A Service Sustainable Live Migration Strategy for Multiple Virtual Machines in Cloud Data Centers

MiGrror: Mitigating Downtime in Mobile Edge Computing, An Extension to Live Migration

“…A subset of these papers provides models and compares various parameters of live migration methods [46][47][48][49], while others employ estimation and optimization techniques to reduce migration costs, such as downtime and migration time [33,50,51]. Despite extensive research on modelling the migration of a single VM/container, few authors focus on modelling multiple-VM/multiple-container migration [18,19,[37][38][39][40][41][42][43][44][45]. Some of these studies focus on the number of VMs/containers and provisioned bandwidth in addition to the stated input values.…”

“…Some of these studies focus on the number of VMs/containers and provisioned bandwidth in addition to the stated input values. Most of these studies focus on modelling multiple-VM migration [18,19,37,38,40,[42][43][44][45] in order to optimize the migration performance of multiple VMs, while authors in [39,41] focus on modelling multiple-container migration. The authors in [18,19,33,[37][38][39][40][41][42][43][44][45][46][47][48][49][50][51] employ only average parameter values, whereas we use non-average values for more precise results.…”

“…In more detail, Sun et al [19] use an M/M/C/C queuing migration model to optimize migration and reduce downtime and migration time for multiple VMs. Satpathy et al [37] compare migration model performance for multiple VMs, including comparisons based on VM size, memory dirtying rate, and available bandwidth. Using a platform based on a softwaredefined network (SDN), He et al [38] evaluate the performance of multiple VM migration models.…”

“…Studies [18,19,33,[37][38][39][40][41][42][43][44][45][46][47][48][49][50][51] use migration modelling to comprehend the future behaviour of a system. Although most research employs average parameter values [18,19,33,[37][38][39][40][41][42][43][44][45][46][47][48][49][50][51] and assumes the input parameter values remain unchanging, our study demonstrates that the results vary since the input parameters can constantly change during the migration. Our results show it is essential to learn if these parameters are higher or lower at the beginning, middle, and end of the migration, considering migration time and downtime as output parameters.…”

“…As a result, new migration models in edge computing are required to achieve more accurate results for multi-container mission-critical 5G/6G mobile IoT applications. This paper presents mathematical models for the pre-copy and MiGrror migration methods, offering a new perspective on modelling by utilizing both average and non-average values during MiGrror migration, while typically, only average values are considered in the literature [18,19,33,[37][38][39][40][41][42][43][44][45][46][47][48][49][50][51]. This research takes into account non-average values of transfer rate, memory dirtying rate and VM/container size for the MiGrror technique in addition to average values.…”