Microsoft SQL Server 2025 on ThinkAgile VX V4 and FX V4

Enterprise organizations manage the full spectrum of data workloads, from high-frequency financial transactions and mission-critical ERP systems to real-time analytics and AI-powered applications. The relational database remains the most trusted foundation for structured, transactional data at enterprise scale, and Microsoft SQL Server 2025 represents a significant evolution of that foundation.

SQL Server 2025 introduces native AI integration, vector data types, intelligent query processing enhancements, and seamless connectivity to Microsoft Fabric, alongside the reliability, security, and high-availability capabilities the platform is known for. For organizations modernizing their data infrastructure without sacrificing governance or performance, SQL Server 2025 offers a compelling path forward.

Lenovo ThinkAgile VX V4 and FX V4 hyperconverged systems, powered by Intel Xeon 6 processors and certified with VMware software, provide the infrastructure foundation for SQL Server 2025 deployments. By converging compute, storage, and virtualization into a single validated platform, ThinkAgile simplifies deployment and operations while delivering the density, performance, and resilience that mission-critical workloads demand. In Lenovo validated testing, a single ThinkAgile VX650 V4 node sustained 9.55 million transactions per minute running SQL Server 2025 Enterprise, with throughput scaling as nodes are added to the cluster.

Figure 1. ThinkAgile VX650 V4 (top) and VX630 V4 (bottom)

CIOs and enterprise IT leaders face converging pressures that make the status quo unsustainable. Infrastructure deployed three to five years ago was not designed for today's data volumes, workload mix, or AI integration requirements, and the gap between business demand and infrastructure capability continues to widen.

• Data Growth and Performance

  ERP, CRM, financial, and IoT workloads generate ever larger datasets. Legacy storage and compute tiers struggle to sustain the required IOPS, throughput, and latency, resulting in degraded performance and missed SLAs.

• AI Integration

  Business intelligence and AI initiatives increasingly require semantic search, recommendations, and retrieval augmented generation alongside traditional OLTP. Bolting separate AI infrastructure onto a relational estate creates integration complexity, data duplication, and governance gaps.

• Complexity, Availability, and Cost

  Three-tier architectures carry separate lifecycles and failure modes at every layer, DRAM and licensing costs continue to climb, and many organizations still protect revenue-critical databases with manually managed high-availability configurations.

• A Closing Modernization Window

  SQL Server 2016 reaches end of extended support on July 14, 2026, and SQL Server 2019 extended support ends January 8, 2030. Because modernization typically takes 12 to 18 months, the window to plan and migrate is closing, and Broadcom's shift away from perpetual VMware licensing adds further urgency to platform decisions.

The Lenovo ThinkAgile VX V4 and FX V4 hyperconverged systems, running SQL Server 2025 on VMware Cloud Foundation, directly address each of these challenges. The platform consolidates compute, storage, and virtualization into a single, factory-integrated stack that simplifies procurement, accelerates deployment, and delivers consistent, predictable performance at scale.

Each ThinkAgile VX650 V4 node houses dual Intel Xeon 6730P processors delivering 32 cores per socket, paired with 1 TB of DDR5 6400MHz memory as tested; the platform supports up to 8 TB of memory per node for workloads that require it. NVMe PCIe Gen 5 SSDs provide high local storage throughput, and Mellanox 25GbE networking connects nodes in a high-bandwidth cluster fabric.

VMware vSphere provides the virtualization layer, and VCF Data Service Manager enables unified provisioning, cloning, and administration of SQL Server databases across the cluster. An optional NVMe tiering drive helps organizations balance CPU and memory usage and extend the value of existing infrastructure without impacting application performance.

ThinkAgile FX V4 extends this foundation with software-defined flexibility, letting organizations change hypervisors without replacing hardware and protecting infrastructure investment as strategies evolve. Complete hardware, software, and workload configuration details for the validated system, together with the measured OLTP performance results, are provided in the Test Methodology and Test Results sections of this brief.

Figure 2. ThinkAgile FX650 V4 (top) and FX630 V4 (bottom) designed for flexible hyperconverged infrastructure

SQL Server 2025 Capabilities

Microsoft SQL Server 2025 pairs this infrastructure with capabilities that extend the platform beyond traditional OLTP:

• Native Vector Data Type and Search

  Vector functions and approximate vector indexing enable semantic search, recommendations, and retrieval augmented generation directly inside the database engine, without a separate vector store.

• Built-in AI Model Integration

  External AI model objects and REST endpoint invocation let T-SQL workloads call Azure OpenAI and other AI services natively, enabling embedding generation and intelligent query augmentation.

• Enhanced Intelligent Query Processing

  Cardinality estimation feedback, optional parameter plan optimization, and degree of parallelism feedback, enabled by default, reduce query plan instability and improve throughput for mixed OLTP and analytics workloads.

• Optimized Locking for High-Concurrency OLTP

  Reduced lock memory consumption, blocking, and lock escalation directly improve concurrency and throughput for transaction-heavy ERP, CRM, and financial applications.

• Native JSON and Microsoft Fabric Mirroring

  A native JSON data type with indexing simplifies document-style data management, and Fabric Mirroring continuously replicates on-premises data into Microsoft Fabric for real-time analytics without ETL pipelines.

• Stronger Security and Availability

  TLS 1.3 with TDS 8.0, PBKDF2 password hashing, Microsoft Entra integration, and improved Always On failover behavior deliver a modern, standards-aligned security and resiliency posture.

SQL Server 2025 on Lenovo ThinkAgile VX V4 and FX V4 addresses four recurring workload patterns on one validated foundation, allowing organizations to consolidate rather than build a separate environment for each. The following table summarizes each pattern, the SQL Server 2025 and ThinkAgile capabilities that enable it, and the resulting business benefit.

SQL Server 2025 on Lenovo ThinkAgile VX V4 and FX V4 is built to run the full range of workloads that make up a modern data estate on one validated platform, rather than requiring a separate environment for each:

• Mission-Critical OLTP Workloads

  High-concurrency transactional systems, such as ERP, CRM, financial, and order management applications, that demand predictable response times at peak load.

• Analytics and AI-Augmented Workloads

  Real-time reporting, semantic search, and retrieval augmented generation built directly on governed operational data, without a separate analytics or vector database.

• Consolidation Workloads

  Multiple SQL Server instances and database workloads unified onto a single HCI stack, reducing rack footprint, platform sprawl, and the number of environments IT must secure and support.

• Resilience and Continuity Workloads

  Databases that cannot tolerate extended downtime, protected by SQL Server 2025 Always On Availability Groups and VMware vSAN built-in data protection.

Complete hardware configuration, test methodology, and measured performance for these workload profiles are provided in the Test Methodology and Test Results sections of this brief.

These workload patterns take different forms in each industry. Regulated sectors weigh data residency, auditability, and control over sensitive data most heavily, while operationally intensive industries prioritize continuity, scale, and analytics close to the point of decision. Because ThinkAgile VX V4 and FX V4 keep data and governance on-premises while still supporting AI and analytics integration, the platform maps well to sectors where public cloud deployment is constrained. The following table outlines the primary requirements, platform benefits, and representative applications for key verticals.

Organizations deploying SQL Server 2025 on Lenovo ThinkAgile VX V4 and FX V4 realize measurable improvements across performance, operational efficiency, total cost of ownership, and long-term infrastructure resilience:

• Performance at Scale

  Validated HammerDB TPC-C style testing demonstrated 9.55 million transactions per minute across twelve concurrent virtual machines on a single ThinkAgile VX650 V4 node, at 75% host CPU utilization and 375K average IOPS. This provides a validated performance baseline for sizing, capacity planning, and workload consolidation assessments.

• Reduced Infrastructure Complexity

  Consolidating compute, storage, and virtualization into a single HCI stack eliminates the operational complexity of separate infrastructure layers. VCF Data Service Manager provides unified database lifecycle management, replacing multiple point tools for provisioning, patching, cloning, and backup, freeing IT teams for higher-value initiatives.

• Lower Total Cost of Ownership

  High-density ThinkAgile VX V4 nodes reduce rack space, power, and cooling requirements compared to equivalent three-tier configurations. Consolidating multiple database workloads reduces per-workload licensing and support costs, and NVMe Memory Tiering can further raise consolidation density where memory cost, rather than raw throughput, is the binding constraint.

• Faster Time-to-Value

  Factory-integrated and pre-validated ThinkAgile systems eliminate lengthy procurement and integration cycles. SQL Server 2025 can be operational within days of hardware delivery, and VCF Data Service Manager provisioning and cloning accelerates application development and deployment cycles, reducing time-to-production from weeks to hours.

• Enterprise Resilience and Business Continuity

  VMware vSAN built-in data protection, SQL Server 2025 Always On Availability Groups, and VCF Data Service Manager automated backup deliver deterministic, low-RTO recovery without manual intervention. Organizations can meet aggressive RTO and RPO commitments while eliminating the complexity and risk of manually maintained DR configurations.

• Future-Proof Investment

  ThinkAgile FX V4 software-defined flexibility allows hypervisor transitions without hardware replacement. SQL Server 2025 hybrid cloud integration with Azure and Microsoft Fabric preserves optionality for future cloud or hybrid deployments, and Intel Xeon 6 architecture provides a high-core-count, high-memory foundation capable of supporting growth for five or more years.

HammerDB is an open source database benchmarking tool that includes an OLTP workload derived from the TPC-C benchmark specification. It simulates a high-concurrency order-entry system with multiple warehouse districts, customers, orders, and stock transactions running simultaneously. HammerDB was deployed on a dedicated external server to generate load against the SQL Server virtual machines running on the ThinkAgile VX650 V4 cluster, so the load generator itself did not consume cluster resources.

Each node in the validated cluster used the configuration in the following table.

Lenovo validated SQL Server 2025 OLTP performance on a four-node Lenovo ThinkAgile VX650 V4 cluster using HammerDB, an open source benchmarking tool with an OLTP workload modeled on the TPC-C benchmark standard. Testing characterized the throughput achievable on a single node running multiple concurrent database virtual machines across three sizing scenarios. Results are not comparable to published TPC-C results and do not comply with TPC-C benchmark standards; actual results will vary based on workload, configuration, and environment.

• SQL Server 2025 VM and Workload Configuration
• Performance Results
• Key Observations

SQL Server 2025 VM and Workload Configuration

Testing used three virtual machine sizing profiles against a single node, each running Microsoft SQL Server 2025 Enterprise; per-scenario VM sizing and results appear in the Performance Results table below. The following parameters were shared across all three scenarios:

Performance Results

The following results were recorded after a 15 minute ramp up period, during steady state operation, across the three virtual machine sizing scenarios tested individually on a single ThinkAgile VX650 V4 node, with and without NVMe Memory Tiering enabled:

Across all three scenarios, the node sustained high transaction throughput while host CPU utilization stayed at or below 76%, leaving headroom for virtualization overhead and VMware vSAN. NVMe Memory Tiering delivered transaction throughput within an acceptable range of DRAM-only performance, with the performance tradeoff growing from approximately 10 percent at the lowest tested memory footprint to more than 20 percent at the highest.

Key Observations

Beyond the throughput and latency figures reported above, the test results point to several practical conclusions for planning a production deployment. The observations below translate the measured data into guidance on compute headroom, scale-out expectations, and the tradeoffs between DRAM-only and NVMe Memory Tiering configurations.

• Storage Is Not the Bottleneck

  375K average IOPS at a 75/25 read write ratio across eight PCIe Gen 5 NVMe drives per node leaves significant storage headroom. Workloads with a higher write ratio or larger database footprints can be accommodated without storage re-architecture.

• Efficient Resource Utilization

  The validated configuration reached 55% to 76% host CPU utilization across the DRAM-only scenarios, reflecting effective use of available compute resources while preserving capacity for virtualization overhead, VMware vSAN, transient workload spikes, and routine database maintenance. Production environments typically benefit from maintaining reserve CPU capacity rather than operating continuously at maximum utilization, and organizations consolidating multiple database servers can use these results as a baseline for modeling consolidation ratios against their own headroom and node-failure reserve policies.

• Linear Scale-Out with VMware vSAN

  Testing characterized single-node throughput under conservative, worst-case conditions. In a production deployment, VMware vSAN distributes workloads across nodes, allowing throughput to scale approximately linearly as nodes are added, so a four-node cluster is expected to sustain roughly four times the single-node throughput measured here.

• SQL Server 2025 Optimized Locking Reduces Contention

  SQL Server 2025 optimized locking reduces lock memory consumption and blocking under high-concurrency OLTP. This is reflected in throughput that scaled consistently across all three tested scenarios, from 35 to 70 concurrent virtual users, without significant lock-wait degradation at peak load.

• NVMe Memory Tiering Trades Some Throughput for Higher Consolidation

  Across the tested scenarios, active memory footprints ranged from 250 GB to 500 GB. NVMe Memory Tiering enabled roughly 1.5 to 2 times higher consolidation density than DRAM-only sizing, with transaction throughput and New Orders Per Minute remaining within an acceptable range of DRAM-only performance, an option worth evaluating wherever memory cost, rather than raw throughput, is the binding constraint.

The following configuration settings were applied during testing and are recommended for production SQL Server 2025 deployments on Lenovo ThinkAgile VX V4 and FX V4 systems:

The following bill of materials lists the components used in the validated Lenovo ThinkAgile VX650 V4 configuration. This represents a single-node BOM. A four-node cluster requires four units of each server-level component. Part numbers and feature codes are provided for reference and procurement purposes. Contact your Lenovo representative or authorized Lenovo Business Partner for current pricing, availability, and configuration guidance.

For ThinkAgile FX V4 configurations or alternative node sizes, contact your Lenovo representative. Full configuration guides and additional BOM options are available at lenovopress.lenovo.com.

Lenovo ThinkAgile VX V4 and FX V4 with SQL Server 2025 give IT leaders a single, validated, factory-integrated stack that handles today's mission-critical databases, tomorrow's AI workloads, and the compliance demands of the most regulated industries. With 9.55 million transactions per minute proven on a single node, NVMe Memory Tiering raising consolidation density without sacrificing throughput, and vSAN resilience built in, the platform is ready to run and scale from day one.

To learn more about Microsoft SQL Server 2025 on Lenovo ThinkAgile VX V4 and FX V4, contact your Lenovo representative or Lenovo Business Partner, or visit the resources below.

References:

• Solution Brief - Modernize Microsoft SQL Server 2025 with Lenovo:
  https://lenovopress.lenovo.com/lp2480
• Lenovo ThinkAgile VX Series Datasheet:
  https://lenovopress.lenovo.com/datasheet/ds0104-thinkagile-vx-series
• Lenovo ThinkAgile VX650 V4 Product Guide:
  https://lenovopress.lenovo.com/lp2135-lenovo-thinkagile-vx650-v4-hyperconverged-system
• Lenovo ThinkAgile FX650 V4 Product Guide:
  https://lenovopress.lenovo.com/lp2338-lenovo-thinkagile-fx650-v4-hyperconverged-system
• Multi-Vendor Hyperconverged Infrastructure with Lenovo ThinkAgile FX V4:
  https://lenovopress.lenovo.com/lp2520-multi-vendor-hyperconverged-infrastructure-with-lenovo-thinkagile-fx-v4
• VMware Data Service Manager:
  https://www.vmware.com/products/cloud-infrastructure/advanced-services/data-services-manager
• Optimizing VCF Private Cloud Economics with NVMe Memory Tiering on ThinkAgile VX:
  https://lenovopress.lenovo.com/lp2518-optimizing-vcf-private-cloud-economics-with-nvme-memory-tiering-on-thinkagile-vx

Chandrakandh Mouleeswaran is a Solution Architect with 18+ years of experience in software development, performance testing and engineering, having worked on designing and architecting many scalable enterprise applications. He has spent a decade in technical enablement and partner solution development for VMware, Nutanix, Oracle and other ISVs across industries and technologies. He specializes in architecting infrastructure solutions for virtualization, VDI, database, cloud, data science, AI/ML solutions and various enterprise workloads.

Chris Honoré is a Solutions Product Manager at Lenovo with deep expertise in datacenter products and solution offerings. He has a strong background in consulting and solution development, helping customers design and support on-premises and hybrid environments. Chris has spent the past 15 years with IBM and Lenovo, specializing in x86 server and data center solutions. Prior to that, he built two decades of experience in the telecommunications industry, serving in both technical and business leadership roles.