How Great Engineers Make Architectural Decisions — ADRs, Trade-offs, and an ATAM-Lite Checklist
Why Decision-Making Matters Without a shared framework, context fades and teams' re-debate old choices. ADRs solve that by recording the why behind design decisions — what problem we solved, what options we considered, and what trade-offs we accepted. A good ADR: Lives next to the code in your repo. Explains reasoning in plain language. Survives personnel changes and version history. Think of it as your team’s engineering memory. The Five Pillars of Trade-offs At Microsoft, we frame every major design discussion using the Azure Well-Architected pillars: Reliability – Will the system recover gracefully from failures? Performance Efficiency – Can it meet latency and throughput targets? Cost Optimization – Are we using resources efficiently? Security – Are we minimizing blast radius and exposure? Operational Excellence – Can we deploy, monitor, and fix quickly? No decision optimizes all five. Great engineers make conscious trade-offs — and document them. A Practical Decision Flow Step What to Do Output 1. Frame It Clarify the problem, constraints, and quality goals (SLOs, cost caps). Problem statement 2. List Options Identify 2-4 realistic approaches. Options list 3. Score Trade-offs Use a Decision Matrix to rate options (1–5) against pillars. Table of scores 4. ATAM-Lite Review List scenarios, identify sensitivity points (small changes with big impact) and risks. Risk notes 5. Record It as an ADR Capture everything in one markdown doc beside the code. ADR file Example: Adding a Read-Through Cache Decision: Add a Redis cache in front of Cosmos DB to reduce read latency. Context: Average P95 latency from DB is 80 ms; target is < 15 ms. Options: A) Query DB directly B) Add read-through cache using Redis Trade-offs Performance: + Massive improvement in read speed. Cost: + Fewer RU/s on Cosmos DB. Reliability: − Risk of stale data if cache invalidation fails. Operational: + Added complexity for monitoring and TTLs. Templates You Can Re-use ADR Template # ADR-001: Add Read-through Cache in Front of Cosmos DB Status: Accepted Date: 2025-10-21 Context: High read latency; P95 = 80ms, target <15ms Options: A) Direct DB reads B) Redis cache for hot keys ✅ Decision: Adopt Redis cache for performance and cost optimization. Consequences: - Improved read latency and reduced RU/s cost - Risk of data staleness during cache invalidation - Added operational complexity Links: PR#3421, Design Doc #204, Azure Monitor dashboard Decision Matrix Example Pillar Weight Option A Option B Notes Reliability 5 3 4 Redis clustering handles failover Performance 4 2 5 In-memory reads Cost 3 4 5 Reduced RU/s Security 4 4 4 Same auth posture Operational Excellence 3 4 3 More moving parts Weighted total = Σ(weight × score) → best overall score wins. Team Guidelines Create a /docs/adr folder in each repo. One ADR per significant change; supersede old ones instead of editing history. Link ADRs in design reviews and PRs. Revisit when constraints change (incidents, new SLOs, cost shifts). Publish insights as follow-up blogs to grow shared knowledge. Why It Works This practice connects the theory of trade-offs with Microsoft’s engineering culture of reliability and transparency. It improves onboarding, enables faster design reviews, and builds a traceable record of engineering evolution. Join the Conversation Have you tried ADRs or other decision frameworks in your projects? Share your experience in the comments or link to your own public templates — let’s make architectural reasoning part of our shared language.Secure Delta Sharing Between Databricks Workspaces Using NCC and Private Endpoints
This guide walks you through the steps to share Delta tables between two Databricks workspaces (NorthCentral and SouthCentral) and configure Network Connectivity Configuration (NCC) for a Serverless Warehouse. These steps ensure secure data sharing and connectivity for your workloads. Part 1: Delta Sharing Between Workspaces Access Delta Shares From your NorthCentral Workspace, go to Catalog. Hover over Delta Shares Received. When the icon appears, click it. → This will redirect you to the Delta Sharing page. Create a New Recipient On the Delta Sharing page, click Shared by me. Click New Recipient. Fill in the details: Recipient Name: (Enter your recipient name) Recipient Type: Select Databricks Sharing Identifier: azure:southcentralus:3035j6je88e8-91-434a-9aca-e6da87c1e882 To get the sharing identifier using a notebook or Databricks SQL query: (SQL) SELECT CURRENT_METASTORE(); Click Create. Share Data Click "Share Data". Enter a Share Name. Select the data assets you want to share. Note: Please disable History for the selected data assets, as the current data snapshot. Disabling the History option on the Delta Share will simplify the share and prevent unnecessary access to historical versions. Additionally, review whether you can further simplify your share by partitioning the data where appropriate. Add the recipient's name you created earlier. Click Share Data. Add Recipient From the newly created share, click Add Recipient. Select your South-Central Workspace Metastore ID. South-CentralWorkspace In your South-Central Workspace, navigate to the Delta Sharing page. Under Shared with me tab, locate your newly created share and click on it. Add the share to a catalog in Unity Catalog. Part 2: Enable NCC for Serverless Warehouse 6. Add Network Connectivity Configuration (NCC) Go to the Databricks Account Console: https://accounts.azuredatabricks.net/ Navigate to Cloud resources, click Add Network Connectivity Configuration. Fill in the required fields and create a new NCC for SouthCentral. 7. Associate NCC with Workspace In the Account Console, go to Workspaces. Select your SouthCentral workspace, click Update Workspace. From the Network Connectivity Configuration dropdown, select the NCC you just created. 8. Add Private Endpoint Rule In Cloud resources, select your NCC, select Private Endpoint Rules and click Add Private Endpoint Rule. Provide: Resource ID: Enter your Storage Account Resource ID in NorthCentral. Note: This can be found in your storage account (NorthCentral). Click on “JSON View” top right. Azure Subresource type: dfs & blob. 9. Approve Pending Connection Go to your NorthCentral Storage Account, Networking, Private Endpoints. You will see a Pending connection from Databricks. Approve the connection and you will see the Connection status in your Account Console as ESTABLISHED. You will now see your share listed under “Delta Shares Received” Note: If you cannot view your share, run the following SQL command: GRANT USE_PROVIDER ON METASTORE TO `username@xxxx.com`.SAP Business Data Cloud Connect with Azure Databricks is now generally available
We are excited to share that SAP Business Data Cloud (SAP BDC) Connect for Azure Databricks is generally available. With this announcement, Azure Databricks customers like you, can connect your SAP BDC environment to your existing Azure Databricks instance – without copying the data – to enable bi-directional, live data sharing. Connecting SAP data with other enterprise data prevents governance risk, compliance gaps, and data silos. In addition, maintenance costs are also reduced and manual building of semantics is no longer needed. SAP data products can now be shared directly via Delta Sharing into your existing Azure Databricks instances ensuring complete context for your business. You can now unify your data estate across Azure Databricks and SAP BDC This makes it easier for you to: Enforce governance Power analytics, data warehousing, BI and AI Connecting SAP BDC to Azure Databricks is simple, secure, and fast. The connection is trusted and requires approval from both platforms to enable bi-directional sharing of data products. Once approved, data products in SAP BDC can be directly mounted into Azure Databricks Unity Catalog and are treated like other assets shared using Delta sharing. As a result, your teams can query, analyze, and gather insights on SAP data in addition to your existing business data in one unified way. Instead of spending time gathering the data in once place, your teams can instead focus on unlocking insights from this unified data quickly and securely. This launch complements SAP Databricks in SAP BDC running on Azure that enables AI, ML, data engineering, and data warehousing capabilities directly inside your SAP environment. We have expanded the list of supported regions for SAP Databricks on SAP BDC running on Azure. To learn more with SAP BDC Connect with Azure Databricks review documentation and get started today.How Azure NetApp Files Object REST API powers Azure and ISV Data and AI services – on YOUR data
This article introduces the Azure NetApp Files Object REST API, a transformative solution for enterprises seeking seamless, real-time integration between their data and Azure's advanced analytics and AI services. By enabling direct, secure access to enterprise data—without costly transfers or duplication—the Object REST API accelerates innovation, streamlines workflows, and enhances operational efficiency. With S3-compatible object storage support, it empowers organizations to make faster, data-driven decisions while maintaining compliance and data security. Discover how this new capability unlocks business potential and drives a new era of productivity in the cloud.
Secure Medallion Architecture Pattern on Azure Databricks (Part I)
This article presents a security-first pattern for Azure Databricks: a Medallion Architecture where Bronze, Silver and Gold each run as their Lakeflow Job and cluster, orchestrated by a parent job. Run-as identities are Microsoft Entra service principals; storage access is governed via Unity Catalog External Locations backed by the Access Connector’s managed identity. Least-privilege is enforced with cluster policies and UC grants. Prefer managed tables to unlock Predictive Optimisation, Automatic liquid clustering and Automatic statistics. Secrets live in Azure Key Vault and are read at runtime. Monitor reliability and cost with system tables and Jobs UI. Part II covers more low-level concepts and CI/CD.Securing Azure Databricks Serverless: Practical Guide to Private Link Integration
The Challenge: Outbound Control in a Serverless World Serverless compute resources run in the serverless compute plane, and is managed by Microsoft, for ease of use. Databricks serverless provides hassle-free compute for running notebooks, jobs, and pipelines and by default, outbound traffic can reach the internet and other networks freely. One of the most common security requirements for customers in the financial and government sectors is to have the ability to retain network paths within a private network for their users to access their data and for system integration. Solution Objective Enforce deny-by-default posture: Control outbound access with granular precision by enabling a deny-by-default policy for internet. By default, all outbound access is blocked unless explicitly allowed via Private Endpoint Rules. Control outbound connections: by specifying allowed locations, connections, FQDN Enforce the traffic to go over the customer network for traffic controls and inspection Solution Overview The solution is designed to route Databricks Serverless outbound traffic to the customer customer-managed Policy Enforcement Point (e.g. Azure Firewall), to allow the customer to securely connect to services hosted on the cloud without exposing the data to the public internet. Essentially, it establishes a private, secure connection between Databricks Control Plane to the customer virtual network. Pre-Requisites Azure Firewall Deploy an Azure Firewall if you don’t already have one. Virtual Networks and Subnets Create a VNET for Databricks and Load Balancer deployment. Set up subnets for the Azure Standard Load Balancer frontend (e.g., 10.0.2.0/26) and backend (e.g., 10.0.2.64/26). Enable Private Endpoint network policy for Network Security Groups and Route Tables on the backend subnet. VNET Peering Peer the Databricks VNET with your hub VNET to allow secure routing. Azure Databricks Workspace Deploy an Azure Databricks workspace if you don’t have one. Follow the official Azure Databricks documentation for detailed steps on creating workspaces and private endpoints Summary of Steps Deploy Azure Firewall and Networking Set up an Azure Firewall and create the necessary virtual networks (VNets) and subnets for your environment. Peer the Databricks VNet with your hub VNet to enable secure routing. Configure the Azure Load Balancer Create an internal Standard Load Balancer. Set up frontend and backend pools using NICs (not IP addresses). Add load balancing rules and configure a health probe (typically HTTP on port 8082). Create a Private Link Service Deploy the Private Link Service behind the load balancer. Associate it with the correct frontend and backend subnets. Set Up Route Tables Create route tables to direct backend VM traffic to the Azure Firewall. Ensure the route tables are associated with the correct subnets (e.g., backend subnet for the router VM). Deploy and Configure the Router VM Deploy a Linux VM to act as a router. Enable IP forwarding on the VM and in Azure settings. Configure IPTables for NAT and traffic forwarding. Install and configure NGINX to serve as a health probe for the load balancer. Configure Network Security Groups (NSGs) Set up NSGs to allow necessary traffic (SSH, load balancer, HTTP/HTTPS, health probe) to and from the router VM. Configure Azure Firewall Application Rules Define application rules to allow outbound access only to approved FQDNs (e.g., microsoft.com). Block all other outbound traffic by default. Configure Databricks Account Portal Enable outbound (serverless) Azure Private Link to customer-managed resources in the Databricks Account Portal. Create Network Connectivity Configurations (NCCs) and attach them to your workspaces. Add private endpoint rules for each Azure resource you want to allow access to. Approve Private Endpoints In the Azure Portal, approve the private endpoint connections created by Databricks for your resources. Troubleshooting Use tools like netstat, conntrack, and tcpdump on the router VM to diagnose routing issues. Double-check route table and NSG associations. Validate private endpoint rule configurations in both Databricks and the Azure Portal. References Serverless compute plane networking - Azure Databricks | Microsoft Learn Configure private connectivity to Azure resources - Azure Databricks | Microsoft Learn Key Takeaway This solution enforces a deny-by-default posture for outbound traffic from Azure Databricks Serverless, only allowing explicitly approved connections via Private Endpoints. All traffic is routed through your network for inspection and control, helping you meet strict compliance and security requirements. Ready to Get Started? Securing your Databricks Serverless environment doesn’t have to be daunting. With Azure Private Link, Azure Firewall, and a smart configuration, you get the best of both worlds: agility and airtight security. For more details, check out the official Azure Databricks documentation and start building your secure analytics platform today. Questions or want to share your experience? Drop a comment below or reach out to the Azure Databricks community.Approaches to Integrating Azure Databricks with Microsoft Fabric: The Better Together Story!
Azure Databricks and Microsoft Fabric can be combined to create a unified and scalable analytics ecosystem. This document outlines eight distinct integration approaches, each accompanied by step-by-step implementation guidance and key design considerations. These methods are not prescriptive—your cloud architecture team can choose the integration strategy that best aligns with your organization’s governance model, workload requirements and platform preferences. Whether you prioritize centralized orchestration, direct data access, or seamless reporting, the flexibility of these options allows you to tailor the solution to your specific needs.General Availability: Automatic Identity Management (AIM) for Entra ID on Azure Databricks
In February, we announced that Automatic Identity Management in public preview and loved to hear your overwhelmingly positive feedback. Prior to public preview, you either had to set up an Entra Enterprise Application or involve an Azure Databricks account admin to import the appropriate groups. This required manual steps whether it was adding or removing users with organizational changes, maintaining scripts, or requiring additional Entra or SCIM configuration. Identity management was thus cumbersome and required management overhead. Today, we are excited to announce that Automatic Identity management (AIM) for Entra ID on Azure Databricks is generally available. This means no manual user setup is needed and you can instantly add users to your workspace(s). Users, groups, and service principals from Microsoft Entra ID are automatically available within Azure Databricks, including support for nested groups and dashboards. This native integration is one of the many reasons Databricks runs best on Azure. Here are some addition ways AIM could benefit you and your organization: Seamlessly share dashboards You can share AI/BI dashboards with any user, service principal, or group in Microsoft Entra ID immediately as these users are automatically added to the Azure Databricks account upon login. Members of Microsoft Entra ID who do not have access to the workspace are granted access to a view-only copy of a dashboard published with embedded credentials. This enables you to share dashboards with users outside your organization, too. To learn more, see share a dashboard. Updated defaults for new accounts All new Azure Databricks accounts have AIM enabled – no opt in or additional configuration required. For existing accounts, you can enable AIM with a single click in the Account Admin Console. Soon, we will also make this the default for existing accounts. Automation at scale enabled via APIs You can also register users, groups, or service principles in Microsoft Entra ID via APIs. Being able to do this programmatically enables the enterprise scale most of our customers need. You can also enable automation via scripts leveraging these APIs. Read the Databricks blog here and get started via documentation today!Part 2: Performance Configurations for Connecting PBI to a Private Link ADB Workspace
This blog was written in conjunction with Leo Furlong, Lead Solutions Architect at Databricks. In Part 1, we discussed networking options for connecting Power BI to an Azure Databricks workspace with a Public Endpoint protected with a workspace IP Access List. In Part 2, we continue our discussion and elaborate on private networking options for an Azure Databricks Private Link workspace. When using Azure Databricks Private Link with Allow Public Network Access setting set to Disabled, all connections to the workspace must go through Private Endpoints. For one of the private networking options, we’ll also discuss how to configure your On-Premise Data Gateway VM to get good performance. Connecting Power BI to a Private Link Azure Databricks Workspaces As covered in Part 1, Power BI offers two primary methods for secure connections to data sources with private networking: 1. On-premises data gateway: An application that gets installed on a Virtual Machine that has a direct networking connection to a data source. It allows Power BI to connect to data sources that don’t allow public connections. The general flow of this setup entails: a. Create or leverage a set of Private Endpoints to the Azure Databricks workspace - both sub-resources for databricks_ui_api and browser_authentication are required b. Create or leverage a Private DNS Zone for privatelink.azuredatabricks.net c. Deploy an Azure VM into a VNet/subnet d. The VM’s VNet/subnet should have access to the Private Endpoints (PEs) via either them being in the same VNet or being peered with another VNet where they reside e. Install and configure the on-premise data gateway software on the VM f. Create a connection in the Power BI Service via Settings -> Manage Connections and Gateways UIs g. Configure the Semantic Model to use the connection under the Semantic Model’s settings and gateway and cloud connections sub-section 2. Virtual Network Data Gateway: A fully managed data gateway that gets created and managed by the Power BI service. Connections work by allowing Power BI to delegate into a VNet for secure connectivity to the data source. The general flow of this setup entails: a. Create or leverage a set of Private Endpoints (PEs) to the Azure Databricks workspace - both sub-resources for databricks_ui_api and browser_authentication are required b. Create or leverage a Private DNS Zone for privatelink.azuredatabricks.net c. Create a subnet in a VNet that has access to the Private Endpoints (PEs) via either them being in the same VNet or being peered with another VNet where they reside. Delegate the Subnet to Microsoft.PowerPlatform/vnetaccesslinks d. Create a virtual network data gateway in the Power BI Service via Settings -> Manage Connections and Gateways UIs e. Configure the the Semantic Model to use the connection under the Semantic Model’s settings and gateway and cloud connections sub-section The documentation for both options is fairly extensive, and this blog post will not focus on breaking down the configurations further. Instead, this post is about configuring your private connections to get the best Import performance. On-Premise Data Gateway Performance Testing In order to provide configuration guidance, a series of Power BI Import tests were performed using various configurations and a testing dataset. Testing Data The testing dataset used was a TPC-DS scale factor 10 dataset (you can create your own using this Repo). A scale factor of 10 in TPC-DS generates about 10 gigabytes (GB) of data. The TPC-DS dataset was loaded into Unity Catalog and the primary and foreign keys were created between the tables. A model was then created in the Power BI Service using the Publish to Power BI capabilities in Unity Catalog; the primary and foreign keys were used to automatically create relationships between the tables in the Power BI semantic model. Here’s an overview of the tables used in this dataset: Fabric Capacity An F64 Fabric Capacity was used in the West US region. The F64 was the smallest size available (in terms of RAM) for refreshing the model without getting capacity errors - the compressed Semantic Model size is 5,244 MB. Azure Databricks SQL Warehouse An Azure Databricks workspace using Unity Catalog was deployed in the East US 2 and West US regions for the performance tests. A Medium Databricks SQL Warehouse was used. For Imports, generally speaking, the size of the SQL Warehouse isn’t very important. Using an aggressive Auto Stop configuration of 5 minutes is ideal to minimize compute charges (1 minute can be used if the SQL Warehouse is deployed via an API). Testing Architecture The following diagram summarizes a simplified Azure networking architecture for the performance tests. A Power BI Semantic Model is connected to a Power BI On-Premise Data Gateway Connection The On-Premise Data Gateway Connection connects to the Azure Databricks workspace using Private Endpoints. Azure Databricks provisions up a Serverless SQL Warehouse in ~5 seconds within the Serverless Data Plane within Azure. SQL queries are executed on the Serverless SQL Warehouse. Unity Catalog gives the Serverless SQL Warehouse a read-only, down-scoped, and pre-signed URL to ADLS. Data is fetched from ADLS and placed on the Azure Databricks workspace’s managed storage account via a capability called Cloud Fetch. Arrow Files are pulled from Cloud Fetch and delivered to the Power BI Service through the Data Gateway. Data in the Semanic Model is compressed and stored in Vertipaq In-Memory storage. Testing Results The following grid outlines the scenarios tested and the results for each test. We’ll review the different configurations tested below in specific sections. Scenario Gateway Scenario Avg Refresh Duration Minutes A East US 2, Public Endpoint 17:01 B West US, Public Endpoint 12:21 C West US, Public Endpoint via IP Access List 15:19 D West US, E VM Gateway Base 12:14 E West US, E VM StreamBeforeRequestCompletes 07:46 F West US, E VM StreamBeforeRequestCompletes + Logical Partitions 07:31 G West US, E VM Spooler (D) 12:57 H West US, E VM Spooler (E) 13:32 I West US, D VM Gateway Base 16:47 J West US, D VM StreamBeforeRequestCompletes 12:19 K West US, PBI Managed Vnet 27:04 Scenario VM Configuration D Standard E8bds v5 (8 vcpus, 64 GiB memory) [NVMe, Accelerated Networking], C Drive default (Premium SSD LRS 127 GiB) E Standard E8bds v5 (8 vcpus, 64 GiB memory) [NVMe, Accelerated Networking], C Drive default (Premium SSD LRS 127 GiB) F Standard E8bds v5 (8 vcpus, 64 GiB memory) [NVMe, Accelerated Networking], C Drive default (Premium SSD LRS 127 GiB) G Standard E8bds v5 (8 vcpus, 64 GiB memory) [NVMe, Accelerated Networking], D drive H Standard E8bds v5 (8 vcpus, 64 GiB memory) [NVMe, Accelerated Networking], E Drive (Premium SSD LRS 600 GiB) I Standard D8s v3 (8 vcpus, 32 GiB memory), C Drive default (Premium SSD LRS 127 GiB) J Standard D8s v3 (8 vcpus, 32 GiB memory), C Drive default (Premium SSD LRS 127 GiB) Performance Configurations 1. Regional Alignment Aligning your Power BI Premium/Fabric Capacity to the same region as your Azure Databricks deployment and your On-Premise Data Gateway VM helps reduce the overall network latency and data transfer duration. It should also eliminate cross-region networking charges. In scenario A, the Azure Databricks deployment was in East US 2 while the Fabric Capacity and On-Premise Data Gateway VM were in West US. The Import processing time when using the public endpoint between the regions was 17:01 minutes. In scenario B, while still using the public endpoint, there is complete regional alignment in the West US region and the Import times averaged 12:21 minutes which is a 27.4% decrease 2. Configure a Gateway Cluster A Power BI Data Gateway Cluster configuration is highly recommended for Prouduction deployments but this configuration was not performance tested during this experiment. Data Gateway clusters can help with data refresh redundancy and for overall volume / throughput of data transfer. This configuration is highly recommended for Production Power BI environments. 3. VM Family Selection The Power BI documentation recommends a VM with 8 cores, 8 GB of RAM, and an SSD for the VM used for the On-Premise Data Gateway. Through testing, it can be proven that using a VM with good performance characteristics can provide immense value in the Import times. In scenario D, data gateway tests were run using a Standard E8bds v5 with 8 cores and 64 GB RAM that also included NVMe, and Accelerated Networking, and a C drive using a Premium SSD. The import times for this scenario averaged 12:14 minutes which was slightly faster than the regionally aligned public endpoint test in scenario B. In scenario I, data gateway tests were run using a Standard D8s v3 with 8 cores and 32 GB RAM and a C drive using a Premium SSD. The import times for this scenario averaged 16:47 minutes which was noticeably slower than using the regionally aligned public endpoint in cenario B which was a 35.96% performance degradation. More tests could certainly be done to determine which VM characteristics help the most with Import performance, but it is clear certain features can be helpful like: Premium SSDs Accelerated Networking NVMe controller Memory optimized instances And while the better E8bds v5 Azure VM costs ~$820 per month in West US at list and the D8s v3 costs ~$610 per month at list (25% more expensive), this feels like a scenario where you pay the premium to get better performance and optimize through Azure VM reservations. 4. StreamBeforeRequestCompletes By default, the on-premise data gateway spools data to disk before sending it to Power BI. Enabling the StreamBeforeRequestCompletes setting to True can significantly improve gateway refresh performance as it allows data to be streamed directly to the Power BI Service without first being spooled to disk. In scenario E, when StreamBeforeRequestCompletes is set to True and restarted, you can see that the average Import times significantly improved to 07:46 minutes which is a 54% improvement compared to scenario A and a 36% improvement over the base VM configuration in scenario D. 5. Spooler Location As discussed above, when using the default setting for StreamBeforeRequestCompletes as False, Power BI spools the data to the data gateway spool directory before sending it to the Power BI Service. In scenarios D, G, and H, StreamBeforeRequestCompletes is False and the Spooler directory has been mapped to the C drive, D drive, and E drives respectively which all correspond to an SSD (of varying configuration) on the Azure VMs. In all scenarios, you can see the times are similar between 12:14, 12:57, and 13:32 minutes, respectively. In all three scenarios the tests were performed with SSDs on the E series VM configured with NVMe. Using this configuration mix, it doesn’t appear that the Spooler directory location provides significant performance improvements. Since the C drive configuration gave the best performance it seems prudent to keep the C drive default configuration. However, it is possible that that the Spooler directory setting might provide more value on a different VM configurations. 6. Logical Partitioning As outlined in the QuickStart samples guide, logical partitioning can often help with Power BI Import performance as multiple logical partitions in the Semantic Model can be processed at the same time. In scenario F, logical partitions were created for the inventory and store_sales table to have 5 partitions each. When combined with the StreamBeforeRequestCompletes setting, the benefit from adding Logical Partitions was negligible (15 second improvement) even though the parallelization settings were increased to 30 (Max Parallelism Per Refresh and Data Source Default Max Connections). While logical partitions are usually a very valuable strategy, combining them with StreamBeforeRequestCompletes, the E series VM configurations, and a Fabric F64 capacity yielded diminishing returns. It is probably worth more testing at some point in the future. Virtual Network Data Gateway Performance Testing The configuration and performance of a Virtual Network Data Gateway was briefly tested. A Power BI subnet was created in the same VNet as the Azure Databricks workspace and delegated to the Power BI Service. A virtual network data gateway was created in the UI with 2 gateways (12 queries can run in parallel) and assigned to the Semantic Model. In scenario K, an Import test was performed through the Virtual Network Data Gateway that took 27:04 minutes. More time was not spent trying to tune the Virtual Network Data Gateway as it was not the primary focus of this blog post. The Best Configuration The Best Configuration: Region Alignment + Good VM + StreamBeforeRequestsCompletes While the Import testing performed for this blog post isn’t definitive, it does provide good directional value in forming an opinion on how you can configure your Power BI On-Premise Data Gateway on an Azure Virtual Machine to get good performance. When looking at the tests performed for this blog, an Azure Virtual Machine, in the same region as the Azure Databricks Workspace and the Fabric Capacity, with Accelerated networking, an SSD, NVMe, and memory optimized compute provided performance that was faster than just using the public endpoint of the Azure Databricks Workspace alone. Using this configuration, we improved our Import performance from 17:01 to 07:46 minutes which is a 54% performance improvement.
