AU2002330651A1 - Object distribution - Google Patents

Description

Object Distribution

The invention relates to a method of locating objects in a distributed electronic system.

In certain distributed systems, in which system components may also be geographically distributed, the problem of deciding where to locate, or home, objects within the system arises. These objects may have geographic dependencies themselves, as may further objects which relate to the first set.

Generally the problem of identifying where an object should be homed and what the optimum configuration with respect to network and processor efficiency is a challenging one. This problem is further compounded when this object relates (with processing dependencies) to a number of further objects within the system, which further objects are diversely located across the system themselves.

US5787284 discloses a method in which programs are grouped together based on the weights of connections, i.e. the execution count between procedures, between the objects and their costs. System-imposed constraints on memory size can be taken into account to avoid creating groupings that overload system capacity. Such a method is generally applicable only in a processor environment and does not deal adequately with the restrictions imposed by operation in a distributed environment.

The present invention seeks to provide a method of distributing objects in an efficient manner with respect to one another.

According to the invention there is provided a method of locating objects in a distributed electronic environment comprising defining a plurality of object-types, which object-types are assigned a plurality of attributes, one attribute being an object- precedence attribute, defining a plurality of home locations for objects, wherein when a new object is instantiated, which object has associations with other objects, the new object is located at the home location of the existing object which has the highest precedence value of all the associated objects. Preferably, the value of the object-precedence is unique for each object-type. Preferably, the precedence value is determined such that objects having a relationship with other objects are located in proximity to one another.

In the method of the invention objects which are related to each other are advantageously arranged in proximity to each other such that overheads associated with processes which reference these objects are minimised. This approach improves the efficiency of processing throughout a system of disparate processors in comparison to the conventional techniques of not organising these relationships, such as a random placing of object through the distributed system or the organised homing of all objects of a given type at a single home.

An exemplary embodiment of the invention will now be described in greater detail with reference to the examples

In a conventional object oriented environment, each object within the system is of a particular class. Following a standard object oriented approach, each object within the system is an instance of such a class. This means that a base description, or template of an object is provided as a class definition, and from this template, an object has been created or instantiated. This object then takes on the properties of this template or object-type.

As part of the object-type definition, an attribute known as object-type precedence is defined. The object-type precedence is defined as having a integer value, which should be unique. Therefore, each and every object within the electronic system is assigned a predetermined object-type precedence value based upon what object-type they are instantiated from. The value of the object-type precedence attribute is that assigned to the object-type precedence attribute within that class. The method of assigning values to the object-type precedence attribute may be manually determined by entering specific values, or may be automated by means of a unique number generator.

The respective values of the pre-determined object-type precedent values are determined by the location of an initial number of objects within the system. These objects may not have any relationship to one another. Once these initial objects have been 'homed', they provide the basis for assigning 'homes' to newly defined objects which relate to them within the system. The intention of this is to reduce interprocess communication and associated overheads, thus increasing efficiency compared to a simple pooling of objects of the same type together at the same processing location.

In an electronic system, it is known to use a policy based management system. A policy is an administrator-specified directive that manages certain aspects of the desired outcome of interactions within a given system. This system may provide an environment for users, applications, services, and infrastructure or any other quantifiable object. A policy provides guidelines for how these different entities within such an environment should be managed in response to interactions within the system.

In the following example, it is assumed for sake of clarity that only one object per object-type may be associated with another object.

It makes sense to 'home' Policy 1 with at least one of the objects which already exist in the system which Policy 1 has associations with (processing dependencies). This is not association through existing 'homes', but is association through precedence value, i.e. it is the precedence value which determines which existing object the new object should be homed with. The home is determined by looking at where that existing object currently resides. An object 'Policyl ' might be associated with the following objects:

Policyl : ObjectA, ObjectB, ObjectC

Each of these objects is of the following object-types:

ObjectA: Object-typeXX ObjectB: Object-typeYY ObjectC: Object-typeZZ

Where each object-type has the following object-type precedence values assigned to them: Object-typeXX: 1 Object-typeYY: 3 Object-typeZZ: 2

Also, the objects have already been assigned 'homes' at the following locations:

ObjectA: located at HomeN ObjectB: located at HomeO ObjectC: located at HomeP

As a result of this, Policy 1 is homed at Home N, if another object is instantiated and is associated with Policy 1 , then it will also be homed at HomeN even if the new object has a higher precedence value itself. This is determined by the precedence value of the existing object with which the new object is associated with rather than the precedence value of the new object. The object Policyl is will be 'homed' or located at HomeN since, of the objects which are associated with Policyl, Object- typeXX has the highest precedence. Therefore since ObjectA is of Object-typeXX and is located at HomeN, therefore Policyl will be also located at HomeN.

The example assumes that '1 ' is the highest precedence value - it need not be, it alternatively could be that the higher the value, the higher the precedence.

It is not a problem to associate more than one object of the same type with another specific object. This simply results in both of the new objects of the same type being homed at the same server.

Claims (3)

Claims

1. A method of locating objects in a distributed electronic environment comprising defining a plurality of object-types, which object-types are assigned a plurality of attributes, characterised in that one attribute is an object-precedence attribute, wherein a plurality of home locations for objects are defined, and in that when a new object is instantiated, which object has associations with other objects, the new object is located at the home location of the existing object which has the highest precedence value of all the associated objects.

2. A method according to Claim 1, wherein the value of the object-type- precedence is unique for each object-type.

3. A method according to Claim 1 or Claim 2, wherein the precedence value is determined such that objects having a relationship with other objects are located in proximity to one another.