The following jobs are loaded into memory using fixed partition following a certain memory allocation method (best-fit, first-fit and worst-fit).
List of Jobs | Size | Turnaround |
Job 1 | 100k | 3 |
Job 2 | 10k | 1 |
Job 3 | 35k | 2 |
Job 4 | 15k | 1 |
Job 5 | 23k | 2 |
Job 6 | 6k | 1 |
Job 7 | 25k | 1 |
Job 8 | 55k | 2 |
Job 9 | 88k | 3 |
Job 10 | 100k | 3 |
Memory Block Size
Block 1 50k
Block 2 200k
Block 3 70k
Block 4 115k
Block 5 15k
B E S T - F I T
Best-fit memory allocation makes the best use of memory space but slower in making allocation. In the illustration below, on the first processing cycle, jobs 1 to 5 are submitted and be processed first. After the first cycle, job 2 and 4 located on block 5 and block 3 respectively and both having one turnaround are replace by job 6 and 7 while job 1, job 3 and job 5 remain on their designated block. In the third cycle, job 1 remain on block 4, while job 8 and job 9 replace job 7 and job 5 respectively (both having 2 turnaround). On the next cycle, job 9 and job 8 remain on their block while job 10 replace job 1 (having 3 turnaround). On the fifth cycle only job 9 and 10 are the remaining jobs to be process and there are 3 free memory blocks for the incoming jobs. But since there are only 10 jobs, so it will remain free. On the sixth cycle, job 10 is the only remaining job to be process and finally on the seventh cycle, all jobs are successfully process and executed and all the memory blocks are now free.
F I R S T - F I T
First-fit memory allocation is faster in making allocation but leads to memory waste. The illustration below shows that on the first cycle, job 1 to job 4 are submitted first while job 6 occupied block 5 because the remaining memory space is enough to its required memory size to be process. While job 5 is in waiting queue because the memory size in block 5 is not enough for the job 5 to be process. Then on the next cycle, job 5 replace job 2 on block 1 and job 7 replace job 4 on block 4 after both job 2 and job 4 finish their process. Job 8 is in waiting queue because the remaining block is not enough to accommodate the memory size of job 8. On the third cycle, job 8 replace job 3 and job 9 occupies block 4 after processing job 7. While Job 1 and job 5 remain on its designated block. After the third cycle block 1 and block 5 are free to serve the incoming jobs but since there are 10 jobs so it will remain free. And job 10 occupies block 2 after job 1 finish its turns. On the other hand, job 8 and job 9 remain on their block. Then on the fifth cycle, only job 9 and job 10 are to be process while there are 3 memory blocks free. In the sixth cycle, job 10 is the only remaining job to be process and lastly in the seventh cycle, all jobs are successfully process and executed and all the memory blocks are now free.
W O R S T - F I T
Worst-fit memory allocation is opposite to best-fit. It allocates free available block to the new job and it is not the best choice for an actual system. In the illustration, on the first cycle, job 5 is in waiting queue while job 1 to job 4 and job 6 are the jobs to be first process. After then, job 5 occupies the free block replacing job 2. Block 5 is now free to accommodate the next job which is job 8 but since the size in block 5 is not enough for job 8, so job 8 is in waiting queue. Then on the next cycle, block 3 accommodate job 8 while job 1 and job 5 remain on their memory block. In this cycle, there are 2 memory blocks are free. In the fourth cycle, only job 8 on block 3 remains while job 1 and job 5 are respectively replace by job 9 and job 10. Just the same in the previous cycle, there are still two free memory blocks. At fifth cycle, job 8 finish its job while the job 9 and job 10 are still on block 2 and block 4 respectively and there is additional memory block free. The same scenario happen on the sixth cycle. Lastly, on the seventh cycle, both job 9 and job 10 finish its process and in this cycle, all jobs are successfully process and executed. And all the memory blocks are now free.