µs |
| Log write latency, 16 KB | ≤ 1000 µs | 420 µs |
| Data write throughput, 16 MB | ≥ 400 MB/s | 2140 MB/s |
| Data read throughput, 16 MB | ≥ 400 MB/s | 3060 MB/s |
All 4 classes can be created under the SAP stack — this makes sense for templates, exports and powered-off secondary systems. The volumes /hana/data, /hana/log and /hana/shared of a productive instance, however, belong exclusively on Datastore NVMe Performance; any other class suspends the acceptance.
| Feature | NVMe PerformanceNVMe/TCP · NFS 4.1 | NVMe StandardNFS 4.1 · iSCSI | CapacityiSCSI · NFS 4.1 | ArchiveNFS 4.1 |
|---|---|---|---|---|
| SAP acceptanceHWCCT log per node | − | − | − | |
| IOPS per datastoreAccesses per second: 4 KB random, 70/30 read/write | 120,000 | 45,000 | 9,000 | 900 |
| Throughput | 6,000 MB/s | 2,400 MB/s | 1,100 MB/s | 350 MB/s |
| Latency p9999% of all accesses are faster than this value | 0.4 ms | 0.9 ms | 4.0 ms | 25.0 ms |
| Price per GB and month | €0.1090$0.1264 | €0.0720$0.0835 | €0.0290$0.0336 | €0.0068$0.0079 |
| Snapshots per GB per month | €0.0210$0.0244 | €0.0140$0.0162 | €0.0060$0.0070 | €0.0009$0.0010 |
| Minimum commitment | 2 TB | 1 TB | 8 TB | 20 TB |
| Snapshot retention | 14 days | 14 days | 30 days | 90 days |
| Backup window | Daily 22:00–04:00 CET, application-consistent via VSS or fsfreeze | Daily 22:00–05:00 CET, application-consistent | Daily 21:00–06:00 CET, crash-consistent | Weekly Sat 20:00–08:00 CET |
| Redundancy | Triply replicated across two fire zones of the location | Triply replicated across two fire zones of the location | Erasure coding 8+2 in the location, cache triply mirrored | Erasure coding 8+2, second copy in another EU region |
The limit of Datastore NVMe Performance is stated above in the note on the IOPS guarantee — it is not repeated here.
An accepted node is not maintained on a rolling basis in a fixed weekly window like a vSphere or HCI cluster. Every change to firmware, hypervisor or datastore affects the acceptance — which is why each one is agreed individually.
Maintenance windows
By arrangement per environment, 21 days lead time, outside your closing periods
The lead time depends on the booked operations level. You store closing periods, inventories and payment runs once; after that, windows will no longer be scheduled during these times.
| Tier | Initial response P1 | Initial response P2 | Availability | Lead time | per environment | per host |
|---|---|---|---|---|---|---|
| Platform operations | 30 min | 4 h | P1 around the clock, P2 and P3 Mon–Fri 08:00–18:00 | 48 h | included | included |
| Platform and guest operations | 15 min | 2 h | P1 and P2 around the clock, P3 Mon–Sat 08:00–20:00 | 24 h | €690.00$800.40 | €39.00$45.24 |
| Full operations | 15 min | 1 h | Around the clock via fixed phone number, named technical contact | 8 h | €2,490.00$2,888.40 | €89.00$103.24 |
The prices apply per protected environment plus a host rate — operations effort grows with the number of systems, not just with the number of nodes. The complete model including the access limit between customer and operator can be found at Managed dedicated hosts.
The class determines the volume layout, datastore and price. The acceptance report is generated on the delivered configuration and is provided to you upon handover. Changing the class, location or datastore requires a new acceptance — which is why it is worth calculating the layout together once.
