VMmark User Guide
After their introduction in 2017, AMD EPYC processors have become the preferred option for both public and private cloud deployments. High core counts, the fastest x86 memory bandwidth in the market, and cutting-edge security features make 4th generation AMD EPYC processors ideal for demanding private cloud workloads. AMD EPYC processors offer a strong platform for building private cloud infrastructures, as do VMware by Broadcom products including VMware Cloud Foundation (VCF) and VMware vSphere Foundation (VVF), which comprise programmes like vSphere, vCenter, vSAN, and NSX. The strategic partnership between AMD and VMware By Broadcom helps organisations install and manage virtualized environments that meet their performance and scalability needs while optimising energy efficiency and TCO.
Two typical deployment scenarios for VCF/VVF exist:
- Conventional VVF infrastructure typically comprises discrete networking, storage, and computation gear. Running this model requires managing each part independently, which frequently calls for specialised knowledge and equipment, but it produces the best results. The computational component is virtualized using software like VMware vSphere, which is controlled by a virtual infrastructure administrator. This configuration’s storage component depends on specialised Storage Area Network (SAN) or Network Attached Storage (NAS) devices under the control of a Storage Administrator.
- A contemporary method of managing data centres is called hyper-converged infrastructure (HCI), which integrates networking, storage, and processing into a single, cohesive system. Typically, HCI systems pool storage resources throughout the entire infrastructure for simple scaling and management by a Virtual Infrastructure Administrator. This is made possible by Software Defined Storage (SDS), which is made possible by products like vSAN. Compared to conventional VCF/VVF installations, this simplified method can maximise savings while enhancing agility and scalability.
VMmark Benchmark
By offering thorough performance measurements that mimic real-world workloads to evaluate CPU, memory, storage, and networking performance, the VMmark benchmark helps enterprises to evaluate the effectiveness and scalability of their virtualized systems. Businesses may decide on hardware configurations, resource allocation, and workload management methods with knowledge thanks to this comprehensive picture of system capabilities under various workload scenarios.
By applying the smallest unit, or “tile,” of load, VMmark3 employs a weighted scoring system to assess a server’s performance. Nineteen virtual machines are included of each tile, which represents a typical enterprise virtualization scenario by executing a wide collection of tasks both concurrently and collaboratively. Each tile’s static nature allows for the size and scalability of the amount of work for every distinct VMmark3 cluster.
With a growth in hosts, sockets, and cores, there are usually more tiles. Each tile also serves as a representation of the maximum theoretical score and a Quality of Service (QoS) metric, which are used to appropriately scale each publication: an excessive number of tiles exceeds the benchmark latency requirements, while an insufficient number of tiles limits the maximum score.
While application performance accounts for a piece of the total score, the quantity of tiles is essential to comprehending the potential of a VCF deployment. Many of the capabilities that are exclusive to VCF, like XvMotion, Storage vMotion, vMotion, and the Distributed Resource Scheduler (DRS), are also utilised by standard virtualized deployments. By utilising these capabilities, VMmark workloads offer a more comprehensive view of what can be accomplished with the entire system as opposed to just one server. Depending on the underlying technology, these infrastructure processes can vary in scope and duration, accounting for 20% of the total score.
Benefits of Generational Performance
AMD is committed to innovation and keeps delivering notable gains in efficiency and performance for important applications. As can be seen in Figure 1, AMD EPYC processors have significantly improved in performance with each iteration when used in VCF scenarios. These assessments include single- and dual-processor installations in both normal settings and vSAN configurations. The consistent dedication of AMD towards meeting the dynamic and ever-increasing demands of contemporary virtualized infrastructures is demonstrated by the notable performance improvements attained with every new generation of AMD EPYC CPUs.
Performance of Throughput
Thanks to continued AMD processor technology advances and collaborative VMware By Broadcom optimisations, the preceding section of this blog revealed constant performance increases across successive AMD EPYC processor generations. The performance domination of AMD EPYC processors over several processing generations is demonstrated in Figure 2, which highlights their unwavering commitment. In VMmark3 performance tests conducted on clusters comprising two dual-socket servers for a total of four processor sockets, the 4th generation AMD EPYC 9654 CPU outperforms even the most recent 5th generation Intel Xeon Platinum 8592+ processor.
Per-Core Efficiency
When it comes to power, space, and server count optimisation, per-server throughput is a crucial measure. Additionally, because per-core performance is frequently a major issue in software licencing models based on core count, it is especially important for workloads that are SLA-critical.
The performance improvements of 32-core AMD EPYC 9374F processors over 32-core Intel Xeon Gold 6548Y+ processors and 64-core AMD EPYC 9554 processors over 64-core 5th Gen Intel Xeon 8592+ processors are shown in Figure 3.
Power Effectiveness
In order to minimise cooling requirements, maximise utility costs, and support sustainability goals, modern data centres must have efficient power consumption. In comparison to both 4th and 5th Gen Intel Xeon processors, Figure 4 illustrates how 4th Gen AMD EPYC 9654 CPUs provide higher power efficiency.
Space for a Data Centre
VMmark Scores
Data centres have tremendous incentives to optimise space, electricity, and cooling for sustainability and TCO. Less servers that may offer more virtual machine consolidation without sacrificing the speed required for corporate operations can be used to achieve these goals. Suppose AMD workloads need to achieve business objectives with resources and quality of service equal to a total VMmark3 score of 500. This aggregate score can be attained with just 26 servers running 4th Gen AMD EPYC 9654 processors, as shown in Figure 5.
AMD will require 38 servers if they choose to employ servers with 5th generation Intel Xeon 8592+ processors. It may take up to 112 servers for customers with 2nd Gen Intel Xeon 8280 processor-based servers to get the same aggregate score! That’s more than 4 times the amount of AMD EPYC 9654 processor-based servers required to meet the same workload in business! This comparison emphasises the significance of not only power efficiency but also making the best use of server rack space in datacenters.
VMmark Benchmark Results
This article demonstrated how VMmark victories and the next generation of AMD EPYC processors can drive consistent performance leadership, giving organisations optimal hardware consolidation and unparalleled x86 workload performance that may help reduce total cost of ownership. Businesses can safely depend on AMD EPYC processors to satisfy their changing virtualized infrastructure demands while maximising available power and capacity. Together, AMD and VMware by Broadcom are still dedicated to promoting innovation that continuously yields record-breaking virtualized environment performance and efficiency. Businesses can achieve new levels of agility, scalability, security, and reliability in their IT infrastructures by utilising Broadcom’s combination of experience in AMD and VMware.
