This long post discusses guest machine performance under Hyper-V looks when there is an over-provisioned Hyper-V Host. When there is an over-provisioned Hyper-V Host, guest machine workloads are subject to a minimal performance penalty, which I will attempt to quantify. This is the eighth post in a series on Hyper-V performance. The series began here.
It is easy to recognize a generously over-provisioned Hyper-V Host machine – its processors are underutilized and machine memory is not fully allocated. When the machine’s logical CPUs are seldom observed running in excess of 25-40% busy, there is ample CPU capacity for all its resident guest machines, especially considering that most Hyper-V Host machines are multiprocessors. Memory can safely be regarded as underutilized when more than 40% of it is Available for allocation by the hypervisor and no guest machine is running at its maximum Dynamic Memory setting.
Note: The dispatching of a guest machine virtual processor is delayed when all CPUs are busy, so it is forced to wait. In a symmetric multiprocessor, the probability that all CPUs are busy simultaneously is the joint probability that all the processors are busy. For example, if there are four CPUs and each CPU is busy 25% of the time, the joint probability of all the CPUs being busy simultaneously is 0.25 * 0.25 * 0.25 * 0.25, or 0.004. The probability that all CPUs are simultaneously busy is only ~0.4%.
When Hyper-V Host machines are over-provisioned, the performance of guest machine applications approaches the level of native hardware. The problem with over-provisioned VM Host machines is that they are not economical. Over-provisioning on a wide enough scale often leads an initiative to increase the degree of sever consolidation by trying to pack more guest machines into the existing virtualization infrastructure.
Discussing over-committed Hyper-V Host machines and under-provisioned guest machines is much more interesting because that is when serious performance problems can occur. The first set of benchmarking runs reported below was directed at showing how these two conditions can be characterized based on various Hyper-V and Windows OS performance measurements.
Several of the benchmarking runs discussed in this section deliberately overload the Hyper-V Host processors or the Host machine’s memory footprint and then look at the Hyper-V and guest machine performance counters that characterize those overloaded conditions. As we have seen, you can implement virtual processor and dynamic memory priority settings to try and protect higher priority workloads from this degradation. Additional benchmarking runs involving physical resources being over-committed reveal how effective these Hyper-V virtual processor and machine memory priority settings are in shielding higher priority workloads from the performance impact when the Hyper-V Host is overloaded.
Scaling up and scaling out
While improving application performance is not among them, virtualization hardware and software technology has evolved to the point where it currently provides compelling benefits in modern data center operations. For example, one of the clear trends in hardware manufacturing that favors virtualization is building more powerful multiprocessor cores, not faster individual processors. CPU manufacturers have resorted to packaging more processors on a chip rather than making processors faster because increases in clock speeds lead to disproportionate increases in power consumption, which also ramps up the amount of heat that has to be dissipated. In semiconductor fabrication, the manufacturers have encountered a “power wall” that resists other engineering solutions.
A second factor that promotes virtualization is industry Best Practices that lead to building and deploying Windows machines that are dedicated to performing a single role, whether they are explicit Server roles or more general purpose desktop and portable workstations handling diverse personal computing tasks. A related practice is the Technical Support group within the IT organization building and then certifying for distribution one or more stable images of the operating system and the application software installed on top of it after a lengthy period of comprehensive Acceptance Testing. This stable image is then cloned each time there is an organizational need to support another copy of this application. Virtualization software that can deploy new copies of these system images through rapid cloning of virtual machines – a process that can also be automated – adds valuable flexibility to data center operations.
Most of the virtual machines configured to handle a single server role are clearly not well matched against the powerful capabilities of the data center machines they would be deployed to. Without virtualization, these data center machines would often be massively over-provisioned if they were only capable of running an individual Windows Server workload. Virtualization technology offers relief from this conundrum, a convenient way to consolidate many of these individual workloads on a single piece of equipment. Essentially, virtualization technology provides a flexible, software-based mechanism that allows system administrators to utilize current hardware more effectively while retaining all the administrative advantages of isolating workloads on dedicated servers.
Still, spinning up a new guest machine from the standard server or workstation Build is not the only possible response to each new request for IT services. There are viable alternatives, including allowing a single instance of IIS, for example, to host multiple application web sites or installing multiple instances of SQL Server on a production or test machine. IT professionals are sometimes reluctant to choose these configuration alternatives because they are concerned about the performance risks associated with multiple web servers sharing a single machine image, for example. Of course, these performance risks do not magically disappear when multiple guest machines are provisioned instead. The problem of over-committing shared computer resources is merely elevated to the level associated with Hyper-V administration.
Finally, having firmly established itself as an integral part of large scale data center operations, virtualization technology continues to evolve other virtual machine management capabilities, including replication, live migration, dynamic load balancing, automatic failover and recovery. The flexibility that virtualization solutions also provide in being able to provision a new machine quickly can benefit the performance of workloads that are running up against capacity limits in their current configuration and need to scale out across multiple machines in an application cluster to achieve higher levels of throughput.
To gain some additional perspective on the performance impact of virtualization, we will look first at some benchmarking results showing the performance of virtual machines in various simple configurations, which we will also compare to native performance where Windows is installed directly on top of the hardware. For these performance tests, I used a benchmarking program that simulates the multi-threaded CPU and memory load of an active ASP.NET web application, but without issuing disk or network requests so that those limited resources on the target machine are not overwhelmed in the course of executing the benchmark program.
The benchmark program I used for stress testing Hyper-V guest machines is a Load Generator application I wrote that is parameter-driven to generate a wide variety of “challenging” workloads. The current version is a 64-bit .NET program written in C# called the ThreadContentionGenerator. It has a main dispatcher thread and a variable number of worker threads, similar to ASP.NET. You set it to execute a fixed number of concurrent tasks, and perform a specific number of iterations of each task. Each task allocates a large .NET collection object that it then fills with random data. It then searches the collection repeatedly, and finally deletes all the data. In this fashion, the program stresses both the processor and virtual memory. Periodically, each active thread simulates an IO wait by sleeping, where the simulated IO rate and the IO duration is also subject to some degree of realistic variation.
The benchmark program is a very flexible beast that can be adjusted to stress the machine’s CPUs, memory or both. You can execute it in a shared nothing environment where the threads execute independent of each other. Alternatively, you can set a parameter that adds an element of resource sharing to the running process so that the threads face lock contention. In contention mode, the main thread sets up some shared data structures that the worker threads access serially to generate a degree of realistic lock contention that can be dialed either up or down by increasing or decreasing the amount of processing spent in the critical section.
For this first set of Hyper-V guest machine performance experiments, I set the number of concurrent worker tasks to 32 and the number iterations to 90:
ThreadContentionGenerator.exe –tasks 32 –iterations 90
There are additional parameters to vary the virtual memory footprint of the program, the duration of IO waits and the rate of lock contention, but for this set of tests I let the program run with default values for those three parameters. With these settings, the program generates a load that is similar in many respects to a busy ASP.NET web application, one that is compute-bound, with requests that can be processed largely independent of each other. Note that the intent was to stress the Hyper-V environment, beginning by stressing the machine’s CPU capacity, without attempting a realistic simulation of a representative or a particular ASP.NET workload.
The hardware was an Intel i7 single socket machine with four physical CPUs (and Intel Hyper-Threading disabled) and 12 GB of RAM. The OS was Windows Server 2012 R2.
Native performance baseline
Running first on the native machine – after re-booting with Hyper-V disabled – the benchmark program ran to completion in about 90 minutes, the baseline execution time we will use to compare the various virtualization configurations that were tested. The only other active process running on the native Windows machine was Demand Technology’s Performance Sentry performance monitor, DmPerfss.exe, gathering performance counters once per minute.
At this stage, the only aspect of the benchmark program’s resource usage profile that is relevant is its CPU utilization. Because each task being processed goes to sleep periodically to simulate I/O, individual worker threads are not CPU-bound. However, since there are 32 worker threads executing concurrently and only four physical CPUs available, the overall workload is CPU-bound, as evidenced in Figure 25, which reports processor utilization by the top 5 consumers of CPU time during a one hour slice when the ThreadContentionGenerator program was active on the native machine.
Figure 25. Native execution of the benchmark program shows CPU utilization near 400% on a single socket machine with 4 physical CPUs. Instantaneous measurements of the System/Processor Queue Length counter, represented by a dotted line chart plotted against the right-hand y-axis, indicate a significant amount of processor queuing.
You can see in Figure 25 that overall processor utilization approaches the capacity of the machine at close to 400% utilization. The dotted line graph in Figure 25 also shows the instantaneous values obtained from the Processor Queue Length counter. The number of threads waiting in the Windows Scheduler Ready Queue exceeds fifteen for some of the observations. We can readily see that not only are the four physical CPUs on the machine quite busy, at many intervals there are a large number of ready threads waiting for service. Figure 26 confirms that the threads waiting in the Ready Queue are predominately from the ThreadContentionGenerator process (shown in blue), which is the behavior I expected, by the way.
Figure 26. This chart charts threads with a Wait State Reason indicating they are waiting in the OS Scheduler Ready Queue. As expected, most of the ready threads in the Ready Queue are from the benchmark program, the ThreadContentionGenerator process.
Standalone in the Root partition
In the next scenario, running standalone on the Root partition under Hyper-V with no child partitions active, the same benchmark executed for approximately 100 minutes, about 11% longer than the native execution baseline. In many scenarios a 10% performance penalty is a small price to pay for the other operational benefits virtualization provides, but it is important to keep in mind that there is always some performance penalty that is due whenever you are running an application in a virtualized environment.
Applications take longer to run inside a virtual machine compared to running native because of a variety of virtualization costs that are not encountered on a native machine. These include performance costs associated with Hyper-V intercepts and Hypercalls, plus the additional path length associated with synthetic interrupt processing. As mentioned above, the benchmark program simulates IO by issuing Timer Waits. These require the timer services of the hypervisor, which are less costly that the synthetic interrupt processing associated with disk and network IO. So, the 10% increase in execution time is very likely a best case of the performance degradation to expect.
Those costs of virtualization are minor irritants so long as the Hyper-V Host machine can supply ample resources to the guest machine. The performance costs of virtualization do increase substantially, however, when guest machines start to contend for shared resources on the Host machine.
Since processor scheduling is under the control of the hypervisor in the second benchmark run, for reliable processor measurements, it is necessary to turn to the Hyper-V Logical Processor counters, as shown in Figure 27. For a one-hour period while the benchmark program was active, overall processor utilization is reported approaching 400%, but you will notice it is slightly lower than the levels reported for the native machine in Figure 25. Figure 27 also shows an overlay line graphing hypervisor processor utilization against the right-hand y-axis, which accounts for some of the difference. The hypervisor consumes about 6% of one processor over the same measurement interval. The amount of CPU time consumed directly by the Hyper-V hypervisor is one readily quantifiable source of virtualization overhead that causes performance of the benchmark application to degrade by 10% or so.
Figure 27. Running the benchmark workload standalone on the Root partition, the hypervisor consumes about 6% of one processor. Overall CPU utilization approaches 400% busy, slightly less busy than the configuration shown in Figure 23.
Reviewing the Hyper-V counter measurement data, we can see that thread execution inside the Root Partition executes on a virtual processor, subject to the hypervisor Scheduler, the same as the virtual processor scheduling performed for any guest machine child partition. When the Windows OS inside the Root Partition executes a thread context switch, the Hyper-V performance counters graphed in Figure 28 show that there is a corresponding hypervisor context switch. For child partitions, there is an additional Hyper-V Scheduler interrupt that requires processing on a context switch, so there is slightly more virtualization overhead whenever child partitions are involved.
Figure 28. Each time the Windows OS inside the Root Partition executes a thread context switch, there is a corresponding hypervisor context switch.
The Hyper-V Logical Processor utilization measurements do include a metric that should be directly comparable to the System\Processor Queue Length measurement that was shown in Figure 25 called CPU Wait Time per Dispatch, which is available at the virtual processor level. Unfortunately, this performance counter is not helpful, however. It is not clear what the units of Wait Time that are reported, although an educated guess is standard Windows 100-nanosecond timer units seems likely. It also reports Wait Time in very discrete, discontinuous measurements, which is strange. Together, these two issue make for problems of interpretation. Fortunately, the System\Processor Queue Length is an instantaneous measurement that remains serviceable under Hyper-V. Figure 29 shows the same set of Process(*)\% Processor Time counters and a Processor Queue Length overlay line as Figure 25. The length of the processor Ready Queue for the Root partition is comparable to the native benchmark run, with even some evidence that the Ready Queue delays are slightly longer in the configuration where virtualization was enabled.
Microsoft strongly suggests that you do not use the Root partition to execute any work other than what is necessary to administer the VM Host machine. There is no technical obstacle that prevents you from executing application programs on the Root partition like I did with the benchmark program. But it is not a practice that is recommended. The Root partition provides a number of high priority virtualization services, like the handling of synthetic disk and network IO requests, which you want to take pains to try not to impact by running any other applications in the Root.
Standalone in a single child partition
Given the prohibition against running applications in the Root, the more useful comparison quantifying the minimum overhead of virtualization would be to compare performance of a guest machine in a child partition with performance on native hardware. So, on the same physical machine, I then created a Windows 8.1 virtual machine and configured it to run with 4 virtual processors. Making sure that nothing else was running on the Hyper-V server, I then ran the same benchmark on the 4-way guest machine. This time the benchmark ran to completion in 105 minutes.
Notice that on the child partition the benchmark run took about 5% longer when a single 4X Guest machine was configured. This virtual machine had access to all the physical CPUs that were available on the physical machine and executed in a standalone environment where it did not have to contend with any other guest VMs for processor resources. 105 minutes in execution time is about 17% longer than it took the same benchmark program to execute in native mode. Figure 30, which shows the rate that the Hyper-V hypervisor processed several types of virtualization-related interrupts, provides some insight into why execution time elongates under virtualization. Notice that hypervisor Scheduler interrupts occur when child partitions are executing – these Scheduler interrupts do not occur when threads are executing inside the Root partition, as illustrated back in Figure 28.
Figure 30. Interrupt processing rates reported for the hypervisor when a child a partition is active.
This configuration was also noteworthy because the hypervisor CPU consumption was reported as about 8%, a slightly higher utilization level (+25%) than any of the other configurations evaluated.
Today, performance testing is often performed on virtual machines due to the fact that they are only intermittently active, plus the ease with which you can spin them up and tear them down again. In my experience it is reasonable to expect the same workload to take about 10% longer to execute if you run inside a VM under ideal circumstances, which implies the VM has access to all the resources it needs on the machine, and there is no or minimal contention for those resources from other resident guest machines. This first set of benchmark tests show that the performance degradation to expect when a guest machine executes on an efficiently-provisioned VM Host is for tasks to run approximately 10% slower. Consider this a minimum stretch factor that elongates execution time due to various virtualization overheads. Furthermore, it is reasonable to expect this stretch factor to increase whenever the guest machine is under-provisioned or the Hyper-V machine is over-committed.
In the next post, this baseline measurement is compared to the other possible VM configurations: an efficiently-provisioned Host machine, an over-committed VM Host machine, and, finally, an under-provisioned guest machine. In the case of an efficiently-provisioned VM Host machine, we can expect a stretch factor comparable to the minimum stretch factor reported here. However, as we will see, when the VM Host machine is significantly over-committed or the guest machine is significantly under-provisioned, quest machine workloads can experience a severe performance penalty.