Patent Application
20120254295
This application claims priority from Japanese patent application No. 2011-072236, filed on Mar. 29, 2011, which is incorporated herein by reference.
1. Field of the Invention
The present invention relates to a service request reception control method, an apparatus and a system, which detect an overload state of service based on the time required for the service provision, and refuse a reception of a new service request.
2. Description of the Related Art
In a server client system that processing requests are given from a large number of clients to the server, independently, if the processing request exceeds the processing capacity of the server, a problem such as a server performance degradation or a system down may happen. To prevent a fall of service quality and the system down due to the overload state, several methods to refuse the reception of a surplus service request are suggested.
The first method is to handle with a software design. This method is to set prospectively a quantity of service request that is possible to receive in the server system, and to refuse the larger service request than it. For example, there is a web login control system limiting the number of the login by comparing the number of the current login with the number of the receivable login (Japanese patent publication No. 2008-242676). Also, there is a regulation method against several kinds of service start requests, the method manages the request by digitizing “load” per the requests and comparing it with a threshold (Japanese patent publication No. 2008-283307).
The second method is to handle with the observation of load situations of resources. This method is to monitor the rate of utilization of physical resources, which constitute the server device and the system, such as CPUs, memories, networks, etc., and to refuse the reception if the load exceeds the threshold. For example, there is a reception control method of the server computer monitoring the load state of the resource every function of the server, and comparing it with the threshold (Japanese patent publication No. 2000-322365). Also, there is a call reception control apparatus and a call reception control method monitoring the CPU utilization directly (Japanese patent publication No. 2004-048580).
However, the prior art 1 is not usable when a bottleneck is not clear, and it is necessary to design to all bottleneck candidates. Further, because the upper limit is decided without depending on a state of the hardware resource, there is a possibility that the hardware resource cannot be used up enough by setting the upper limit level in the design too low. In other words, the performance upper limit is limited with the upper limit level in the design rather than the hardware performance. On the contrary, the hardware resource becomes the bottleneck by setting the upper limit level in the design too high, there is a problem to have possibilities to fall into the overload state.
Also, the prior art 2 must know what hardware resource becomes the bottleneck prospectively, there is a problem that the design is difficult. Also, if the software implementation cannot finish up the hardware performance enough, there is a problem that the detection of the tightness of the hardware resource is difficult (an example: in the case of multi-core CPUs, the utilization upper limit can vary by the software implementation).
Therefore, it is an object of the present invention to realize the service request reception control method, the apparatus, and the system with the reception refusal, (1) which are possible to finish up the hardware resource without knowing the available quantity of hardware resource prospectively, and (2) which are possible to detect the reception of the surplus service request without knowing the bottleneck reason prospectively.
To realize the above object, according to a service request reception control method of the present invention, a service request reception control method for reception control in a server device receiving a processing request from a plurality of client apparatuses comprises the steps of: calculating a mean time required for a series of processing to the request from the client apparatus; determining a load state of service by comparing the mean time with a targeted service processing time; and refusing a request reception from the client apparatus based on the load state.
Further, it is also preferable that said calculating step calculates the mean time from a start time of the processing of the request from the client apparatus and an end time of the processing of the request from the client apparatus, the start time being a received time of a reservation request message in a two phase commit technique and the end time being a transmitted time of a commitment response message in the two phase commit technique.
Further, it is also preferable that said refusing step only selects the reservation request message in the two phase commit technique and refuses it.
Further, it is also preferable that said determining step determines again after refusing the request reception, when a predetermined number of the requests during processing are finished or when all requests are finished.
To realize the above object, according to a service request reception control apparatus of the present invention, a service request reception control apparatus for reception control in a server device receiving the processing request from a plurality of client apparatuses comprises: a means for calculating a mean time required for a series of processing to the request from the client apparatus; a means for determining a load state of service by comparing the mean time with a targeted service processing time; and a means for refusing a request reception from the client apparatus based on the load state.
To realize the above object, according to a service request reception control system of the present invention, the system comprises a plurality of client apparatuses, a server device receiving a processing request from a plurality of said client apparatuses, and a service request reception control apparatus described above.
According to the feature “which is possible to finish up the hardware resource without knowing the available quantity of hardware resource prospectively”, the present invention has the advantage to prevent a surplus and a lack of the hardware resources, which occurs by deviating a reception upper limit value in a design set by OS or software from the available quantity of hardware resources.
According to the feature “which is possible to detect the surplus service request without knowing the bottleneck reason prospectively”, the present invention has the advantage to prevent the overload without depending on the hardware configuration at the time of the system introduction.
An embodiment of the present invention will be described below with reference to the drawings in detail.
The message processing unit 11 has a function to receive a request message from the client 2 and to extract a classification (it is the reservation request or the commitment request, or the other control messages) and a parameter of the message. Also, the unit 11 has a function to transmit a reply message to the client 2.
The request admission judgment unit 12 decides whether the reception of the reservation request is possible based on the current load situation. When the reception of the reservation request is possible, the unit 12 notifies to the control processing unit 14 so as to carry on the reservation request. When the reception is impossible, the unit 12 replies a negative reply. Also, as for the commitment request, the unit 12 notifies to the control processing unit 14 regardless of the load.
The processing time monitoring unit 13 accumulates a history of a time from the reception of the reservation request to the transmission of the commitment response (=the processing time), and notifies the load situation based on the processing time to the request admission judgment unit 12. When it is detected that the mean processing time exceeds than the threshold (the targeted service processing time), which is set as the target prospectively, the unit 13 notifies so as not to receive the further reservation request. Also, when a reception refusal state can be released, the unit 13 notifies to the request admission judgment unit 12 so as to receive the reservation request again (the cancellation of the reception refusal state will be described in the following in detail).
The control processing unit 14 controls the control object devices 3 based on the request. The unit 14 replies the success or failure of the processing to the message processing unit 11, after waiting all replies from the control object devices 3.
When the refusal of new request is valid, the unit 12 replies a negative reply as it is impossible to receive the reservation request in the current load situation.
When the refusal of new request is invalid, since it is possible to receive, the unit 12 notifies a reservation request reception time to the processing time monitoring unit 13, and hands the parameter of the request message of the reservation request to the control processing unit 14.
In the case of the request that cannot be controlled by parameter abnormality, as control impossibility, the unit 14 replies a negative reply to the message processing unit 11. The message processing unit 11 notifies this negative reply to the clients 2. Note that, in this case, the unit 14 notifies a stop of the processing time measurement to the processing time monitoring unit 13.
In the case of the request that be controllable, it goes to the next processing.
From the (1) to the (9), a series of processing from the reservation request from the client 2 to the commitment response is performed. Independently with it, the processing time monitoring unit 13 performs a threshold excess judgment of the mean processing time and a measurement of the remaining request number as follows.
In the processing time monitoring unit 13, the mean processing time is detected as follow.
The processing time of one request is the time from the reception of the reservation request message to the transmission of the commitment response. However, to take into account the remaining request, the processing time of the request during processing is the larger one between “from the reception of the reservation request to the current time” and “the mean processing time by the present”. For example, in
To remove the processing time of an extremely old request, the calculation object request of the mean processing time is the smaller one between “the most recent N requests” and “the number of the requests received in the most recent Trec seconds”. In the case of
In the processing time monitoring unit 13, the cancellation of the reception refusal state is carried out as follows. When either following is satisfied, the processing time monitoring unit 13 confirms whether the reception refusal state can be released.
In this case, the processing time monitoring unit 13 compares the mean processing time due to the most recent data and the threshold. If the mean processing time is less than or equal to the threshold, the reception refusal state is released, if the mean processing time is larger than the threshold, the reception refusal state is maintained (it waits for the completion of the M requests or all remaining requests).
Note that, a server administrator sets the various control parameters (Tth, Trec, N, M) described in the above to the control request reception server 1 prospectively manually.
The present invention is applicable to the control server system of the network apparatus. The control server of the network apparatus manages a plurality of apparatuses with one server, and a concentration of setup requests may occur. Also, since there is a possibility that the time for the control depends on the overload out of the server (the apparatus of the control object), it is not easy to estimate the bottleneck. Therefore, the present invention functions effectively.
As described above, the present invention monitors the processing time of the request from the clients, detects the degradation of the processing performance, and performs the reception control. Thus, it is easy to detect the performance stringency without being conscious of the point of the bottleneck (it does not need the conscious of characteristics of the hardware resources usage in the software implementation), further, the hardware performance of the system can be used up without being limited by the upper limit level in the design. Especially, the present invention can be applied to the server system that is hard to understand the bottleneck (examples: multistage server, apparatus control server) easily.
All the foregoing embodiments are by way of example of the present invention only and not intended to be limiting, and many widely different alternations and modifications of the present invention may be constructed without departing from the spirit and scope of the present invention. Accordingly, the present invention is limited only as defined in the following claims and equivalents thereto.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2011-072236 | Mar 2011 | JP | national |
