Efficient Online Virtual Machines Migration for Alert-Based Disaster Resilience

Abstract

Several recent weather-based disasters had very negative impacts on cloud networks, causing Data Center (DC) shutdown, consequent data-loss and intolerable downtime of cloud services. This has put the reactive disaster-resilient design of cloud networks on top the agenda of several cloud DC operators. DC operators are investigating approaches to avoid downtime of cloud services in case a DC is affected by a disaster. Thanks to virtualization most cloud services run on Virtual Machines (VMs) hosted by DCs, so it is possible to keep these services alive if the VMs are evacuated (namely, migrated) before the disaster from a DC affected by the disaster to a DC in a safe location, in an online technique. This technique is known as online "VM migration", which results without or with a minimal service downtime. In this paper, we present an Integer Linear Programming (ILP) model for efficient online VMs migration in case of an alerted disaster (e.g., most weather-based disasters, as hurricanes) such as to avoid service downtime. The ILP performs scheduling and assigns route and bandwidth to the migration of VMs towards a safe DC within an alert time, with the objective of maximizing the number of VMs migrated and minimizing service downtime, network resource occupation and migration duration. We present a comparative analysis of offline and online migration strategies such as to quantify the trade-off between downtime, network resource utilization and migration duration. Moreover, we investigate the impact of the memory dirtying rate on the online migration process, i.e., the number of VMs evacuated and network resource occupation.

Downtime Data Center: Memahami Penyebab, Dampak, dan Solusi Efektif

“…Tujuan dari penggunaan Data Center tergantung berdasarkan organisasi atau industri seperti apa yang akan menggunakannya, seperti institusi pendidikan, organisasi berorientasi laba, atau organisasi pemerintahan. Menurut Ayoub et al, (2019) serta Hao et al, (2019) tujuan dari penggunaan Data Center yang utama adalah memberikan dukungan operasional kritis dan menyediakan layanan digital dengan menyimpan dan mengelola infrastruktur IT. Selain itu, Data Center juga digunakan untuk memaksimalkan penggunaan sumber daya yang tersedia untuk meningkatkan ketangguhan suatu proses bisnis sebuah organisasi (Bari et al, 2014;Shi et al, 2022;Varasteh & Goudarzi, 2015).…”

“…Setelah terjadi downtime, langkah-langkah tanggap dan solutif diterapkan untuk mengidentifikasi sumber permasalahan, mengisolasi kerusakan, dan memulihkan fungsi sistem dengan secepat mungkin. Dengan adanya strategi downtime overcome and recovery yang terintegrasi, perusahaan dapat meminimalkan dampak negatif dari kegagalan dan memastikan ketersediaan layanan Data Center yang optimal Berdasarkan artikel Ayoub et al, (2019) dan Le (2020), tujuan dari migrasi VM adalah meminimalkan downtime, memungkinkan administrator sistem untuk melakukan pemeliharaan perangkat keras atau perangkat lunak pada host tanpa harus mematikan mesin virtual. Ini dapat meningkatkan keterasediaan sistem.…”

A Framework for Spatial and Temporal Evaluation of Network Disaster Recovery

“…hurricanes) may also benefit from the forecasts and trajectory projections. Alert-based migration of virtual machines to safer datacenter is performed in [12]. The work in [13] combines the awareness of both disaster impact and remediation in terms of content evacuation and incorporates it into the datacenter placement.…”

Online Virtual Machine Evacuation for Disaster Resilience in Inter-Data Center Networks

“…We refer to this problem as "Alert-based Online VM Migration for Disaster-Resilience" (Alert-VMmig). In our previous work [9], we proposed an Integer Linear Programming (ILP) model to solve this problem. In this paper, we enhance the ILP model and propose various heuristic approaches to investigate large instances of the problem.…”