JP2000224171A - Network state measurement method and apparatus using multiple services - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a system that can measure whether a change in the reply time in a network service is caused by a network or a server only at a client side, without using special devices, and to provide its device. SOLUTION: A client 1 is provide with a Ta measurement section 11 an a Tb measurement section 12, that transmit a request instruction to services A, B of a server 2 to measure respective reply times and with a discrimination section 13 that compares the reply times of the services A, B to discriminate whether the cause to a reply delay results from a network or the server so as to discriminate the cause to a change in the reply time without the need for special devices at the server side.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for measuring a network state in a network service by a client-server type system on a client side and an improvement of the system.

[0002]

2. Description of the Related Art In recent years, with the spread of the Internet, W
2. Description of the Related Art Network services based on client-server systems such as WW services are widely used.

A client-server type system is
Computer using the service (client)
And a computer (server) that provides a service via a network, and manages each resource such as data by the server. When a client uses a service, data is exchanged with a server via a network according to a predetermined procedure called a protocol.

At this time, there is some time difference (response time) between the time when the client issues a request and the time when the response to the request returns to the client, due to the time required for transmitting the data through the network path and processing the data at the server. Thus, as a characteristic of the client-server type system, the response time changes with time. This may be caused by a plurality of factors, such as a change in time required for data to be transmitted on the network path, a change in processing time in the server, and the like.

The response time is defined as (response time) = (transmission time on network) + (processing time on server), where the total time required for processing the network service by the client / server type system is taken as the response time. Can be roughly divided into two: network factors and server factors.

FIG. 1 shows how a client-server system performs a network service. In FIG. 1, time T ₁ is the time at which the client sends a data request, and T ₂ is the time at which the server sends this data request. time of receiving the, T ₃ is the time when the corresponding data server sends the request, T ₄ is a time when the client receives the data.

[0007] The main factors that determine the time (T ₂ -T ₁ and T ₄ -T ₃ ) for transmitting the network route are: Factor 1: Network bandwidth (how much data is stored in a unit time) Factor 2: the processing performance of the router that controls the packet (the time required for control / transfer processing); factor 3: the amount of data on the network at that point in time. Among the above three factors, factor 3 is the one that is most likely to change with time.

The factors that determine the time required for data processing in the server (T ₃ −T ₂ ) are: • Factor 4: server processing performance; • Factor 5: a program that provides services on the server is accepted.・ The amount of data being processed (transaction amount) is mainly cited. Of these two, the factor 5 is likely to change with time.

Conventionally, there has been considered a method of obtaining various network service response times by performing various measurements from the client side.

The basic idea is to record the time when the client device sends data to the server, and calculate the difference from the time when the response to the data is returned. . As this method, an Internet Protocol (ICMP) (Internet Content Control Protocol) is generally used on an IP protocol currently generally used in the Internet.
A method called Ping, which measures a response time by using a packet of a protocol called “Rollmanagement Protocol”, is generally known.

However, in practice, the response time of ICMP and the response time of other services are often different,
In many cases, the response time of Ping does not immediately indicate the response time of another service on the server. Also, on protocols that provide other common services,
T ₂ on the server side in order to obtain a T _2, T ₃ in FIG. 1,
Not incorporated a special mechanism for measuring T ₃ as well, the only client-side device was difficult to determine the breakdown of response time.

[0012]

There are the following problems in the conventional Ping and the calculation of the network service processing time by the same type of system. -The response times of the ICMP providing the ping and the protocol of another service do not always match and often differ. -On a general protocol other than ICMP, it is necessary to incorporate a special mechanism to acquire the processing time on the server side. Therefore, in the measurement by the device on the client side only, the time required for processing the network service (response time) cannot be classified into the arrival time of the packet and the processing time of the server.

An object of the present invention is to determine whether the network state in a network service provided by a client-server type system, that is, whether the response time changes due to an intermediate network path or the service providing server. The present invention provides a method and a device capable of measuring only on the client side without requiring a special mechanism.

[0014]

SUMMARY OF THE INVENTION The present invention solves the above problem by focusing on a plurality of services and a plurality of protocols running on a server.

Basically, it is published on a server,
Focusing on the fact that there are many services running in addition to the services that are receiving a large number of accesses, sending packets to each of the multiple services, and the response time change factors Is a network or server.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Generally, a network model has a hierarchical structure as shown in FIG. In this structure, each layer exists only to provide a function to a higher layer. As a result, there is an advantage that even if a specific function is changed, the operation of the entire model is not affected, and even when a new service is created, it is not necessary to create the entire service from a lower layer.

In a client-server system, a server can operate a plurality of services on a single protocol. At that time, there is an example in which a number called a port is used to distinguish each service on the same protocol (eg, TCP / I
P).

In general, the number of services that can be provided simultaneously and stably is limited due to the processing performance of the server and the specifications and capabilities of programs running as services. If the number of connection requests exceeds the limit, the service becomes overloaded, causing a waiting time until the process is accepted. Problems often arise.

From the viewpoint of load distribution and the like, a server that is expected to receive a large amount of access often limits the services to be provided. However, there are cases where other services are running in addition to the one service provided by the server.

At this time, the time required for the response processing of a service with a large number of accesses, which is the main purpose of providing by the server, is different from the time required for the response processing of another service with a low concentration of accesses. .

FIG. 3 shows the influence of concentration of access to the server. Here, assuming that a service that is disclosed to the outside as a service mainly provided by a certain server, for example, X is A, and services that are implemented but are not mainly provided are B, C, and D. May cause a delay in response time due to the concentration of accesses. That is,
The response time of the service A is (response time) = (transmission time in the network) + (normal processing time in the server) + (delay time of processing caused by access concentration). On the other hand, the services B, C, and D have a small number of accesses, and connection requests exceeding the connection limit hardly occur.

At this time, a change in the response time of the service A,
The main factor of the difference between the change in the response time of the services B, C, and D is the factor resulting from the aforementioned factor 5. Using this, the network factor and the server factor as the breakdown of the response time are calculated.

FIG. 4 is a system configuration diagram showing an example of an embodiment of a network state measurement system using a plurality of services according to the present invention.

In FIG. 4, reference numeral 1 denotes a client that uses the service, 2 denotes a server that provides the service, and 3 denotes a network that connects the client 1 and the server 2.
The client 1 includes a Ta measuring unit 11 for measuring a response time Ta of the service A, and a response time Tb of the service B.
Tb measuring unit 12 for measuring the measured Ta and Tb
The server 2 includes a service A providing unit 21 that provides the service A and a service B providing unit 2 that provides the service B.
2 is provided.

When the number of accesses to the service A is large and the number of accesses to the service B is small, the breakdown of Ta and Tb is Ta = 2N + _Swait + _{Sexe # a, and} Tb = 2N + _{Sexe # b} as shown in FIG. . Here, S _{exe #a} and S _{exe #b} are the processing times of services A and B, respectively. S _wait is the waiting time (delay time) of the service A process caused by the access concentration. N is a one-way network transmission time, and 2N is required for a round trip to receive a service. It is assumed that the transmission time of the network can be represented by a value of N without abrupt change in a short time.

When the difference between the processing times of the services A and B can be considered to be sufficiently smaller than the variables as other factors, or when the measurement method for each service is taken so as to satisfy such a condition, that is, S _{exe # a} = _{Sexe # b} = _Sexe , then Ta = 2N + _Swait + _Sexe Tb = 2N + _Sexe Taking the ratio of Ta and Tb gives Ta / Tb = 1 + S _wait / (2N + S _exe ).

N and S _wait are large, and _Sexe is Ta / T
In a model that does not show an important element as the value of b, S _exe ≪
Since N and S _wait , Ta / Tb ≒ 1 + S _wait / 2N is obtained. The numerical comparison between the response time S _wait of the server factor and the response time N of the network factor is S _wait / 2N = 1−Ta / Tb, from which it is possible to determine whether the change in response time is due to the network or the server. FIG. 6 shows a flowchart of the measurement in this example.

In an actual application on the Internet, service A is provided by http which provides a WWW service, service B is provided.
May be considered as ICMP.

[0029]

As described above, according to the present invention,
No special mechanism is required on the server side, and it is possible to determine whether the cause of the service response delay is caused by the network or the server by measurement using only the device prepared on the client side.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram of a breakdown of a response time required for a network service process.

FIG. 2 is an explanatory diagram of a network model.

FIG. 3 is an explanatory diagram showing an influence of concentration of access to a server;

FIG. 4 is a system configuration diagram showing an example of an embodiment of the present invention.

FIG. 5 is an explanatory diagram of a response time for each service.

FIG. 6 is a flowchart showing a measurement process according to the method of the present invention.

[Explanation of symbols]

1: client, 2: server, 3: network, 1
1: Ta measuring unit, 12: Tb measuring unit, 13: response delay factor determining unit, 21: service A providing unit, 22: service B
Provision department.

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5B089 GA11 GA21 GB09 MA07 5K030 GA11 HA04 MB06 MC00 9A001 BB02 BB03 BB04 CC06 CC07 DD10 JJ25 JJ27 KK37 KK56 LL08

Claims (2)

[Claims] In a client-server system in which a client and a server are connected via a network, a response time is measured by sending a request command for a plurality of services from the client to the server, and each service is measured. A network state measurement method using a plurality of services, which determines whether the cause of the response delay is caused by the network or the server by comparing the response times of the network conditions. 2. In a client-server type system in which a client and a server are connected via a network, measuring means for sending a request command for a plurality of services to the server and measuring a response time on the client side. And a determination means for comparing the response time of each service to determine whether the cause of the response delay is caused by the network or the server.