Upload
sarmad-makhdoom
View
1.744
Download
1
Tags:
Embed Size (px)
DESCRIPTION
Survey on Virtual Machine migration with in the intra-LAN network as well as the long distance migration.
Citation preview
Sarmad Makhdoom2012-03-0019
Kamran Khalil2012-03-0012
Hafeez ur Rehman 2012-03-0031
Challenges in Cloud Computing – VM Migration
Introduction
• Cloud computing is an umbrella term used to refer to Internet based development and services
• A number of characteristics define cloud data, applications services and infrastructure:
▫ Remotely hosted: Services or data are hosted on remote infrastructure. ▫ Ubiquitous: Services or data are available anywhere.▫ Commodified: The result is a utility computing model similar to traditional
that of traditional utilities, like gas and electricity - you pay for what you would want!
Cloud Computing
•A virtual machine provides interface identical to underlying bare hardware▫ i.e. all devices, interrupts, memory, page tables etc.
•Applications of Virtual Machines•Virtualization Software▫VMWare▫ZAP▫Xen▫QEMU
Virtual Machines
▫Benefits of Virtual Machines Virtualization help making efficient use of hardware resources Facilitates a greater degree of abstraction Easily move from one piece of hardware to another Replicate them at will Create more scalable and flexible infrastructure Snapshots
▫Cloud computing has taken that degree of efficiency and agility realized from virtualization Pooled resources Geographic diversity Universal connectivity
Virtual Machines in Cloud
Research Problems
•Automated services provisioning•Virtual machine migration•Server consolidation•Energy management•Traffic management and analysis•Data security•Software frameworks•Storage technology and data management•Novel cloud architecture
Research Problems
Motivation
•Consider a data center consisting of “n” physical machines (PM) hosting “m” VMs implementing one customer application each
•Resources(CPU, Network, Memory, I/O) are allocated to each VM to handle the workload and operate at certain performance level (SLA)
•Each VM sees workload fluctuation from time to time => resource requirement changes
Motivation
Network Bandwidth = 10 GbpsMemory = 16 GBCPU = 8 cores
VM1N = 5GbpsM = 8GBC = 4 cores
VM2N = 5GbpsM = 8GBC = 4 cores
Resource Allocation
PM CapacityVM1 VM2
# of user visit increases
cricinfo mail server
Virtualization LayerHardware
•An increase in workload can be handled by allocating more resources to it, if idle resources are available
•Main Issues:▫What if PM does not have (enough or no) idle resources to satisfy VM's
requirement? Performance of the application degrades SLA violation occurs
•Key Ideas▫Replication VMs▫Migrating VMs
Motivation
Virtual Machine MigrationIt is impossible or impractical to bring the data (or devices) close to the
computation engines.
•Why we need migration?•When we need to migrate?•How migration is done?•Issues in long distance migration (across data centers)
Virtual Machine Migration
•Hotspots can cause SLA violations▫Burden on some Virtual or Physical Machines are called hotspots
•Hotspot Detection (Sandpiper)▫Black-box Monitoring
CPU (/proc) Network (/proc/net/dev) Memory (swap)
▫Gray-box Monitoring Gather OS level statistics and application logs
•A hotspot is flagged only if thresholds or SLAs are exceeded for a sustained time
When we need to migrate? [NSDI’ 07]
Wood T et al (2007) Black-box and gray-box strategies for virtual machine migration. In: Proc of NSDI
14
Gathering resource usage
statistics on that server
Gathers processor, network and memory swap statisticsfor each VM Implements a
daemon to gatherOS-level statistics and application logs
Construct resource usage profiles foreach virtual server(Predict PM workload)
Monitors usage profiles to detect hotspots.Hotspot: any resource exceeds a threshold(or SLA violation) for a sustain period
Determine:What virtual servers should migrateWhere to move themHow much of a resource to allocate the virtual servers after migration
Wood T et al (2007) Black-box and gray-box strategies for virtual machine migration. In: Proc of NSDI
•SLA violation detection▫Mapping low-level resource metrics to high-level SLAs▫Crude data maps to user requirements such as
CPU speed maps to Response Time Occupied memory size maps to number of concurrent clients
▫Predictive Strategy for detection of possible SLA violations▫Detection interval
Short measurement intervals may degrade performance Long measurement intervals may cause ignorance of heavy SLA violations
When we need to migrate? [FGCS’ 12]
Towards autonomic detection of SLA violations in Cloud infrastructures, Future Generation Computer Systems, 2012
•Memory Migration▫Pre-copy
Push phase Stop-and-copy phase Pull Phase
▫Pure demand-migration•File System Migration▫In case of distributed file system, there is no need to copy▫Alternatively, copy only changed local files to the destination using Virtual
Machine Manager’s API.
How migration is done?
S. Venkatesha, S. Sadhu, S. Kintali, and S. Barbara, "Survey of virtual machine migration techniques" - Memory, 2009
•Network Migration▫If both source and destination are on same LAN switch
an unsolicited ARP reply from the migrating host is provided▫Alternatively, on a switched network
the migrating OS can keep its original Ethernet MAC address, relying on the network switch to detect its move to a new port
•Device Migration▫Three type of device support
Emulation Virtualization Non-migratable
How migration is done?
S. Venkatesha, S. Sadhu, S. Kintali, and S. Barbara, "Survey of virtual machine migration techniques" - Memory, 2009
•Testbed specification▫6 Servers (1 head node, 5 VM hosts)▫Intel Xeon (2.33 GHz Quad-core with 2x6MB L2 Cache)▫4GB memory and 7200rpm hard drive▫64-bit Ubuntu Linux 8.04 Server Edition▫Apache 2.2.8 and MySQL 5.2.4-2
•Workload▫Olio as a Web 2.0 application (http://incubator.apache.org/olio/)
▫Faban Load generator (http://faban.sunsource.net)
•Experiments▫10 minute and 20 minute benchmark runs with 600 concurrent users
Migration: A Performance Evaluation
Cost of virtual machine live migration in clouds: A performance evaluation, International Conference on Cloud Computing, 2009
Migration: A Performance Evaluation
Cost of virtual machine live migration in clouds: A performance evaluation, International Conference on Cloud Computing, 2009
•Need for VM mobility across data centers▫Data center maintenance without downtime▫Disaster avoidance▫Data center migration/expansion▫Workload balancing across multiple sites
• Issues▫RTT is 1,000 times greater than sub-networks, ‘word wide wait’▫Trust to a remote execution environment▫Interoperability at the level of Web Services, Java etc▫Migration across multiple domains are vulnerable to security exploits
Migration across data centers
• Introduced an intermediate traffic controller to facilitate migration which consists of:▫VM Traffic Controller
Provisioning of network resources and the re-provisioning of the IP tunnel to ensure seamless layer
▫AAA (Authentication, Authorization and Accounting) Pre-allocation of extra VM-resources required for migration
▫DRAC (Dynamic Resource Allocation Controller) Exposes a service-oriented API for coupling with applications
▫Preservation of TCP and higher-level sessions Dynamically configured IP tunnels allow client connectivity
Related Work: Migration over MAN/WAN [FGCS’06]
“Seamless Live Migration of Virtual Machines over the MAN/WAN”, Elsevier Future Generation Computer Systems 2006
•HP Cluster Extension (CLX) provides flawless mirroring capabilities for disaster recovery
•VM data is already replicated in single data center cluster▫Modifications to spread it across multi-site:
Multi-site Disaster Recovery solution is implemented CLX enables Hyper-V Live Migration across sites
Related Work: HP Cluster Extension and Microsoft Hyper-V™
(VMware, Hyper-V, HP-VM and AMD-V)Our limitation is non-availability of peer referenced or published
material of these architectures to discuss their techniques in detail
•Security▫Layer 2 Link Encryption (IEEE 802.1AE) may be used to help ensure privacy
and confidentially ▫Token based security - Authorization message sequence to thwart
resource theft•1,000 times rise in RTT doesn’t matter much because▫All states will be transferred before the switch▫Downtime will be based on the time required in redirection
Which is only 5-10 times than the intra-LAN setup
Proposed Solution
Conclusion
•Virtual machine migration provide significant benefits in cloud computing
•State of the art work is already done on resource monitoring, live VM migration over the LAN and MAN/WAN
•Downtime is 60ms to 3s which they assume is negligible•Small transient spikes does not trigger needless migrations, if threshold
or SLAs are exceeded for sustained time•A slowdown is expected due to cache warm-up at the destination after
migration•Hybrid approach has been suggested to harness the benefits of all
technqiues
Conclusion
Questions and Answers