Distributed Process Management

Thebestindonesia.com – Process Migration sends a certain number of processes from one machine to another for an execution process on the target machine. The background of this concept stems from research aimed at balancing the load between several computer network systems.

Purpose of the Migration Process

Process migration is significant in distributed systems for some reasons, including:

1. Load sharing (burden sharing)

By moving processes from a fully loaded system to a less heavily laden system, the loads can be balanced to improve system-wide performance.

2. Communication Performance (communication performance)

Processes that interact intensively with each other can be moved to the same node to reduce the communication costs incurred by their interaction. Likewise, when a process performs data analysis on a file or group of files that exceed the size of the process, it is advantageous to move the process closer to the data than vice versa.

Read also : Computer Networks and Distributed Systems

3. Availability (availability)

Processes that take a long time must be moved to continue on another system if the system used at that time will be turned off for specific reasons.

4. Utilization of Special Ability

A process can be moved to benefit from using software or hardware at a particular node.

Read also : Computer and Network Security 

Migration Process Mechanism

Some things to consider in designing migration process facilities include the following:

  • Who started the migration.
  • Which parts of the process need to be moved.
  • What happens to the signals and messages that migration generates.

Start Migration

The factors determining who will initiate migration depend on the purpose of the migration facility.

If the migration is to balance the load, then several modules in the operating system that monitor system load are responsible for determining when the migration should occur. This module will perform the following tasks:

  • Give a signal to the process intended to migrate.
  • Determine the location of the migration destination. For this reason, the module must communicate with similar modules on other systems to monitor the load description on other systems. If the goal of migration is to move a process closer to a particular resource, then the process can migrate on its initiative.

What was moved?

During the migration process, what needs to be done is to delete the process on the source system and create a circle on the destination system. Because the migration process is not a process replication, it must also be moved. Consequently, the process image, which stores process block control information, must also be transferred. In addition, every link between the process and other processes (in the form of messages or signals) must be updated.

Before Migration Process
After Migration Process

Process 3 on machine P is migrated to process four on machine Q. All links associated with that process point to the same target before and after migration. The operating system performs the link capture task.

Read also : Communication Network

The transfer of the process control block is carried out directly. The most complicated part of this section is related to the process space and the open files associated with the process. For this reason, a particular strategy is needed to handle it. Two alternatives can be used:

  1. Migrate the entire address space during migration. This is a cleaning fix. Thus there are no process traces left on the old system/machine.
  2. Only move the part of the address space that resides in memory. Additional blocks of the virtual address space will only be transformed if they are needed. With this strategy, the amount of data transferred can be minimized. But on the other hand, the source system maintains some of the process address segment table entries.

If the process being migrated is not expected to use too much nonresident address space (for example, a method that only temporarily switches to another machine to perform a function on a file and then returns to the source machine), then the second strategy may be a more efficient option. However, if the required non-resident address space is too large and will be accessed frequently, then a second strategy is needed to handle it.

Tinggalkan Balasan

Alamat email Anda tidak akan dipublikasikan. Ruas yang wajib ditandai *