IL194654A - Method and system for transaction resource control - Google Patents

Description

22966/08 ywrm ΌΝΒ¾ np n "?y ηϋ1^ roiysn π ^ METHOD AND SYSTEM FOR TRANSACTION RESOURCE CONTROL 194654 ,7'JI I 453506 rnx METHOD AND SYSTEM FOR TRANSACTION RESOURCE CONTROL Field of the Invention The present invention is related to computer resources management. More particularly, the invention is related to a method for providing prioritizing to important processes by controlling respective resource consumption of processes and transactions (such as processes of the operating system, application processes, database transactions, network activity, I O requests) and for allocating more resources to one or more processes at the expense of the others.

Background of the Invention In any given computerized system, a set of processes/transactions compete for resources. For example, processes that are related to the Central Processing Unit (CPU), Input/Output (I O) devices (the communication the computer and the user or another information processing system), network resources, or any other shared computer resource. A "heavy" process which requires many resources causes the other processes to slow down or stop (e.g., not to respond) due to lack of available resources. Such processes/transactions may include: A process that is a part of an Operating System (e.g., Windows, · UNIX, Linux, OS2), such as device drivers, kernel (the core that provides basic services for all other parts of the operating system), services, daemons (A daemon is a program that runs continuously and exists for the purpose of handling periodic service requests that a computer system expects to receive).

• An application process: any process or sub-process (such as a thread - a placeholder information associated with a single use of a program that can handle multiple concurrent users. It allows a program to know which user is being served as the program alternately gets re-entered on behalf of different users) that is running in an operating system. For instance, computer games, antivirus, Office tools, utilities, mail servers, ERP applications, etc.

• Database Transaction - a database operation (such as SELECT / INSERT / UPDATE / DELETE) or a set of operations.

• Network Traffic - a stream of IP packets from any source to any destination.

• I O Traffic - a stream of I/O operations between a computer and a storage or an I O device.

AU the conventional systems described above have not yet provided satisfactory solutions to the problem of dynamically prioritizing important processes executed by a computerized system and for allocating more resources to such processes.

It is therefore an object of the present invention to dynamically prioritize important processes executed by a computerized system and for allocating more resources to such processes at the expense of the others.

It is another object of the present invention to improve performance and to maintain a normal level of performance and response time to users while heavy transactions are running in the background.

It is a further object of the present invention to improve performance and to free resources for important users/actions.

It is still another object of the present invention to improve performance and to improve protection and stability of applications against heavy transactions/operations that may cause a Denial-of-Service" (DoS).

It is yet another object of the present invention to more effectively handle peak demands and for reducing the peak stress from the computerized system.

It is a further object of the present invention to improve performance and maintain a desired Quality of Service (QoS) for one or more transactions.

Other objects and advantages of the invention will become apparent as the description proceeds. 194654/2 The present invention is directed to a method of controlling resource consumption of running processes, sub processes and/or threads (such as a database or an application transaction) in a computerized system. Accordingly, resources consumed by less important processes are freed by periodically suspending (by causing them to demand less resources) and resuming these processes transparently to other entities of the computerized system and externally to the OS without intervention in its inherent resource allocation mechanism and allowing the OS of the computerized system to allocate the free resources to other running processes.

The amount of resources consumed by a running process is preferably controlled by repeatedly limiting its demand for system resources for a first period, during which no demand is presented and a second period, during which the preceding demand is resumed, until the process is finished. In order to control the system resources given to a process, allocation of execution time units is used instead of allocating CPU and/or I/O cycles.

Preferably, processes are accelerated by slowing down other processes. Running processes may be accelerated or slowed down by using a GUI (e.g., a sliding virtual throttle) for controlling the resources allocated to 194654/2 them. Suspending running processes may be carried out through an inherent API, a debugger or directly through the OS system calls, or by using a binary code or any other representation of a command, for suspending and resuming processes.

The amount of consumed resources may be controlled for a process that is already running and consuming resources. Preferably, the amount of resources consumed by an application is controlled by a set of rules, which can be determined externally to this application.

The present invention is also directed to computerized system, in which resource consumption of running processes sub processes and/or threads are controlled by freeing resources consumed by one or more processes by periodically suspending and resuming said one or more processes and allowing said computerized system to allocate the free resources to other running processes.

Brief Description of the Drawings The above and other characteristics and advantages of the invention will be better understood through the following illustrative and non-limitative detailed description of preferred embodiments thereof with reference to the appended drawings, wherein: Fig. 1 is a general flowchart of controlling the resource consumption of a process, according to a preferred embodiment of the invention; Fig. 2 schematically illustrates a: general architecture of a computerized system, in which resource consumption of processes can be controlled, according to a preferred embodiment of the invention; Fig. 3A is a graph that illustrates an example of how resources are controlled for a long running process, according to a preferred embodiment of the invention; Fig. 3B is a graph that illustrates an example of how resources are controlled for a long running process, according to another preferred embodiment of the invention; and Figs. 4A and 4B are pie-charts that illustrate how important transactions in the system can be given more resources based on the method of controlling and limiting the execution time of other running transactions; Fig. 4C illustrates the results of controlling the resource allocation to a transaction, even when not all resources are used; Fig. 5 illustrates an intuitive dashboard for showing the status of the system; and Fig. 6 illustrates a Define Rule window for slowing down the system when a transaction has started by sh'ding a virtual throttle slider.

Detailed Description of Preferred Embodiments The present invention is directed to a method of controlling computerized resources, based on allocating chunks of execution time in order to control the system resources given to a process. Generally, the method is based on stopping and resuming the transactions and processes on the fly in realtime, and by limiting the consumption of computer resources to some processes while allocating more resources to other transactions in the system, without requiring a priority level to be determined on launching the processes.

Different embodiments of the present invention additional intend to: ■ Improve Performance: Maintain a normal level of performance and response time to users while heavy transactions are also running in the background.

■ Resource handler: Free resources for important users/actions on the expense of others.

Improve Protection & stability: Protect the DB/OS/Application from heavy transactions/operations that may cause a Denial-of- Service" (DoS). Such processes may consume all resources while leaving nothing to other processes (such as excessive CPU consumption by a looping (executing the same set of instructions a given number of times or until a specified result is obtained) application, antivirus, background procees, etc.).

■ Priorities: Priorities users, clients, queries, etc, in the database/OS/Application based on various factors.

■ Handling peak; Taking a peak demand for resources and distributing it over a greater timescale, and by that, reducing the peak stress' from the system.

Fig. 1 is a general flowchart of controlling the resource consumption of a process, according to a preferred embodiment of the invention. At the first step 101, the processes and transactions in the Operating System are monitored in real-time. At the next step 102, assessment is made to decide whether or not one of the running processes satisfies one or more criteria associated with a need for resource consumption control. If it is, then at the next step 103 the system checks the predefined resource limitations for that process. At the next step 104, the resources for that process is controlled by periodically suspending and resuming it at a predetermined frequency, while each time checking at step 105 if the process is still running, until it is completed. Conventional methods for resource management in which requests of processes for resources are managed without prioritizing more important processes entail a situation where inherent CPU and/or I/O algorithms allocate, inter alia, high amount of resources to processes that increase the load but are not necessarily important. This causes many other processes to get a little amount of resources, until freezing or denial of service. In contradiction, the method proposed by the present invention allows all processes to run and new processes to start running by optimally managing the resources allocated to them such that less important processes get less resources, to thereby free resources for allowing more important processes to run at the same rate or faster, while avoiding denial of any request.

Pig. 2 schematically illustrates a: general architecture of a computerized system, in which resource consumption of processes can be controlled, according to a preferred embodiment of the invention. In this example, several users 201 are connected to an application 202 that runs on a server 203. Normally, resource, allocation is carried out by the operating system, which controls the operation of the CPU 204 and the mass storage devices (the hard disks) 205. According to the present invention, a unique interface 206 is added for controlling the operation of corresponding processes/threads at the operating system level. This interface is adapted to each environment, since the processes/threads may vary from one environment (i.e., Operating System) to another. It is possible to connect to the operating system through its standard Application Programming Interface (API, which is the method prescribed by a computer operating system by which a programmer writing an application program can make requests of the operating system). For example, intervention can be achieved by adding a "stop" command whenever required using the inherent debugging feature (a process of locating and fixing or bypassing bugs/errors in computer program code) which exists in each operating system. The method proposed by the present invention uses the debugger to manipulate and manage resource consumption. For example, several more important processes are self accelerated by causing less important processes to cancel their demand for resources for predefined time periods. As a result, the operating system will allocate the free resources to the other running processes which are more important. This is carried out without actual intervention in the inherent resource allocation mechanism of the operating system (the result is manipulated, rather than the operations of the OS).

Resource control is based on the following steps of identifying and controlling transaction/process within a given environment: Step 1: Collecting statistics raprarriinfl transactions/processes In order to control the resource consumptions of a given process/transaction, the system is monitored by any conventional existing method. For example, in an Oracle database, an internal database views and delivers various factors and statistics regarding transaction (such as username, connected terminal, transaction content, amount of resources used, etc.). Internal commands or even third party tools can be used, for example, in order to collect statistics about running processes/transactions and target those that need their resources to be controlled.

Step 2: Crossing statistics with resource rules The next step is to cross the statistics collected in the previous step 1· to a set of Rules. A rule is defined as a set of thresholds for each gathered statistic/factor. For example, if a transaction is currently active, doing more than 1000 physical reads per second, scanning table "X" or is running during the peak hours (10:00 am - 4:00 pm), its resource consumption is reduced to 20%. According to another example, if a process is executed by the manager of an enterprise, it has been running for 5 minutes and has not been completed yet while other processes were also executed by the employees, then the resource consumption of the other processes (of the employees) is reduced to 20%. According to another example, if a process is an "Anti Virus" that is working while Office tools are running, then its resource consumption is reduced to 5%.

Step 3: Controlling transaction resources usage In this step the transaction resource consumptions is controlled by allocating chunks of execution times. The method of allocating chunks of execution, according to embodiments of the present invention is based on suspending and resuming the processes, which can be done in several ways in a computerized system.

Generally, resource control is based on identifying the exact process/thread at the operating system level that is running and managing the transaction. Then, the process/thread is suspended and resumed in a frequency that is defined by the rule, causing the effect of slowing it down and controlling the amount of resource it uses. The following examples illustrate how to use the method proposed by the present invention in different environments: Byample 1 : limiting resource consumption of one process to 20% in an Oracle database In this example, the process/thread in the Operating System that performs the transaction is identified and suspended it for 4 seconds using OEADEBUG SUSPEND command (through SQLPLUS utility). Then it is resumed for 1 second using OEADEBUG RESUME command. The Suspend/Resume operations are repeated until the transaction is completed. The advantage in database applications is that according the proposed method, a transaction is first identified in the database and then in the operating system. Then, the resources allocated to this transaction are managed in the OS. The result is that the load in the database is affected without managing the database.

Example ¾- limiting resource consumption of one process to 20% in a Unix/Iinux environment In this example, the process thread in the Operating System that performs the transaction is identified and suspended it for 4 seconds using "KILL ~ STOP" command (in the shell environment). Then it is resumed for 1 second using "KILL -CONT" command. The STOP/CONT operations are repeated until the transaction is completed. Alternatively, this may be done by using directly the signal code number. For example : "KILL -23" & "KILL -25" command in Solaris OS (of Sun Microsystems), or "KILL -17" & "KILL -19" command in Free BSD (free and open source operating system that is based on the Berkeley Software Distribution (BSD) version of the Unix operating system) or some Linuxes.

The KILL command is used at the "Shell" (the interactive user interface with UNIX operating system) level. Alternatively, the system calls of a specific OS can be used directly. For example in Unix/Iinux OS the commands "KILLipiiSIGSTOPr & '¾ILL(pid,SIGCONT)" can be used.

Example 3: limiting resource consumption of one process to 20% in a Windows environment In this example, there is an interfacing stage with processes in the Operating System to get their handles, as well as with threads of a specific process, to get its handles. Then the thread is suspended for 4 seconds using "THREAD.SUSPEND" command (in .NET, for example). Then it is resumed for 1 second using "THREAD. RESUME' command (in .NET). The Suspend/Resume operations are repeated until the transaction is completed.

Example 4r limiti g resource consumption of one process to 20% in an I/O (Storage) subsystem p irnpm βτιϊ.

In this example, the storage device is connected and the corresponding I/O stream is identified. Then the I O stream is suspended for 4 seconds and then it is resumed for 1 second. The Suspend/Resume operations are repeated until the I/O stream finishes.

Example κ· limiting resource consumption of one process to 20% in a network interface (such as hubB. switches etc.) In this example, the corresponding stream of IP packets is identified. Then the IP packets delivery is suspended for 4 seconds and then it is resumed for 1 second. The Suspend Resume operations are repeated until the transaction finishes.

Fig. 3A is a graph that illustrates an example of how resources are controlled for a long running process, according to a preferred embodiment of the invention. In the shown example, a single heavy duty process (local disk usage time of a process, expressed in physical reads per second) is executed in a computer, causing a high I/O consumption (shown on the right). The resource consumptions of the process are limited to a level of up to 50%. This limitation is enforced by suspending the process for 5 seconds, and then resuming the process for 5 seconds and repeating the suspend/resume steps until the query is finished (shown on the left). It can be clearly seen that the distribution of the average I/O readings over time is substantially reduced, compared to uncontrolled resource allocation.

Fig. 3B is a graph that illustrates an example of how resources are controlled for a long running process, according to another preferred embodiment of the invention. In the shown example, a single heavy duty process (CPU usage time of a process) is executed in a computer, causing a high CPU consumption (shown on the right). The process is suspended and resumed while limiting the resource consumptions to a level of up to 50% (the CPU usage time with resource control is shown on the left). This limitation is enforced by suspending the process for 5 seconds, and then resuming the process for 5 seconds and repeating suspend/resume steps until the query is finished. It can be clearly seen that the distribution of the average CPU usage time is substantially reduced, compared to uncontrolled resource allocation.

Figs. 4A and 4B are pie-charts that illustrate how important transactions in the system can be given more resources based on the method of controlling and limiting the execution time of other running transactions.

Fig. 4A illustrates the results of controlling the resource allocation in case of two transactions, one of which is defined as an important transaction. In case when the default (uncontrolled) resource management of the Operating system is used, resources are allocated equally to both transactions (left pie-chart), such that all transactions consume the same amount of time from all resources. In case when the controlled resource management of the Operating system is used according to the present invention, 90% of the resources are allocated to the more important transaction, while limiting the resource consumption of the less important transaction to 10% (right pie-chart). Less important transactions are slowed-down by allocating them smaller chunks of execution time in order to free resources time for more important transactions.

Fig. 4B illustrates the results of controlling the resource allocation in case of several transactions, one of which is defined as an important transaction. In case when the default (uncontrolled) resource management of the Operating system is used, resources are allocated equally (20%) to all five transactions (left pie-chart). In case when the' controlled resource management of the Operating system is used according to the present invention, 80% of the resources are allocated to the more important transaction, while limiting the resource consumption of the four remaining less important transactions to 5% each (right pie-chart). In both cases, this is done by allocating smaller chunks of execution time to less important P T IL2007/001621 transactions, to thereby slowing them down, while adding the obtained free execution time to the more important transaction(s).

Fig. 4C illustrates the results of controlling the resource allocation to a transaction, even when not all resources are used. In this case, a specific less important transaction which normally consumes 40% of the resources when there are available free resources (left pie-chart) is also managed according to the method proposed by the present invention, to consume only 20% of the resources (right pie-chart), to thereby free more resources.

Fig. 5 illustrates an intuitive dashboard for showing the status of the system. A System Health Gauge uses displays the comprehensive status of the system using green color for a healthy system and red color for a system in distress. In this example, the resource usage indicators 51 show the level of usage of both CPU and I/O. The health bar 52 shows a system overall managed load of 13%. Virtual resource partition chart 53 shows the degree of accelerated processes. Panes 54, 55 and 56 show the CPU usage history, I/O transactions history and currently viewed running transactions, respectively.

When a transaction has started, the Define Rule window, shown in Fig. 6, can be used for slowing down the system by sHding the virtual throttle slider to the left, in order to prevent the selected transaction from dominating the available resources. Suitable rules may be created for type of transaction.

The above examples and description have of course been provided only for the purpose of illustration, and are not intended to limit the invention in any way. As will be appreciated by the skilled person, the invention can be carried out in a great variety of ways, employing more than one technique from those described above, all without exceeding the scope of the invention.

Claims (21)

22966/npl/07 194654/3 - 19 - fcAMENDED CLAIMS:

1. A method of controlling resource consumption of running processes, sub processes and/or threads in a computerized system, comprising: a) communicating with said running processes through the OS or lower layer program system calls; b) controlling the resource consumption is by directly affecting said running processes and manipulating the operations of the inherent scheduler of said OS or lower level program; c) freeing resources consumed by one or more processes by periodically suspending and resuming said one or more processes; and d) allowing said computerized system to allocate the free resources to other running processes.

2. Method according to claim 1, wherein the other running processes are more important.

3. Method according to claim 1, wherein allocation of resources is carried out by an OS.

4. Method according to claim 1, wherein the running process is a database transaction.

5. Method according to claim 1, wherein the running process is an application transaction. 22966/npl/07 194654/3 - 20 -

6. Method according to claim 1, wherein the amount of resources consumed by a running process is controlled by repeatedly limiting its demand for system resources for a first period during which no demand is presented and a second period, during which the preceding demand is resumed, until said process is finished.

7. A method according to claim 1, wherein in order to control the system resources given to a process, allocation of execution time units is used instead of allocating CPU and/or I/O cycles.

8. A method according to claim 1, wherein suspending and resuming of running transactions/processes is carried out transparently to other entities of the computerized system.

9. A method according to claim 8, wherein suspending and resuming of running transactions/processes is carried out externally to the OS without intervention in its inherent resource allocation mechanism.

10. A method according to claim 1, wherein one or more processes is accelerated by slowing down other processes. 22966/npl/07 194654/3 - 21 -

11. A method according to claim 10, wherein processes are suspended by causing said processes to demand fewer resources.

12. A method according to claim 1, wherein running processes are accelerated or slowed down by using a GUI for controlling the resources allocated to said running processes.

13. A method according to claim 11, wherein the GUI includes a sliding virtual throttle.

14. A method according to claim 1, wherein suspending is carried out through an inherent API.

15. A method according to claim 1, wherein suspending is carried out through a debugger.

16. A method according to claim 1, wherein suspending is carried out directly through OS system calls or using a binary code or any other representation of a command for suspending and resuming processes.

17. A method according to claim 1, wherein the amount of consumed resources is controlled for a process that is already running and consuming resources. 22966/npl/07 194654/3 - 22 -

18. A method according to claim 1, wherein the amount of resources consumed by an application is controlled by a set of rules, which can be determined externally to said application.

19. A computerized system, in which resource consumption of running processes, sub processes and/or threads are controlled at its OS level by: a) communicating directly with said OS through its native system calls; b) controlling the resource consumption is by directly affecting said running processes and manipulating the operations of the inherent scheduler of said OS; c) freeing resources consumed by one or more processes by periodically suspending and resuming said one or more processes; and d) allowing said computerized system to allocate the free resources to other running processes.

20. A method according to claim 1, wherein the control is done via the OS, a Virtual Machine or hardware embedded software.

21. A method according to claim 1, wherein communication with the running processes is made directly through the OS or lower layer program system calls. LUZZATTO & LUZZATTO