Microsoft Fabric Consulting: OneLake, Direct Lake, and a Capacity Bill You Can Actually Justify

Microsoft Fabric is a unified, SaaS-based analytics platform that combines Power BI, Data Factory, Synapse Data Engineering, Synapse Data Warehouse, Synapse Real-Time Intelligence, and Data Activator on a single OneLake storage foundation. Businesses typically hire a Fabric consultant when they already run Power BI Premium, Azure Synapse, or a third-party warehouse like Snowflake and want to consolidate workloads onto one platform, when Copilot in Power BI is on the roadmap and the capacity question is now blocking, or when an F-SKU is provisioned and the actual cost per month came in higher than expected.

Petronella Technology Group, Inc. delivers Fabric Foundation, Fabric Enterprise, and F-SKU Right-Sizing Audit engagements with security, compliance, and AI augmentation built into every milestone.

Power BI is the visualization and semantic-model layer. Azure Synapse was Microsoft's previous generation of separate data-engineering, warehouse, and analytics services that you provisioned and stitched together yourself. Microsoft Fabric is a unified SaaS that brings those experiences plus Power BI together on shared OneLake storage with a single capacity (F-SKU) billing model.

The practical difference is that you stop paying for separate Synapse pools, separate Power BI Premium capacities, and separate ADLS Gen2 storage; you also stop copying data between them. For most mid-market businesses with an existing Power BI Premium footprint plus a small to mid-size warehouse, Fabric is a real consolidation. For very large or specialized analytical workloads, Synapse-on-Spark or a dedicated warehouse still has a place, we will tell you when that is the case rather than recommending Fabric by default.

OneLake is the single storage foundation underneath Microsoft Fabric, Microsoft positions it as "the OneDrive for data." Per Microsoft Learn, OneLake is automatically provisioned with every Fabric tenant, organizes data in workspaces, and stores everything as open Delta-Parquet tables. The result is that the same dataset is reachable from every Fabric experience, Data Engineering, Data Warehouse, Data Science, Real-Time Intelligence, Power BI, without copies.

OneLake also exposes shortcuts to external sources (ADLS Gen2, Amazon S3, Google Cloud Storage, Dataverse) so you can query data in place. For governance-conscious clients, the single-copy story dramatically simplifies sensitivity-label enforcement and audit logging.

Direct Lake is a Power BI semantic-model storage mode unique to Fabric. Per Microsoft Learn, Direct Lake loads Delta-Parquet data directly from OneLake into Power BI's columnar in-memory engine at query time, without the scheduled refresh of Import mode and without the per-query database round-trip of DirectQuery.

The result is dashboards that read from the freshest data in the lake at near-Import performance, with no refresh schedule to manage and no source-system load. For executive dashboards, near-real-time operations dashboards, and Copilot-in-Power-BI scenarios that need consistent low latency, Direct Lake is the most important architectural shift in Fabric. Petronella Technology Group, Inc. evaluates Direct Lake eligibility (table sizes, transformations, fallback behavior) as part of every Fabric Foundation engagement.

A Lakehouse in Fabric is the data-engineering home in OneLake, Delta-Parquet tables, files, and notebooks together in one workspace artifact. The medallion architecture is a layered design pattern Microsoft documents on Fabric Learn: a Bronze layer for raw ingested data as-is, a Silver layer for cleaned and conformed data with standardized schema, and a Gold layer for analytics-ready, business-validated tables that feed semantic models.

The pattern keeps lineage clean, allows reprocessing without re-ingesting, and aligns to Direct Lake performance expectations. Petronella Technology Group, Inc. uses the medallion pattern by default and adds a parallel Compliance layer for clients with regulatory data classification needs.

An F-SKU is a Fabric capacity SKU billed per Capacity Unit (CU) per hour, with pause and resume controls in the Azure portal. Per Microsoft Learn, F-SKUs range from F2 (smallest) through F2048. The minimum F-SKU that enables Copilot in Power BI is F2, meaning yes, Copilot in Power BI requires Fabric capacity (or Power BI Premium P1 or higher), and a Pro license alone is not sufficient.

Many organizations provision an F64 because that SKU also includes per-user Power BI viewing for free (a parity with the Power BI Premium P1 license model). Right-sizing the F-SKU to actual workload is where most of our F-SKU Right-Sizing Audit savings come from.

Data Activator is the Fabric experience that watches data, streaming events, Lakehouse tables, Power BI semantic models, for conditions you define, and triggers actions when those conditions hit. Per Microsoft Learn, actions include sending email or Teams alerts, calling Power Automate flows, and invoking custom endpoints.

Practical examples: alert the plant manager when scrap rate on a production line crosses a threshold, page the on-call analyst when a refresh fails, escalate to vCISO when a sensitivity-label-restricted dataset is shared outside the boundary. We treat Data Activator as the operational glue that turns dashboards from observation into action.

Yes to all three, with different effort profiles. From Azure Synapse Dedicated SQL Pools, Microsoft provides a documented migration assistant and the SQL surface inside Fabric Warehouse is broadly compatible, most schemas, stored procedures, and views migrate with light revision.

From Snowflake, the lift is larger because external table syntax, stored procedures, and time-travel semantics differ; we typically rebuild the warehouse layer rather than line-by-line translate. From on-premises SQL Server data warehousing, we use Data Factory pipelines in Fabric plus shortcuts and mirroring to land data in OneLake while keeping the legacy warehouse running in parallel until cutover. Petronella Technology Group, Inc. delivers migration as either a standalone engagement or as Phase 1 of a Fabric Enterprise build.

Copilot in Power BI focuses on the report-author and report-consumer experience, natural-language questions over a semantic model, narrative summaries of visuals, DAX measure generation. Copilot in Fabric extends that to every Fabric experience: writing SQL or KQL queries, generating PySpark notebooks in Data Engineering, building Data Factory pipelines from a natural-language description, and authoring Data Activator alerts.

The licensing dependency is the same, Fabric F2 or higher capacity with admin enablement, but the surface area is broader. For regulated clients where Copilot cannot touch restricted-label data, Petronella Technology Group, Inc. recommends Penny, our private AI assistant running on PTG-owned GPU infrastructure, as the conversational layer instead.

Yes with the right tenant configuration. For HIPAA, Microsoft Fabric is covered under the Microsoft 365 BAA on commercial and GCC tenants when configured correctly, sensitivity labels, RLS, audit-log retention, BAA-aware sharing controls, and Publish-to-Web blocked.

For CMMC Level 2 and CUI workloads, Fabric on commercial is generally not the right answer; we use a three-pattern decision tree (GCC High Fabric / Petronella encrypted enclave with de-identified data flowing into commercial Fabric / Fabric Warehouse on private capacity behind authorization boundary) depending on the data flow. Our CMMC Power BI Reporting page documents the three patterns in detail.

An F-SKU Right-Sizing Audit examines four signals: actual Capacity Unit consumption telemetry from the Fabric Capacity Metrics app, the workload mix (interactive Power BI, scheduled Data Engineering jobs, Direct Lake refresh patterns, Real-Time Intelligence streaming), Copilot enablement status, and Power BI per-user license counts against the F64-or-higher viewer-included threshold.

We then model two to three alternative SKU configurations including a pause-and-resume schedule for non-business-hours capacity, and present a savings forecast. Where the audit finds genuine over-provisioning, mid-market clients have seen meaningful reductions in monthly Fabric spend, we publish the actual results of each audit rather than market-average claims.

Fabric Foundation (capacity provisioning, one Lakehouse with bronze and silver layers, one to two Direct Lake semantic models, two to three role-based dashboards, governance handbook) ships in four to six weeks. Fabric Enterprise (multi-domain Lakehouse with full medallion, Data Factory pipelines, Real-Time Intelligence for one workload, Data Activator alerts, Copilot governance, vCISO oversight, sensitivity-label rollout) ships in eight to twelve weeks. F-SKU Right-Sizing Audit ships in two to three weeks. All engagements are fixed-fee with quoted starting points and are paid 100% upfront at contract execution.

Request a Microsoft Fabric quote

Yes, Copilot in Power BI requires either a Microsoft Fabric capacity (F2 or higher) or a Power BI Premium capacity (P1 or higher), with Copilot enabled at the tenant and capacity level. A Power BI Pro license alone is not sufficient. Per Microsoft Learn, as of April 28, 2025 the Fabric Copilot Capacity (FCC) is available starting at F2, earlier guidance required F64, which was relaxed once FCC shipped.

Tenants on commercial Power BI Premium P1+ remain eligible without a separate F-SKU. For organizations that want Copilot-style natural-language analytics without exposing regulated or restricted-label data to a Microsoft-hosted model, we offer Penny, our private AI assistant running on PTG-owned GPU infrastructure, as an alternative conversational layer.

F SKUs are the unified Microsoft Fabric capacity model, they bill per Capacity Unit (CU) per hour with pause and resume in the Azure portal, cover every Fabric experience (Power BI, Data Factory, Synapse Data Engineering, Data Warehouse, Real-Time Intelligence, Data Activator) on shared OneLake storage, and are the SKU family Microsoft is investing in going forward.

P SKUs (P1, P2, P3) were the legacy Power BI Premium capacity tier, Power-BI-only, billed monthly with no pause, and tied to per-user license inclusions inside Power BI. Microsoft has communicated that P SKUs are being phased out in favor of F SKUs; existing P-SKU customers are being migrated to equivalent F-SKU capacities, and new Premium purchases default to F SKUs. F64 is the parity point where the F-SKU model includes free Power BI viewing for all users in the tenant, matching the old P1 license-inclusion behavior. For any new Fabric or Power BI Premium engagement in 2026, Petronella Technology Group, Inc. provisions F SKUs, not P SKUs.