advertisement

Resource Allocation using Virtual Machine Migration: A Survey

50 %
50 %
advertisement
Information about Resource Allocation using Virtual Machine Migration: A Survey
Education

Published on February 17, 2014

Author: idescitation

Source: slideshare.net

Description

As virtualization is proving to be dominant in
enterprise and organizational networks there is a need for
operators and administrators to pay more attention to live
migration of virtual machines (VMs) with the main objective
of workload balancing, monitoring, fault management, low-
level system maintenance and good performance with minimal
service downtimes. It is also a crucial aspect of cloud computing
that offers strategies to implement the dynamic allocation of
resources. Virtualization also enables virtual machine
migration to eliminate hotspots in data centers .However the
security associated with VMs live migration has not received
thorough analysis. Further, the negative impact on service
levels of running applications is likely to occur during the
live VM migration hence a better understanding of its
implications on the system performance is highly required.
In this survey we explore the security issues involved in live
migration of VMs and demonstrate the importance of security
during the migration process. A model which demonstrates
the cost incurred in reconfiguring a cloud-based environment
in response to the workload variations is studied. It is also
proved that migration cost is acceptable but should not be
neglected, particularly in systems where service availability
and response times are imposed by stringent Service Level
Agreements (SLAs). A system that provides automation of
monitoring and detection of hotspots and determination of
the new mapping of physical to virtual resources and finally
initiates the required migrations based on its observations is
also studied. These are experimented using Xen Virtual
Machine Manager. Migration based resource Managers for
virtualized environments are presented by comparing and
discussing several types of underlying algorithms from
algorithmistic issues point of view.
advertisement

Poster Paper Proc. of Int. Conf. on Advances in Information Technology and Mobile Communication 2013 Resource Allocation using Virtual Machine Migration: A Survey Ts‘epo Mofolo1, R. Suchithra2, N. Rajkumar3 1,2 MS (IT) Department Jain University, Bangalore, India {mofolotc, suchithra.suriya}@gmail.com 3 Department of Software Engineering Ramakrishna College of Engineering, Coimbatore, India Nrk29@rediffmail.com Abstract- As virtualization is proving to be dominant in enterprise and organizational networks there is a need for operators and administrators to pay more attention to live migration of virtual machines (VMs) with the main objective of workload balancing, monitoring, fault management, lowlevel system maintenance and good performance with minimal service downtimes. It is also a crucial aspect of cloud computing that offers strategies to implement the dynamic allocation of resources. Virtualization also enables virtual machine migration to eliminate hotspots in data centers .However the security associated with VMs live migration has not received thorough analysis. Further, the negative impact on service levels of running applications is likely to occur during the live VM migration hence a better understanding of its implications on the system performance is highly required. In this survey we explore the security issues involved in live migration of VMs and demonstrate the importance of security during the migration process. A model which demonstrates the cost incurred in reconfiguring a cloud-based environment in response to the workload variations is studied. It is also proved that migration cost is acceptable but should not be neglected, particularly in systems where service availability and response times are imposed by stringent Service Level Agreements (SLAs). A system that provides automation of monitoring and detection of hotspots and determination of the new mapping of physical to virtual resources and finally initiates the required migrations based on its observations is also studied. These are experimented using Xen Virtual Machine Manager. Migration based resource Managers for virtualized environments are presented by comparing and discussing several types of underlying algorithms from algorithmistic issues point of view. Keywords: Virtualization, Migration, Virtual Machines, allocation, resources. I. INTRODUCTION Live migration of virtual machines (VMs), the process of mirroring a VM from one Virtual Machine Manager (VMM) to another without stopping the execution of a guest operating system, often between different physical hosts has resulted in new opportunities in traditional computing as well Cloud Computing. Live migration is of great importance in achieving factors such as high-availability of services, transparent mobility, consolidated mobility and workload balancing. [1, 6, 7] Virtualized infrastructures have proved to be a key component to drive the emerging Cloud Computing paradigm. Migration of VMs aims at improving the manageability, 99 © 2013 ACEEE DOI: 03.LSCS.2013.2.556 performance and fault tolerance of systems. Reasons that justify VM migration in a production system include: the need for a balanced system workload, which can be achieved via migrating VMs out of overloaded or overheated servers and the need to power off servers for maintenance after migrating their workload other servers. Hypervisors such as Xen and VMWare allow the migration of a VM while it still continues to provide services to various applications [4, 8, 9, 12]. The greatest advantage of live migration is the possibility to migrate a VM with near-zero downtime, a crucial feature when applications are being served. [4]Clouds have a remarkable advantage over traditional data centers in providing elasticity as well as attaining high resource utilization. A customer has the ability and flexibility of increasing and decreasing the amount of resources it requires for itself. For a cloud provider, elasticity is the ability to transparently exchange resources from one customer to another in response to variations in demand thus enabling the cloud to operate at high resource utilization. [2]The following benefits are provided by live migration in multiple VM-based environments: [10, 11] Load balancing, online maintenance and proactive fault tolerance and power management. Live migration introduces some significant security challenges. A VMM that facilitates a vulnerable implementation may lead to the exposure of both the guest and host operating system attacks and hence result in absolute system integrity compromise. Reconfigurations (dynamic allocation) in a cloud may result in performance issues from hosted applications. It consumes resources and may result in resource contention for applications. Hence it is of utmost importance to have a thorough understanding of the following aspects: (i) The frequency of reconfigurations in typical cloud environment (ii) The impact of certain reconfigurations on the hosted applications [2] The dynamic resource allocation requirements of a workload can be satisfied by changing the capacity of a virtual machine at runtime. The pre-copy algorithm [3, 12] addresses the issue of downtime by reducing it to the magnitude of milliseconds. There are other issues which remain unresolved though: When the rate at which pages are dirtied is faster than that of pre-copy process, all pre-copy work will be ineffi

Poster Paper Proc. of Int. Conf. on Advances in Information Technology and Mobile Communication 2013 cient. This would mean memory-intensive applications would not benefit from pre-copy algorithm and the downtime may increase to several seconds. This drawback makes the algorithm applicable in high speed LANs. A novel live migration is proposed. The main objective is to minimize the migration downtime and network bandwidth consumption and this is implemented based on full-system trace and replay systemRevirt. [13] Checkpointing/recovery and trace/replay technology is adopted to provide fast, transparent VM migration in both LAN and WAN environments. A trace daemon is deployed and continuously keeps logs of the non-deterministic events of the VM while sacrificing very little performance. The execution trace file logged at the original host is iteratively mirrored to the target host and used to synchronize the migrated VM’s execution state. [10]Other factors to be considered include the total migration during which state on both the source and target hosts is synchronized and hence may affect system reliability. Further, it has to be ensured that migration does not interfere with active services through resource contention for instance CPU and network bandwidth with the migrating OS. This can be achieved through pre-copy algorithm in which pages of memory are mirrored iteratively from the source host to the target host, all without stopping the execution of the virtual machine being migrated. The final stage pauses the VM, mirrors only remaining pages to the target host and restarts execution there as explained Clark et al. General live migration process based management system is presented in this paper. [5]Impact of reconfiguration in a cloud setting has been studied and a model presented to characterize it. B. The Migration Planning Phase This is the most sensitive part of the processing. The VMM has to produce a convenient migration plan. A typical migration plan consists of source physical host(s), VM(s) to be migrated and the physical host(s). C. The Migration Execution Phase It is the responsibility of the VMM to put the migration plan into action. The specific details of this phase depend on the adopted virtualization technology. III. DESIGN CONSIDERATIONS ON LIVE VM MIGRATION Migrating VMs in cluster server environments leads to consideration of the physical resources used in such architectures, in particular, the memory, network and disk. A. Migrating Memory “Reference [3]”, Clark et al, Mirroring VM’s memory from one physical host to another can be performed in several ways. When a VM is executing a live service, this mirroring process has to be carried out in way that strikes a balance of the requirements of minimizing the downtime and total migration time. Downtime will be directly observable to the end users of the VM as service disruption. Memory transfer is achieved in three phases: Push phase: The source VM continues its execution while certain memory pages are pushed across the network to the target physical host. Memory pages modified (dirtied) during this process must be resent in order to ensure consistency. Stop and copy phase: The source VM is halted and pages of memory are copied to the target host and the execution of the new VM is started on the physical target host. Pull phase: The new VM executes, and if it accesses a memory page that has not yet been mirrored this page is pulled across the network from the source VM. The stopand-copy mechanism means downtime and total migration time are directly proportional to the amount of physical memory allocated to a VM. This is highly likely to result in detrimental outage if the VM is executing a live service. The pre-copy mechanism balances these issues by combining a parameterized iterative push phase and a very short stopand-copy phase. Iteratively scanning and transferring a VM’s memory image between two physical hosts in a cluster could subsequently consume the entire network bandwidth available between them and deprive the executing services of resources. The service degradation is likely to occur to a certain extent during any live migration process. This issue is addressed by employing strategies that ensure that the live migration process does not disrupt the active traffic or any processing. II. GENERAL LIVE MIGRATION PROCESS USING RESOURCE MANAGER This section presents the major phases for a general VM live migration process which use a resource manager. The main focus is on operating system virtualization. On every physical host there is a resident Virtual Machine Manager (VMM), also known as a hypervisor which enables many virtual machines to have the physical system resources. The process of migrating VMs is carried out by the VMM. In a shared hosting model an individual application is represented by one VM, however there are situations where it can be represented more than one VM. [15, 16] The resource manager execution occurs in iterations. In every iteration, three major phases occur: A. The Pre-allocation Phase The main responsibility of the VMM is to gather usage data from the executing nodes within a predefined measurement interval by using a particular monitoring tool. The features of this tool depend mainly on the adopted virtualization technology and the needed data to be collected. Through these gathered data, the VMM can keep a general view about the performance in the executing nodes. Reallocation is initiated if there are violations of the predefined triggering conditions. © 2013 ACEEE DOI: 03.LSCS.2013.2.556 B. Local Resources Considering network resources, it is desirable that the migrated VM should have all network connections open without any dependencies on forwarding techniques on the original host, which maybe shutdown after the migration process. A migrating VM will include the protocol state and 100

Poster Paper Proc. of Int. Conf. on Advances in Information Technology and Mobile Communication 2013 migration based on the resource manager are thoroughly analyzed in this section. [5] will still retain its IP address when migrated. C. Design Overview In Figure 1 the migration process is viewed as a transactional interaction between two physical hosts under consideration as presented by to Clark et al [3]. A. Ordering algorithms These algorithms should allow the resource manager to answer questions such as: from where to migrate? Which VM to migrate? Where to migrate? Ordering algorithms such as Dynamic Management Algorithm (DMA) have some limitations. Using DMA makes it impossible to judge the performance of the resource manager under network intensive applications. The Fuzzy Decision making model based on TOPSIS techniques suffers from the problem of time consuming calculations. In [14] management algorithms for cloud computing paradigm are proposed. A load-trend based sorting algorithm is used to select candidate VMs for migrations and the target hosts. It distributes the migrated VMs by assigning each receiving host only one VM. This leads to avoidance of load balance shifting, which is an unwanted implication that results in frequent fluctuations that negatively impact the system performance and stability. B. Constraint Programming It has the limitations of only considering only viable processing nodes and CPU-memory resources. The explored resource manager (Entropy) has the capability to combine and automate application provisioning problem. However this also has a problem of being limited to CPU-memory dimensions. C. Genetic Algorithms (GAs) Gas based model enables for multiple-SLA representation for each VM. The main challenge of Gas results from the infeasible solutions that can appear as a byproduct of the evolution process.GA based approach is flexible and faster for VM packing problems and represents a promising mechanism. It can greatly fasten constraint programming expensive processing. Figure 1: VM Migration Timeline Step 0: Pre-Migration The process begins with VM executing on source physical host A. A target host maybe preselected and has to have enough resources by the migrated VM. Step 1: Reservation A request to migrate VM from host A to host B is issued. Resources available on B are verified and a container for VM’s size is reserved. Step 2: Iterative Pre-Copy All pages are transferred from A to B. During round n, pages dirtied during round n-1 are transferred. Step 3: Stop-and-Copy VM in A is halted for a while and its network traffic is redirected to B. The VM copy at A is still considered to be primary and can be resumed in case of migration failure. Step 4: Commitment Host B acknowledges to A that it has successfully received a consistent copy of the VM. Host A acknowledges B’s message as commitment of the migration process. Host A can get rid of the original VM B takes over as the primary physical host. Step 5: Activation The VM that resides on B is activated. Post-migration code is executed to advertise the migrated IP address. This technique ensures that at least one physical host has a consistent VM copy during the entire migration process according to Clark et al [3]. V. CONCLUSION Virtualization is a significant technology that can be deployed in data centers and cloud platforms to provide powerful resource allocation strategies. In this paper we surveyed the benefits offered by VM machine migration in resource allocation. Challenges associated with the VM migration have also been explored including the security threats inherent in live VM migration. Strategies and techniques to overcome these issues have also been outlined. The checkpointing/ recovery and trace/replay technology provides fast and transparent VM for both LAN and VM environments and this gives it a good advantage over other migration strategies.[10] The Xen VMM provides rapid transfer of interactive workloads within clusters and data centers. The total downtime is also reduced below thresholds. [1] Access control policies should be provide administrators with the privileges to manage the migration process. Techniques that ensure very little utilization of resources by the migration process should be developed as it leads to high downtime and subsequently service degradation and this is not desirable in cloud IV. ANALYSIS OF RESOURCE MANAGEMENT ALGORITHMS Algorithms that are used to implement the surveyed live © 2013 ACEEE DOI: 03.LSCS.2013.2.556 101

Poster Paper Proc. of Int. Conf. on Advances in Information Technology and Mobile Communication 2013 platforms. Dynamic consolidation methods which aim at minimizing number of migrations as much as possible should be employed in cloud computing. This should enable achievement of high resource utilization. [8] P. Barham, B. Dragovic, K. Fraser, S. Hand, T. Harris, A. Ho, R. Neugebaher, I. Pratt, A. Warfield, “Xen and the art of virtualization”, In: SOSP ’03: Proceedings of the 19th ACM Symposium on Operating System Principles, New York, NY, USA, ACM (2003) pages 164-177, in press [9] VMWare Inc: http://www.VMWare.com [10] H. Jin, H. Liu, X. Liao, L. Hu, P Li, “Live Migration of Virtual Machine Based on Full-System Trace and Replay” 01/2009; In Proceeding of: Proceedings of the 18th ACM International Sysmposium on High Performance Distributed Computing, HPDC 2009, Garching, Germany, June 11-13,2009, in press [11] R. Nathuji, K. Schwan. “Virtual Power: Coordinated Power Management in Virtualized Enterprise Systems”, Proceedings of the 22nd ACM Symposium on Operating Systems Principles (SOSP ‘07), October 14-17, 2007, Skamania Lodge Stevenson, WA, in press [12] M. Nelson, B.H. Lim, and G. Hutchins, “Fast Transparent Migration for Virtual Machines”, Proceedings of USENIX Annual Technical Conference (USENIX), April 10-15, 2005, Marriot Anaheim, Anaheim, CA, USA, pp. 391-394, in press [13] G.W. Dunlap, S.T. King, S. Cinar, M.A. Basrai, P.M. Chen, “Revirt: Enabling Intrusion Analysis through Virtual-Machine Logging and Replay”, Proceedings of the 2002 Sysmposium on Operating Systems Design and Implementation (OSDI), pp. 211-224, 2002, in press [14] M. Andveolini, S. Casolari, M. Colajanni, and M. Messori, “Dynamic Load Management of Virtual Machines in Cloud Architectures”, First International Conference on Cloud Computing (ICSI CLOUDCOMP 2009), Munich, Germany, October 19-21, 2009, in press [15] H.N Van, F.D. Tran, “Autonomic Virtual Resource Management for Service Hosting Platforms”, In Proc. Of the 2009 ICSE Workshop on Software Engineering Challenges of Cloud Computing, pp. 1-8, 2009, in press [16] P. Ruth, J. Rhee, D. Xu, R. Kennell, S. Goasghen, “Autonomic Live Adaptation of Virtual Computational Environments in a Multi-Domain Infrastructure”,In Proc. IEEE ICAC’06, pp. 5-14, June, 2006, in press REFERENCES [1]J. Oberheide, E. Cooke and F. Jahanian. “Empirical Exploitation of Live Virtual Machine Migration” Proceedings of BlackHat DC convention, 2008, in press [2] A. Verna, G. Kumar, R. Koller, A. Sen.”CosMig: Modelling the Impact of Reconfiguration in a Cloud” IEEE/ACM MASCOTS 2011, in press [3] C. Clark, K. Fraser, S. Hand, J.G. Hansen, E. Jul, C. Limpach, I. Pratt, A. Warfield, “Live Migration of Virtual Machines”, NSDI ’05:2nd Symposium on Networked Systems Design and Implementation, 2005,in press [4] W. Voorsluys, J. Brobery, S. Venugopal, and R. Buyya, “Cost of Virtual Machine Migration in Clouds: A Performance Evaluation.” Proceedings of the 1st International Conference on Cloud Computing (cloudcom 2009), pages 254-265, in press [5]O. Abdul-Rahman, M. Munetomo and K. Akama.”Live Migration-based Resource Managers for Virtualized Environments: A Survey.” CLOUD COMPUTING 2010: The First International Conference on Cloud Computing, GRIDs and Virtualization, in press [6] A. Ganguly, A. Agrawal, P. Oskar Boykin, R. Figuiredo, “WOW: Self-organizing Wide Area overlay Networks of Virtual Workstations.” Proceedings of the 15th IEEE International Symposium on the High Performance Distributed Computing (HPDC), pages 30-41 2006, in press [7] P. Ruth, J. Rhee, D. Xu and S. Goasquen. “Automatic Live adaptation of Virtual Computational Enviroments in a MultiDomain Infrastructure”, IEEE International Conference on Automatic Computing (ICAC’06), 2006, in press © 2013 ACEEE DOI: 03.LSCS.2013.2.556 102

Add a comment

Related presentations

Related pages

Resource Allocation using Virtual Machine Migration: A Survey

As virtualization is proving to be dominant in enterprise and organizational networks there is a need for operators and administrators to pay more ...
Read more

Resource Allocation using Virtual Machine Migration: A Survey

Poster Paper Proc. of Int. Conf. on Advances in Information Technology and Mobile Communication 2013 Resource Allocation using Virtual Machine Migration:
Read more

th Vibrant Resource Allocation Algorithms using Virtual ...

Vibrant Resource Allocation Algorithms using Virtual Machine in ... of virtual machine migration. ... Survey on Resource Allocation Strategies ...
Read more

Poster Paper Proc. of Int. Conf. on Advances in ...

Resource Allocation using Virtual Machine ... In this survey we explore ... These are experimented using Xen Virtual Machine Manager. Migration based ...
Read more

Threshold based Dynamic Resource Allocation using Virtual ...

Threshold based Dynamic Resource Allocation using Virtual Machine Migration ... made a survey on the ... Dynamic Resource Allocation Using Virtual Machine ...
Read more

A Survey on Energetic Service Allocation Using Virtual ...

algorithm that helps our resource allocation system ... A Survey on Energetic Service Allocation Using Virtual ... virtual machine migration,” in Proc ...
Read more

Resource Management Using Virtual Machine Migrations ...

“Dynamic resource management using virtual machine ... “A survey on resource allocation in ... using fast virtual machine migration.”IBM ...
Read more

A Survey on Resource Allocation Strategies in Cloud Computing

A Survey on Resource Allocation Strategies in ... contention and increases resource utilization by using different ... C. Virtual Machine ...
Read more

A Dynamic Resource Allocation Model for Virtual Machine ...

A Dynamic Resource Allocation Model for Virtual Machine Management on Cloud. ... Virtualization VM Migration Dynamic Resource Allocation Cloud Computing.
Read more