VMmark is a free tool used by hardware vendors and others to measure the performance, scalability, and power consumption of virtualization platforms.
The VMmark benchmark:
A cloud environment typically collects several diverse workloads onto a virtualization platform — a collection of physical servers accessing shared storage and network resources. Traditional single-workload performance and scalability benchmarks for non-virtualized environments were developed with neither virtual machines nor cloud environments in mind. Even single-server virtualization benchmarks did not fully capture the complexities of today's virtualized data centers.
VMmark 1.x pioneered single-server virtualization benchmarking with its unique tile-based multi-application design. VMmark 2.x expanded this to multiple servers and platform-level workloads. Applications have evolved since then, and VMmark 3 addresses these changes in application architecture and platform-level operations. More than ever before, users are running highly scalable workloads, often referred to as third platform applications, and increasingly complex OLTP workloads.
VMmark 3 features a highly automated setup and tile-cloning process that makes benchmark deployment fast and easy, with the need for very little manual intervention. VMmark 3 uses free or open-source software throughout, eliminating the need to purchase software licenses, and the VMmark .ova includes all the needed software in one downloadable template.
The VMmark benchmark combines commonly-virtualized applications into predefined bundles called "tiles." The number of VMmark tiles a virtualization platform can run, as well as the cumulative performance of those tiles and of a variety of platform-level workloads, determine the VMmark 3 score.
Before being published, VMmark results must be submitted to a review panel composed of a variety of companies that have published VMmark benchmark results, ensuring the fairness and integrity of the benchmark.
Since its inception in 2007, more than 200 VMmark results have been published on the VMmark website and VMmark has become the standard by which the performance of virtualization platforms is evaluated.
VMmark uses workloads representative of the highly scalable and complex applications commonly found in the data center. VMware has worked closely with its partners to design and implement the benchmark and has gathered extensive customer feedback to understand how these applications are typically used in virtualized environments.
The unit of work for a virtualized data center can be usefully defined as a collection of virtual machines executing a set of diverse workloads. The VMmark benchmark refers to this unit of work as a "tile." Each VMmark tile is paired with a client system that drives the tile’s virtual machines to perform a variety of tasks, some internal to each virtual machine, some involving other virtual machines in the tile, and some involving the client system.
The total number of tiles that multiple systems in the data center can accommodate, while infrastructure operations are performed in the background, provides a coarse measure of that data center's consolidation capacity. The performance of the workloads within those tiles provides a fine measure of the data center's overall performance and, combined with the performance of the administrative operations, is used to calculate a VMmark benchmark score.
Power and cooling expenses are a substantial — and increasing — part of the cost of running a data center. Additionally, environmental considerations are a growing factor in data center design and selection. To address these issues, VMmark enables optional power measurement in addition to performance measurements. VMmark 3 benchmark results can be any of three types:
- Performance only (no power measurement)
- Performance with server power
- Performance with server and storage power
VMmark results with power measurement allow hardware purchasers to see not just absolute performance, but also absolute power consumption and performance per kilowatt. This makes it possible to consider both capital expenses and operating expenses in the selection of new datacenter components.
VMmark 3 can automatically collect ESXTOP power data from the servers hosting VMmark application workloads. When enabled, this provides power usage tracking based on the server's power supply sensor.
Note: Though VMmark uses the SPEC® PTDaemon, VMmark results are not SPEC metrics and cannot in any manner be compared to SPEC metrics. ESXTOP power data cannot be used in published results or compared to PTDaemon metrics.
The rapid pace of innovation has quickly transformed typical server usage by enabling easier virtualization of bursty and heavy workloads, dynamic virtual machine relocation (vMotion), dynamic datastore relocation (storage vMotion), and automation of many provisioning and administrative tasks across large-scale multi-host environments. In this paradigm, a significant proportion of the stresses on the CPU, network, disk and memory subsystems can be generated by the underlying infrastructure operations. Load balancing across multiple hosts can also greatly affect application performance. Any relevant benchmarking methodology must still focus on user-centric application performance while accounting for the effects of this infrastructure activity on overall platform performance. VMmark 3 generates a realistic measure of platform performance by incorporating a variety of platform-level workloads such as shared nothing migration, virtual machine migration, clone and deploy, snapshotting, and storage migration operations, in addition to traditional application-level workloads.
Elasticity is critical in measuring the real-world workloads that are found in today’s datacenters. The act of adding and subtracting resources to meet demands is now more common than ever for self-scaling applications. VMmark 3 adds this component alongside a cyclical application profile to more accurately represent today’s bursty environments.
During a VMmark benchmark run, which lasts at least three hours, individual performance metrics are collected every 60 seconds. Each of these metrics represents the performance of an individual application or infrastructure workload.
The application workload metrics for each tile are computed and aggregated into a score for that tile by normalizing the different performance metrics, such as operations/second or transactions/second, with respect to a reference system. A geometric mean of the normalized scores is then computed as the final score for the tile. Finally, the resulting per-tile scores are summed to create the application workload portion of the final metric.
A similar calculation is used to create the infrastructure workload portion of the final metric except that, unlike the application workloads, the infrastructure workloads are not scaled explicitly by the user. Consequently, the infrastructure workloads are compiled as a single group and no multi-tile sums are required.
The final benchmark score is computed as a weighted average: 80% to the application workload component and 20% to the infrastructure workload component. These weights were chosen to reflect the relative contribution of infrastructure and application workloads to overall resource demands.
In order for the resultant benchmark score to be considered compliant, the benchmark run must also meet a number of conditions, including minimum quality-of-service requirements.
In addition to the overall benchmark score, a VMmark full disclosure report also includes the raw and normalized results for each underlying workload and complete details of the virtualization platform configuration. In some cases, studying the workload metrics along with the platform configuration can provide insight into system performance and scaling.
This section provides an abbreviated outline of the VMmark 3 system requirements. For a detailed list, see the most recent version of the VMmark User’s Guide (on the VMmark download page, available after accepting the VMmark EULA).
VMware vCenter Server installed on a separate, dedicated server
One Prime Client virtual machine with:
One client virtual machine per tile, each with:
Virtual client hosts with the following:
vMotion networking (10 Gbps network recommended)
To get started with VMmark, follow these steps: