Electrical computers and digital processing systems: multicomput – Distributed data processing – Client/server
Reexamination Certificate
1999-04-09
2002-05-21
Barot, Bharat (Department: 2154)
Electrical computers and digital processing systems: multicomput
Distributed data processing
Client/server
C709S201000, C709S202000, C709S227000, C709S238000, C709S241000, C709S241000, C707S793000, C707S793000
Reexamination Certificate
active
06393458
ABSTRACT:
STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH
Not Applicable
FIELD OF THE INVENTION
The present invention relates to load balancing and more specifically, to static and dynamic load balancing within a distributed object architecture.
BACKGROUND OF THE INVENTION
Sophisticated financial transaction, manufacturing and data processing systems are increasingly being implemented in computerized modern installations. In a typical large installation, hundreds and sometimes thousands of machines, transaction types, objects or nodes may be defined in a distributed computing environment or infrastructure.
A simplified typical distributed object computing environment
10
is illustrated in FIG.
1
. The distributed object computing environment
10
includes a plurality of machines
12
. Machines
12
may be actual computers, such as work stations or PCs commonly known in the art or any other computers useful as either server or client machines, special purpose machines, or virtual machines. As used herein, the term “host” generally refers to an actual computer that may be executing a plurality of virtual machines, processes or threads. In the context of a manufacturing execution system, special purpose machines might include bar code software devices that operate on a computer, but appear to software that runs on and interacts with the virtual machine to be a complete computer. A common example of a virtual machine known in the art is the Java virtual machine, however, other types of virtual machines are available and may be used herein.
It is common that in these large modern installations that only a portion of the machines, nodes and/or objects in the network may be used at any one time. Thus, a common problem with these complicated infrastructures is the potential for overload or bottlenecking at any one node or object and the resultant inability of that node or object to accept or process further transactions or events. This condition can occur if, for example, the node experiences unusually high transaction activity due to an unexpected event. In such systems it is vital that processing in the system occur with maximum efficiency while still affording maximum flexibility for future expansion and modification. In order to achieve this efficiency and prevent overloading, load balancing must be undertaken within the system.
Basically, load balancing is the reallocation and reapportionment of resources within a system to achieve a uniform or balanced distribution of work or load within the system. Load balancing can be generally classified as either static or dynamic. In static load balancing, the required events or processes occur regardless or independent of the current system loads or state. Commonly, static load balancers are manually implemented by a system operator who observes how work is apportioned among the nodes in a system and based on these observations, and re-apportions the work to even out the load. Static load balancing is relatively uncomplicated to implement but poses problems in operation because static load balancing disregards changes and fluctuations in the current system state. Essentially, when the system is in operation, the balancing is fixed and thus the system cannot respond to unusual circumstances or changes in the system's usage patterns. If changes or modifications are required, operator intervention is required to incorporate any changes in the system.
In dynamic load balancing, the status and conditions of all nodes or members of the system are accounted for on a routine basis in order to decide where to process events or tasks based on the current usage in the system. However, dynamic load balancing is inherently more complicated and requires significantly more computing power than static load balancing since it requires constant maintenance of the states in the system.
Further, modern requirements demand that any load balancing system be scalable. Scalability provides a system a way to limit the degradation of system response time as new members, nodes or objects are added, removed or modified within the existing infrastructure.
Referring to
FIG. 2
, a prior art load balancing system
20
for distributing requests from client objects to server objects is shown. The load balancing system generally is divided into a client side
22
and a server side
24
. The client side
22
may include a number of client or calling objects
26
,
28
and
30
. These client objects interact with a single load balancer object
32
on the server side
24
. The load balancer
32
then communicates with a number of server or receiving objects
34
,
36
and
38
. In the prior art, the single load balancer object
32
is subject to bottlenecking since all events are passed through this single load balancer. In operation, the load balancer object
32
will redirect incoming events or calls from the client objects to specific server objects. Essentially, the load balancer
32
will determine which specific server object should receive the event and then direct that event to that specific server object. The receiving server object will then process the event or call which was received. Such a system is both inefficient and cumbersome because it implements two objects, the load balancer and server object, which have the same interface which results in maintenance and operating problems involved with the duplicative code. Furthermore, since the load balancer is implemented on the server side, the potential for overloading the server is increased.
Accordingly, it would be desirable to have load balancing which incorporates the benefits of both static and dynamic load balancing and is infinitely scalable in an object oriented environment.
SUMMARY OF THE INVENTION
The invention provides a method, system and computer program product for load balancing in a distributed computing environment. The system for balancing the distribution of event messages in a distributed object computing environment includes at least one client publishing an event containing information and a plurality of server classes residing on one or more server hosts, at least one server class subscribing to the event published by the client, and a plurality of load balancers. Each load balancer queries the server hosts to calculate a load parameter for each server host.
The system provides for the registering of a plurality of server classes as subscribers for selected events. For each subscribing server class, the system also provides for the registering of one or more server hosts as capable of running an instance of the class. The client then publishes an event which is received by only one of the plurality of load balancers.
The system then provides for the selection of a server host for each subscribing server class registered to subscribe to the event by the load balancer based on load parameters calculated and the registration of hosts capable of running an instance of the subscribing server classes. The selected server host identity for each subscribing server class is then coupled to the event for further publication of the event into the distributed object computing environment by the load balancer. The event is then received by a plurality of event subscribers, each event subscriber reviewing the server host identity information for a server host affiliated with that event subscribers. The event subscribers pass the event on to the identified subscribing class for each class identified for processing on a server host affiliated with the event subscriber. At least one instance of each subscribing class for the published event then performs logical operations in accordance with the event.
In another embodiment, each server host may also be assigned a certain power rating. The load balancing object then takes the current load of the server host and divides the load by the power rating to determine a relative load.
REFERENCES:
patent: 5603029 (1997-02-01), Aman et al.
patent: 5790789 (1998-08-01), Suarez
patent: 5794005 (1998-08-01), Steinman
patent: 5847957 (1998-12-01), Cohen et al.
patent:
Boole Ward L.
Gigliotti Samuel S.
Barot Bharat
Genrad, Inc.
Nutter & McClennen & Fish LLP
LandOfFree
Method and apparatus for load balancing in a distributed... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for load balancing in a distributed..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for load balancing in a distributed... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2867302