JDI: Virtualization is no replacement for physical network servers
While the attraction of virtual computing and the ability to run multiple operating systems optimized for different tasks on a single physical machine is compelling, there are also potential drawbacks, according to an article in this month’s Journal of Digital Imaging.
Steve G. Langer, PhD, and Todd L. French from the Mayo Clinic in Rochester, Minn., attemped to gain insight into the impact of virtualization on performance by benchmarking metrics including local memory and disk bandwidth, network bandwidth and integer and floating point performance.
“High performance requirements need to be carefully considered if they are to be executed in an environment where the running software has to execute through multiple layers of device drivers before reaching the real disk or network interface,” the authors wrote.
A 32-bit virtual machine was chosen as the benchmark platform for portability as a 32-bit virtual machine can run on either a 32- or 64-bit host operating system while the converse is not true, according to Langer and French. “This is important because a cost-sensitive user running a virtual environment on Microsoft tools may not be able to afford the additional charges that are incurred for that company’s 64-bit high performance products.”
Results were broken into three groups based on whether the test appliance was operated on bare metal, a thin hypervisor or a thick hypervisor residing on a host operating system. The authors found that read/write (R/W) performance in RAM, local disk and network disk comprise three distinct areas, and the winner is not consistent.
“One would expect, and indeed we certainly did, that the thin hypervisor group would be closest to bare metal results,” the authors wrote. “However, the results are more complex than that, and…selecting the ‘best’ virtual machine environment depends on the target application’s behavior; is it compute limited, R/W limited or a combination of both?”
Langer and French, based on their results, deduced that for high speed network file or web-sharing needs, no virtual machine result is better than about 25 percent of bare metal performance. “Hence, virtual machine methods cannot be recommended as a competitive replacement for physical network file servers at this time.
“For various reasons we have found it very productive to adopt virtualization in our practice, but this direction is not without its drawbacks,” the authors concluded, particularly noting read performance on local and network disk is negatively impacted as is floating point performance. “Applications that are very sensitive to these requirements may not provide satisfactory performance in a network environment. Also, in contrast to expectations the best performance was often seen from a thick virtualization tool rather than the thin hypervisor environment.”
Steve G. Langer, PhD, and Todd L. French from the Mayo Clinic in Rochester, Minn., attemped to gain insight into the impact of virtualization on performance by benchmarking metrics including local memory and disk bandwidth, network bandwidth and integer and floating point performance.
“High performance requirements need to be carefully considered if they are to be executed in an environment where the running software has to execute through multiple layers of device drivers before reaching the real disk or network interface,” the authors wrote.
A 32-bit virtual machine was chosen as the benchmark platform for portability as a 32-bit virtual machine can run on either a 32- or 64-bit host operating system while the converse is not true, according to Langer and French. “This is important because a cost-sensitive user running a virtual environment on Microsoft tools may not be able to afford the additional charges that are incurred for that company’s 64-bit high performance products.”
Results were broken into three groups based on whether the test appliance was operated on bare metal, a thin hypervisor or a thick hypervisor residing on a host operating system. The authors found that read/write (R/W) performance in RAM, local disk and network disk comprise three distinct areas, and the winner is not consistent.
“One would expect, and indeed we certainly did, that the thin hypervisor group would be closest to bare metal results,” the authors wrote. “However, the results are more complex than that, and…selecting the ‘best’ virtual machine environment depends on the target application’s behavior; is it compute limited, R/W limited or a combination of both?”
Langer and French, based on their results, deduced that for high speed network file or web-sharing needs, no virtual machine result is better than about 25 percent of bare metal performance. “Hence, virtual machine methods cannot be recommended as a competitive replacement for physical network file servers at this time.
“For various reasons we have found it very productive to adopt virtualization in our practice, but this direction is not without its drawbacks,” the authors concluded, particularly noting read performance on local and network disk is negatively impacted as is floating point performance. “Applications that are very sensitive to these requirements may not provide satisfactory performance in a network environment. Also, in contrast to expectations the best performance was often seen from a thick virtualization tool rather than the thin hypervisor environment.”