p2 p grid
Download as PPTX, PDF1 like373 views
The document proposes a Fault Tolerance Policy called FTCPM that uses checkpointing to improve reliability in peer-to-peer grid systems. It introduces a checkpoint strategy that saves job state to stable storage. If a failure occurs, it uses timestamps and rollback to identify the cause and reschedules the job to the correct node using an algorithm called SMF. The system architecture is layered, with services like job management, fault tolerance, and SMF. It also details CP and SMF algorithms that schedule checkpoints dynamically and select the most fitting resource for job assignment.
p2 p grid
- 2. In Grid Computing In grid computing environment, resources are made available across geographically distributed locations. Since resources are distributed there may a chances of failure. Techniques for handling failures.
- 3. Abstract: To reduce the probability of failures a checkpoint strategy is introduced. The checkpoint strategy works along with a timestamp. Delay in time stamp- Failure occurs, checkpoint triggers a roll back mechanism to identify origin and cause of failure. Rectifies and the process is rescheduled to the correct node by using SMF( select most fitting resource).
- 4. Peer-to-Peer (P2P) computing The robust availability of distributed resources is a very important issue. P2P grid system has to guarantee the correctness when faults occur. To improve reliability, fault-tolerant mechanism for such systems is mandatory. Although extensive fault tolerance policies were proposed but the fact that mechanisms was seldom taken into consideration.
- 5. Fault Tolerance Policy on Check Point Monitoring (FTCPM) Here we are proposing a Fault Tolerance policy – FTCPM. For improving the system reliability, FTCPM duplicates jobs into different sites to tolerate failures. Check pointing is a process to save the state of a running application to a stable storage that can be used to resume execution of the application in case of any failures. Thus reduce the execution time.
- 6. Previous Work: Fault Tolerance policy on Dynamic Load Balance: In order to provide the uninterrupted services, there were many proposed works. Fault tolerance mechanisms include the redundant strategy uses redundant tasks to carry out concurrent computations on different nodes for keeping a higher availability. When a failure occurs, the system will take a redundant copy of the failed job from another node. Such an approach wastes resources.
- 7. In the P2P grid system: each site has a backup storage two queues, named Ready Queue (REQ) and Running Queue (RQ). Each site regularly sends "heartbeat" messages to its neighbour site. neighbour site fails. The fault tolerance policy is triggered.
- 8. Proposed System: Check pointing Techniques: New way approach of check point technique is introduced. The checkpoint strategy which helps to keep the job’s turnaround time stable. Timestamp delay - rollback process takes place. The check pointing mechanism reschedules the job to the correct node using the SMF (Select Most Fitting Resource) algorithm. Improves the system reliability, can achieve better job execution and job completion rate, and thereby reduce failure probability.
- 9. System Architecture: Organized in a layered architecture. The job allocation testing data The job allocation services are MM, EM, configure, FT, SMF, FT, Finding fault tolerance node is based on 3 mechanisms.
- 10. CP (check pointing) Algorithm: The algorithm schedules check pointing interval in a dynamic way: 1. Schedule initial check pointing at short time period after job starts. 2. Next schedule depends on machine mean failure interval by topping up (inew= Iold+Iinit). 3.If inequality remaining execution time<mean failure interval & Inew< a* execution time is false then new checkpoint interval decreased to Inew= Iold-Iinit.
- 11. SMF Algorithm (Select Most Fitting Resource) 1. Resources register to GIS( geo info system) – Machines are grouped into L discrete levels based on node speed. 2. User submits jobs with resource requirement and closeness factor. 3. Starting from smallest level, it searches for an available node that meets the needs. 4. If node found, the job is assigned. Else it will search for a free node upward, from level rj+1, rj+2, rj+3 to rL, and assign job to the first free node found.
- 12. 5. If node found, the job is assigned. Else it finds downward from level rj-1, rj-2, rj-3 to r1, and assign job to the first free node found. 6. If the job still not assigned, it will wait in queue. 7. When job starts running, it schedules initial check pointing at a short time period. 8. During first check pointing interval, executing job image is sent to resource’s storage machine. Storage saves job image and informs resource upon completion. 9. Algorithm schedules next checkpointing based on remaining execution time and machine failure interval information kept at GIS.
- 13. 12. During next intervals, instead of whole job image, resource only sends updated attributes/properties of the job to storage. Storage updates job. 13.job is completed. It notifies storage machine to remove the saved checkpoint to free storage space. 14. On machine failure or when the resource is out of order, storage machine notifies job’s corresponding user whether to load saved checkpoint. 15. When user receives the load checkpoint request, it will use SMF to choose a fitting free machine in same or from other resources, to resume execution of the job.
- 14. Job Allocati on (Testing Data) Gridlets creation: Job Allocation (Testing Data): Check point Controlling: Applying Rollback Mechanism: check point triggers a roll back mechanism to identify cause & origin of failure. Finding Fault Tolerant Node in the Network Time Stamp Event Tracki ng Mem ory Reso urce Allo catio n Fault Toler ant File Trans fer Che ck Poin tnt Con trol SMF (Select Most Fitting Resourc e) Find Fault Tolera nt & Rollback OU TPU T (Tas k Exe cuti on)
- 15. Thank you!!! Thank you for Giving us this GOLDEN opportunity!!!
- 16. Queries???