Researchers Submit Patent Application, "Framework and Method for Monitoring Performance of Virtualized Systems Based on Hardware Base Tool", for Approval
No assignee for this patent application has been made.
News editors obtained the following quote from the background information supplied by the inventors: "In the past, the performance of computer software system, including operating system (OS) and its applications, has been a concern of software vendors and their customers, since system performance relates directly to the service quality and success of software product sales. In general there are two angles to view the issue of performance:
"The first angle views the 'consequence' of application software, or the 'user experience' as performance. For example, software code is inserted into applications to measure the 'response time' from the return key to the response appearing on computer screen. The performance measurement code may be combined with application functions to form a complicated Performance Monitor, such as Couture's patent US20080235075 'Enterprise application performance monitors'.
"The second angle views the 'cause' that impacts the performance of running software. The impacting root cause is, of course, the hardware resources such as CPU, memory, IO and network. In the past there is no way to measure performance from hardware, hence the second best way can only utilize the OS performance commands to measure performance of the resources like threads, processes, etc. Lately, new hardware monitoring base tool emerges, such as x86-based hardware register capable of monitoring CPU, called Performance Monitor Unit (PMU). There are quite a few PMU patents: e.g. Davidson's U.S. Pat. No. 6,718,403 'Hierarchical selection of direct and indirect counting events in a PMU', Mericas's U.S. Pat. No. 7,500,138 'Simplified event selection for a PMU' and Mericas's US20060167658 'Method and product of PMU for sampling all performance events generated by a processor'. There are also non-PMU patents that are hardware and performance monitoring based, such as Fowles's US20060277395 'Processor performance monitoring'; Kosche's US20080177756 'Method and apparatus for synthesizing hardware counters from performance sampling', and Hunter's US20080294944 'Processor bus for performance monitoring with digests'.
"Two kinds of registers help monitoring performance: performance control register (PMC) and performance data register (PMD). PMC monitors CPU at regular time intervals, and saves monitored data into PMD. PMC is event-based or time-based to consolidate the collected event information and report the monitoring results to higher software layers. Advantages of hardware monitoring are: (1) in the past, certain software-based monitors require to continuously change the source code of the monitored software to reach its goal, e.g. optimizing a Java virtual machine routine. On the contrary, hardware-based monitoring requires no source code. (2) The performance of monitored software is not affected by monitoring activities. (3) Very low-level kernel code can be monitored. Most importantly, (4) Capability to monitor cache is nowhere found in any other monitoring means. One example is TLB miss and hit, where TLB stands for 'translation look-aside buffer', the CPU cache hardware for memory management to improve the speed of virtual H physical address translation. TLB takes virtual address as search keyword. The search result is physical address: if the needed address can be found in TLB, it is called TLB hit. Otherwise, it is called TLB miss. Frequent TLB misses indicate the performance of resource is degraded. The details of PMU are described in the document written by Eranian,
"There are not very many PMU applications. Moreover, there is no existing method to utilize PMU and combine OS performance commands to measure the performance of 'virtualized systems'. Here is the explanation of virtualized systems:
"The technology of OS virtualization leads to the result of a physical machine is capable of running multiple 'guest' OSs (or Virtual Machine, briefly VM, or simply 'guest'). The VMs run on VMM (virtual machine monitor, or hypervisor), and Hypervisor runs directly on physical machine. In most cases, the monitored physical machines are servers (or host) in the data center. There are different OS virtualization techniques, such as para-virtualization or full virtualization. In general, a host has at least two spaces to be monitored: one is the space where VMs are running, called (1) guest space; the other is the space where a hypervisor is running, called (2) kernel space. Some virtualization technique, such as Linux KVM, keeps the use of the original (3) user space. The above explains the internals of a host. Multiple such hosts form a cluster, with a lead host called master, and several subordinate hosts called slaves. Multiple clusters become a network system, being distributed, centralized, or mixed. This explains 'virtualized systems' of the disclosed invention, a network system covering large geographical areas.
"Tang's patent US20110296411 'Kernel bus system to build Virtual Machine Monitor and the performance service framework & Method therefore' describes how virtualized systems analyze performance data, perform VM migration, and mediate workload. The patent however does not describe how virtualized systems acquire performance data."
As a supplement to the background information on this patent application, VerticalNews correspondents also obtained the inventors' summary information for this patent application: "In order to resolve the existing technical issue, the objective of the disclosed invention is to provide a framework and method of monitoring performance of virtualized systems based on hardware base tool, to apply PMU (or similar tool) to the performance monitoring of virtualized systems. The said method monitors CPU, memory, cache, IO, network, process, etc. of multiple hosts in virtualized systems. Meanwhile, the hardware-based monitoring method of the disclosed invention resolves the issue of how to acquire performance data for virtualized systems.
"One of the disclosed inventions involves the said framework and method to monitor the performance of virtualized systems based on hardware base tool, comprise at least one master host, each of the said master host includes user space components, guest space components, and kernel space components, wherein
"The said user space components comprise connected policy manager, workload mediator, monitoring library, and host performance monitor. The host performance monitor is further connected with workload mediator, and the host performance monitor comprises user space monitor and kernel space monitor;
"The said guest space components comprise at least one virtual machine (VM) connected with said host performance monitor via kernel serial channel;
"The said kernel space components comprise performance monitor base tool application interface (API) and performance monitoring base tool core code module, connected with said monitor library, as well as task scheduler, memory management unit, network driver, file system and device driver, wherein the said network driver is connected with said workload mediator;
"The said hardware supports peripheral performance monitoring base tool, wherein the hardware comprises PMU connected with said performance monitoring base tool core code module, PMD and PMC respectively connected with PMU, and CPU, memory, hard disk and network card wherein the said network card is connected with network driver.
"The said framework to monitor virtualized systems based on hardware base tool comprises at least one slave host. Each of the said master host is connected with the said at least one slave host via net pipe to form a cluster. Multiple said master hosts and their respective slave hosts form multiple clusters, and said clusters communicate with each other via centralized, distributed or mixed network architecture. The said slave host comprises said host performance monitor wherein the host performance monitor of slave host is connected to the network card of said master host.
"The second disclosed invention involves the said method to monitor performance of the virtualized systems based on hardware base tool, comprising:
"Firstly, administrators via graphical interface, or user applications via said performance monitor APIs, make performance mediating requests to said policy manager;
"Secondly, when said policy manager interprets the said performance mediating requests, it instructs the said workload mediator making request to said master host performance monitor, asking the monitor to report individual slave resource performance conditions via net pipe;
"Finally, said policy manager reports overall performance condition to said administrators or said user applications.
"Because of the above said technical solution, the disclosed invention is capable of monitoring performance of at least one cluster, wherein the cluster comprises at least one host and the host comprises at least one virtual machine. Therefore, the disclosed invention is also capable of monitoring the performance of hosts. The disclosed invention provides for enterprises and cloud service providers a framework and method to monitor the performance of virtualized systems. The disclosed invention is not about the PMU hardware itself. Rather, it uses a method based on hardware base tool to monitor, calls the APIs of the tool to realize the method, and builds a software performance monitor. System administrators may also interact with the interface provided by the monitor, e.g. an application to adjust virtual and physical resources may call the said monitor interface to obtain the rough ideas of idle/busy situations of the entire clusters, and proceed to adjust these resources.
BRIEF DESCRIPTION OF THE DRAWINGS
"FIG. 1a is the architecture block diagram of the framework to monitor performance of virtualized systems based on hardware base tool;
"FIG. 1b is the upper layer class diagram of the framework to monitor performance of virtualized systems based on hardware base tool;
"FIG. 2 is the flowchart of the policy manager in the framework to monitor performance of virtualized systems based on hardware base tool;
"FIG. 3a the flowchart of the workload mediator in the framework to monitor performance of virtualized systems based on hardware base tool;
"FIG. 3b the state diagram of the workload mediator in the framework to monitor performance of virtualized systems based on hardware base tool;
"FIG. 4 the flowchart of the host performance monitor in the framework to monitor performance of virtualized systems based on hardware base tool;
"FIG. 5a the flowchart of the system internal resource monitor in the framework to monitor performance of virtualized systems based on hardware base tool;
"FIG. 5b the flowchart of the system external resource monitor in the framework to monitor performance of virtualized systems based on hardware base tool;
"FIG. 6 the lower layer class diagram of the framework to monitor performance of virtualized systems based on hardware base tool;
"FIG. 7 the flowchart of the monitor library in the framework to monitor performance of virtualized systems based on hardware base tool;
"FIG. 8 the flowchart of the monitor in the framework to monitor performance of virtualized systems based on hardware base tool;
"FIG. 9 the flowchart of the monitored object in the framework to monitor performance of virtualized systems based on hardware base tool;
"FIG. 10 the flowchart of the event executor in the framework to monitor performance of virtualized systems based on hardware base tool."
For additional information on this patent application, see: Tang, Changbin; Cai, Penghao. Framework and Method for Monitoring Performance of Virtualized Systems Based on Hardware Base Tool. Filed
Keywords for this news article include: Patents, Software, Legal Issues.
Our reports deliver fact-based news of research and discoveries from around the world. Copyright 2014, NewsRx LLC
Most Popular Stories
- U.S. Families 'Extraordinarily Vulnerable': Yellen
- Hillary Clinton to Address CHCI Conference
- Larry Ellison Steps Down as Oracle CEO
- Alibaba Prices IPO at $68 a Share
- Veterans to Get Training as Solar Panel Installers
- Apple Locks Itself Out of Devices
- Hispanics Doubt Marco Rubio's Chances
- Wildfires Rage in California
- John Cantlie Delivers ISIS Message to Save Life
- Alibaba: Today China, Tomorrow the World