STREAM-Intelligence · USB4SAP · Focused Comparison
USB4SAP vs Simplement Roundhouse, dab Nexus, and Theobald Xtract — table CDC and semantic model delivery to Microsoft Fabric, OneLake, and cloud. April 2026.
Scope of this comparison. This table focuses on the specific parameters that matter when choosing a tool to stream SAP table changes and semantic business models to Microsoft Fabric, OneLake, or cloud data platforms — covering CDC method, semantic readiness, RISE compatibility, clean-core alignment, and live access. All four tools can move SAP data to cloud; this table shows how and at what cost to your SAP environment. Sources: vendor documentation, SAP notes, and Microsoft partner certification records, April 2026.
Scroll horizontally to see all vendors →
| USB4SAP Zero-copy streaming API · Patented | Simplement Roundhouse · SAP app-layer CDC | dab Nexus SAP app-layer extraction | Theobald Xtract On-prem SAP connector suite | |
|---|---|---|---|---|
| SAP Strategic Alignment | ||||
| SAP clean-core aligned No custom ABAP, no modification of SAP core | Yes No ABAP. No custom dev. SAP Note 3255746 compliant. NetWeaver certified. | Yes Operates via SAP application layer; no Z-table creation or core mods | Yes Application-layer extraction; no ABAP modifications required | Partial Custom ABAP package installed for table extraction; Table CDC uses proprietary logging |
| SAP RISE Public Cloud compatible Works without direct DB access or Z-table writes on RISE | Yes No DB access. No Z-tables. Fully RISE Public Cloud compatible. | Yes Application-layer approach works on RISE; no DB-level access required | Yes Application-layer; RISE Public Cloud compatible | Partial ABAP add-on requires transport; RISE Public Cloud blocks Z-table creation; Private Cloud more permissive |
| SAP Note 3255746 compliant ODP-RFC path deprecated; OData/REST required for table delta | Yes REST/ODP/RFC supported. Fully compliant. SAP-certified via NetWeaver. | Yes App-layer approach bypasses deprecated ODP-RFC path | Yes Application-layer; not reliant on deprecated ODP-RFC for table delta | Yes Table CDC avoids ODP-RFC; ODP + OData path used; deprecated path removed |
| Change Data Capture (CDC) & Table Streaming | ||||
| CDC for tables — without SLT Table-level delta without SAP Landscape Transformation Server | Yes Timeseries & Doc# series CDC. No SLT. No change pointers. 0.143s on 100M+ row table. | Yes Application-layer CDC; no SLT, no change pointers required | Yes No SLT required; app-layer delta method | Yes Table CDC component with proprietary logging; no SLT required |
| CDC delta method Technical mechanism for detecting changes | Timeseries & Doc# series Field-level combination delta. Detects inserts, updates, deletes. No change pointers activation. Not timestamp-only. | SAP application layer Reads via RFC/BAPI; change detection through SAP update records or timestamp fields | SAP application layer App-layer polling / delta via SAP standard mechanisms | Proprietary table logging Near real-time; custom log table written on SAP side; not sub-second streaming |
| Delete detection in CDC Captures hard-deleted records, not just inserts/updates | Yes Field-level delta natively detects deletes. No additional configuration. | Partial Soft deletes via SAP logical deletion flags; hard deletes not guaranteed without additional config | Partial App-layer approach; hard delete capture depends on SAP object configuration | Partial Proprietary log captures some deletes; completeness depends on table and log configuration |
| CDC performance at scale Throughput on large tables (100M+ rows) | 0.143s delta Benchmarked on 100M+ row table. Co-validated with Microsoft. | Partial App-layer throughput constrained by SAP dialog/RFC work process limits; no published benchmark | Partial Application-layer throughput; SAP work process dependent; no published benchmark | Partial Near real-time; no sub-second published benchmark for 100M+ row tables |
| On-demand / live extraction (no schedule) Pull live SAP data per API request without a pipeline | Yes On-demand REST API call. No pipeline. No schedule. Live data per request. | No Schedule or event-triggered replication; no live per-request API | No Batch/scheduled extraction model; no live REST query | No Scheduled near-real-time; not true on-demand live API query |
| Semantic Model Delivery | ||||
| Pre-built SAP semantic models Named business objects (Purchase Order) not raw tables (EKKO/EKPO) | Yes 450K+ pre-built semantic models. Purchase Order arrives as Purchase Order. No reverse-engineering required. | No Raw table and DataSource extraction; no pre-built business model catalog | No Table and extractor focused; no semantic catalog | Partial BEx queries carry some semantic context; no pre-built business model catalog |
| Business field labels alongside technical names Consuming app sees both LIFNR and "Vendor Number" | Yes Business & technical field names. Always. Every object. | Partial Technical field names primary; some label metadata available via RFC metadata call | Partial Metadata available; not consistently surfaced at delivery layer | Yes Technical and business field metadata exposed via ODP / OData metadata service |
| BW model migration support InfoCubes, DSOs, BEx queries extractable to cloud | Yes BW extractors, InfoCubes, DSOs, BEx queries, CDS — all natively. Key 2027 migration path. | Partial BW DataSources via ODP; InfoCube/DSO layer requires BW extractor configuration | Partial BW extractors via standard SAP mechanisms; not primary focus | Yes DeltaQ + ODP; BW/4HANA extractors, BEx queries — primary use case |
| Semantic mirroring to OneLake Business-named models land in OneLake, not raw table dumps | Yes 450K+ semantic models provisioned directly into OneLake. Business-ready on arrival. | No Replicates raw tables/DataSources; no semantic translation at landing | No Table-level landing; semantic layer must be built post-load | Partial Fabric Open Mirroring supported; semantic naming depends on source object (BEx adds context) |
| Microsoft Fabric & Cloud Delivery | ||||
| Microsoft Fabric Open Mirroring support Native SAP-to-Fabric mirroring without intermediate storage | Yes Microsoft-certified. ADLS, OneLake, Fabric, Power BI, Excel. Azure Marketplace published. | Yes Roundhouse supports Fabric Open Mirroring; SAP-to-OneLake path | Partial Fabric delivery supported; Open Mirroring native integration not confirmed in public docs | Yes Xtract Universal + Fabric Open Mirroring connector (2024); OneLake and ADLS targets |
| Live Power BI / Excel without scheduled refresh SAP data live in BI tools per query, not on a copy | Yes Microsoft-certified. Live in Excel, Power BI, Fabric, API-M. No scheduled refresh. No export. | No Replication model; Power BI reads from replicated copy | No Extraction-based; BI tools read from loaded copy | Partial Power BI via scheduled refresh from extracted dataset; not live SAP query |
| Multi-cloud targets (AWS, GCP, Databricks, Snowflake) Delivery beyond Microsoft Fabric | Yes AWS, GCP, Databricks, Snowflake — REST API works with any target. No vendor lock-in. | Partial Primary focus Microsoft Fabric; other cloud targets via partner integrations | Partial Multiple cloud targets supported; SAP-to-cloud focused | Yes Snowflake, S3, Redshift, Databricks, ADLS — broad target support via Xtract Universal |
| Security & Data Sovereignty | ||||
| SAP native authorizations inherited Roles and auth objects enforced at extraction — no shadow security build | Yes 100% native SAP application authorizations. Role changes reflect immediately. No shadow copies outside auth model. | Partial Runs as SAP user; role-based access applies at RFC/BAPI layer; downstream copy is not auth-aware | Partial SAP credentials respected at extraction; downstream copy loses auth context | Partial SAP user credentials used; no automatic role inheritance in target; separate auth layer required |
| Zero-copy data sovereignty Data never leaves customer environment without explicit consent | Yes STREAM Activate: data never leaves your environment. No external pipeline. Full sovereignty. | Partial Data replicated to target; customer controls destination but data leaves SAP perimeter | Partial Data moves to target environment; no persistent copy in vendor infrastructure | Yes On-premise Xtract server in customer environment; data does not transit vendor infrastructure |
| No-Code & AI Readiness | ||||
| Zero customization on SAP side required No CDS annotation, no transport, no ABAP per object | Yes No CDS creation. No ABAP. No annotations. All 1.9M+ objects available on-demand. | Yes No custom ABAP or CDS annotation needed; app-layer approach is configuration-only | Yes Configuration-based; no per-object SAP customization required | Partial Custom ABAP package install required; some CDS objects need ODP annotation prep |
| Modern REST API — no SAP SDK required for consumers Any app can call SAP data without JCo, RFC, or ABAP knowledge | Yes 1.3M+ objects as no-code REST API. No SAP SDK. No on-premise agent. Works with any REST consumer. | No Pipeline/replication tool; downstream apps read from replicated copy, not live SAP via REST | No Extraction tool; no live consumer REST API for SAP objects | Partial yunIO (separate Theobald product) provides REST API; requires separate setup and licensing |
| AI agent ready (MCP / live SAP context) Suitable for AI agents querying live SAP data per inference | Yes MCP catalog for AI agents. Live SAP data. Verifiable-at-source AI. P2P Agent autonomous since Nov 2023. | No Replication latency; downstream copy not suitable for real-time agent queries | No Batch extraction model; not a live API for AI agents | No Extraction / replication pipeline; not designed for live agentic queries |
| Summary | ||||
| Primary use case fit | Live SAP API for analytics, AI, automation & semantic mirroring — no pipeline, no copy, no staging | SAP CDC replication to Fabric / cloud — application-layer approach, RISE-safe, no ABAP | SAP table & extractor delivery to cloud — configuration-based, RISE-compatible | SAP extraction to BI & analytics — broad object coverage, strong BW/BEx support, on-prem focus |
| Key limitation | No built-in transformation layer (by design — schema-on-read at destination) | No semantic model catalog; raw table landing requires downstream modeling | No semantic catalog; Open Mirroring native integration not fully confirmed; no live API | ABAP add-on limits RISE Public Cloud deployments; no live REST API for consumers; no semantic catalog |
Sources & research notes (April 2026):
Simplement Roundhouse: simplement.us · Roundhouse product documentation. Application-layer SAP CDC confirmed. Fabric Open Mirroring support confirmed via partner announcements. No Z-table or direct DB access — RISE Public Cloud compatible. No published sub-second CDC benchmark. No semantic model catalog in public documentation.
dab Nexus: dabnexus.com · Product documentation and partner materials. SAP application-layer extraction. RISE-compatible. Fabric delivery supported. Open Mirroring native integration status unconfirmed in public product docs as of April 2026.
Theobald Xtract: theobald-software.com/en/xtract-universal · Table CDC component confirmed (no SLT required). Fabric Open Mirroring connector released 2024. Custom ABAP package for table extraction — limits RISE Public Cloud compatibility. yunIO (separate product) for REST API. ODP + OData extension. ERPConnect SAP-certified.
USB4SAP: SAP Note 3255746. SAP NetWeaver certification. Microsoft certification (learn.microsoft.com/connectors/usb4sap). USPTO patent. Azure Marketplace published. In production since 2021. CDC benchmark (0.143s on 100M+ rows) co-validated with Microsoft. All object counts (1.9M+, 450K semantic models) independently verified.
This table is maintained by Ecoservity and reflects publicly available vendor documentation. Vendor features evolve rapidly — if you believe a cell is incorrect, contact us with a source reference and we will update. Last full research pass: April 2026.
See USB4SAP stream your SAP data live
30-day free trial. No SAP project. No on-premise agent. Live from day one.