The case for vm based cloudlets in mobile computing
2 likes1,500 views
This paper proposes using cloudlets, which are localized data centers, to overcome limitations of mobile computing that relies solely on distant cloud services. It discusses how cloudlets can provide nearby compute resources to allow mobile devices to function as thin clients, improving latency issues. The paper presents an approach called dynamic VM synthesis to rapidly customize cloudlet infrastructure by delivering a VM overlay from a mobile device to launch on the cloudlet. It discusses challenges in deploying this system at scale and suggests areas for future performance improvements.
The case for vm based cloudlets in mobile computing
- 1. The Case for VM-based Cloudlets in Mobile Computing Mahadev Satyanarayanan Paramvir Bahl Ramon Caceres Nigel Davies Asanka Nissanka Published in 2009
- 2. Only a dream in the mid-1990s Now a reality “Information at my finger tips at any time and place”
- 3. Awaiting discovery is an entirely new world in which mobile computing seamlessly augments the cognitive abilities of users using compute- intensive capabilities Natural Language Processing Computer Vision and Graphics Speech Recognition Machine Learning Augmented Reality Planning Decision Making
- 4. This Paper Discuss about technical obstacles of compute-intensive mobile computing Propose a new system architecture to overcome those obstacles Discuss deployment challenges of the proposed system Suggest possible performance improvements to the proposed system
- 5. Resource-Poor Mobile Hardware Mobile hardware is necessarily resource-poor relative to static client and server hardware Improving size, weight and battery life are higher priorities than enhancing compute power
- 6. Cloud Computing for Mobile Cloud Services Cellular Network Mobile Device
- 7. Limits of Cloud Computing Latency Hurts WAN delays in the critical path of user interaction Humans are acutely sensitive to delay and jitter Immersive tasks become sluggish WAN Latency Unlikely to Improve Prime targets Bandwidth Security Energy Efficiency Manageability Transmission of large data sets Wireless Internet (HSDPA) is 2 Mbps Bandwidth Induced Delays
- 8. Cloudlets Data Center in a Box Self Managing Low Power Consumption Internet Connectivity Access Control
- 9. How Cloudlets can Help Nearby resource-rich cloudlet Mobile device functions as a thin client Significant computations occur in the cloudlet
- 10. Cloudlet vs Cloud Cloudlet Cloud State Only soft state Hard and Soft state Management Self-managed Professionally administered Environment “Data Center in a box” at business premises Machine room with power conditioning and cooling Ownership Decentralized ownership by local business Centralized ownership by Amazon, Yahoo etc Network LAN latency/bandwidth Internet latency/bandwidth Sharing Few users at a time 100s-1000s of users at a time
- 11. Transient Cloudlet Customization A key challenge is to simplify cloudlet management Solution is transient customization of cloudlet infrastructure using hardware VM technology Pre-use customization and Post-use cleanup Delivering VM state to infrastructure VM migration approach Dynamic VM synthesis (Focus of this paper)
- 12. Approach 1 : VM Migration First suspend already-executing VM VM’s processor, disk and memory state are transferred VM execution is resumed at the destination Feasibility has been confirmed by Internet Suspend/Resume (ISR) System SoulPad
- 13. Approach 2 : Dynamic VM Synthesis Performance is determined solely by local resources Bandwidth to cloudlet Compute power of the cloudlet WAN failures do not affect synthesis Overlay is delivered from the mobile
- 14. Feasibility of Dynamic VM Synthesis Kimberley proof-of-concept prototype Transient binding between device and cloudlet controlled by Kimberley Control Manager (KCM) Supports browsing and publishing of services using Avahi mechanism in Linux Uses VirtualBox as VMM Kimberlize tool is used to create VM overlays Differences launch VM with base VM to obtain overlay
- 15. Dynamic VM Synthesis Bindings to Cloudlet Infrastructure Establish secure TCP tunnel using SSL between KCM on a device and a cloudlet Cloudlet KCM fetches the VM overlay from mobile Decrypts and decompress the overlay and apply to the base VM Launch the VM Speed of VM Synthesis VM Overlay is 100-200MB Takes less than 1min at 100Mbps
- 16. Improving Performance Overlay transmission Higher-bandwidth short-range wireless network Decompressing and applying the overlay on Cloudlet Exploit parallelism to reduce decompression and overlay application times Partitioning the VM image so that multi- core cloudlet can synthesize the parts Use caching and prefetching techniques VM Synthesis Time Depends on
- 17. Deployment Challenges Choosing between bottom-up and top-down deployment Structuring pricing plans for top-down deployment Deciding on processing, storage and networking capacity needed Defining management policies for maximum user experience and minimum cost Trust and Security Trust establishment Reputation based trust
- 18. Conclusion Resource poverty is a fundamental constraint that severely limits the class of applications that can be run on mobile devices Connecting to distant cloud over WAN introduces delays and jitter Connecting to nearby Cloudlet is fast Proximity of the cloudlet make crisp interactive response for immersive applications that augment human cognition Rapid customization of cloudlet infrastructure, is achievable through dynamic VM synthesis
- 19. Future Work Explore about performance improvements
- 20. Thank You Any Questions ?