SYNOPSIS ON Going Back and Forth: Efficient Multi deployment and Multi snapshotting on Clouds




3.3 Cloud Infrastructure 4. 3 4 5 6 7 8 8 9 11 11 12 15 16 17 References .1 Advantage 6. Problem Definition Abstract Aim & Objective Literature Review 4. Plan of Action Software & Hardware Requirements Page No. 4.1 Existing System 4. Scope 5.Table of Contents Title 1.1 Data flow diagram 6. 2.4 Application State 5.2 System Architecture 7. High Level Design 6.2 Disadvantages 4. 8.

.1. for check pointing or off-line migration) to another cluster or cloud . starting from a set of VM images previously stored in a persistent fashion • Multisnapshotting : Many VM images that were locally modified need to be concurrently transferred to stable storage with the purpose of capturing the VM’s state for later use (e. . Problem Definition To provide the basic functionalities for the use of Virtual Machine’s (VM’s) over the cloud given as: • Multideployment : The operation of Infrastructure As A Service (IaaS) is the need to deploy a large number of VMs on many nodes of a datacentre at the same time.g.

it is important to enable efficient concurrent deployment and snapshotting that are at the same time hypervisor independent and ensure a maximum compatibility with different configurations.2. With datacenters growing rapidly and configurations becoming heterogeneous. Once the VM instances 1are deployed. . This paper addresses these challenges by proposing a virtual file system specifically optimized for virtual machine image storage. another challenge is to simultaneously take a snapshot of many images and transfer them to persistent storage to support management tasks. such as suspend-resume and migration. It is based on a lazy transfer scheme coupled with object versioning that handles snapshotting transparently in a hypervisor-independent fashion. ensuring high portability for different configurations.Abstract Infrastructure as a Service (IaaS) cloud computing has transform the way we think of acquiring resources by introducing a simple change: allowing users to lease computational resources from the cloud provider’s datacenter for a short time by deploying virtual machines (VMs) on these resources. This new model raises new challenges in the design and development of IaaS middleware. One of those challenges is the need to deploy a large number (hundreds or even thousands) of VM instances simultaneously.

Aim and Objectives In this project.2. We want to reduce the contention on current system & allow maximum number of user to access VM’s with quick resume. . A focus on the use of reusable frameworks to provide cost and times benefits. restart & suspend operation. Our aim is to create and implement load balancing mechanisms. for managing numbers of users at the same time and within the context of time slices. One resource can be available to the all the users . They’re focused on coming up with solutions that serve customer requirements today and anticipate future anticipating their requirements we provide them without confliction. we are creating cloud infrastructure which allows users to lease computational resources from cloud provider.

g. since it enables users to expand or shrink their resources according to their computational needs. We refer to this pattern as multideployment. A typical deployment consists of hundreds or even thousands of such images. starting from a set of VM images previously stored in a persistent fashion. The on-demand nature of IaaS is critical to making such leases attractive. This problem is particularly acute for VM images used in scientific computing where image sizes are large (from a few gigabytes up to more than 10 GB). a similar challenge applies to snapshotting the deployment: many VM images that were locally modified need to be concurrently transferred to stable storage with the purpose of capturing the VM state for later use (e.. Literature Survey In recent years. Therefore. One of the commonly occurring patterns in the operation of IaaS is the need to deploy a large number of VMs on many nodes of a datacentre at the same time. For example.4. Leasing of computation time is accomplished by allowing users to deploy virtual machines (VMs) on the datacentre’s resources[1]. which limits the ability to easily migrate VMs among different hypervisors. by using external resources to complement their local resource base. Custom image formats are not standardized and can be used with specific hypervisors only. Infrastructure as a Service (IaaS) cloud computing has emerged as a viable alternative to the acquisition and management of physical resources. such approaches generate a large number of files and interdependencies among them. configurations are becoming more and more heterogeneous. Furthermore. with growing datacentre trends and tendencies to federate clouds. This emerging model leads to new challenges relating to the design and development of IaaS systems. which are difficult to manage and which interfere with the ease-of-use rationale behind clouds. We refer to this pattern as multisnapshotting.[1] Once the VM instances are running. This can make the response time of the IaaS installation much longer than acceptable and erase the ondemand benefits of cloud computing. Conventional snapshotting techniques rely on custom VM image file formats to store only incremental differences in a new file that depends on the original VM image as the backing file. not counting the time to boot the operating system itself. Such a large deployment of many VMs at once can take a long time. Conventional deployment techniques broadcast the images to the nodes before starting the VM instances. With IaaS. IaaS leasing is equivalent to purchasing dedicated hardware but without the long-term commitment and cost. users can lease storage and computation time from large datacentre’s. multisnapshotting must be handled in a transparent and portable . When taking frequent snapshots for a large number of VMs. Since the user has complete control over the configuration of the VMs using on-demand deployments. a process that can take tens of minutes to hours. this pattern occurs when the user wants to deploy a virtual cluster that executes a distributed application or a set of environments to support a workflow. for check pointing or off-line migration to another cluster or cloud).

and data and processing outsourcing. Since the patterns are complementary. 4. storage space. we investigate them in conjunction. Cloud Computing is a “buzz word” around a wide variety of aspects such as deployment. Clouds have been defined just as virtualized hardware and software plus the previous monitoring and provisioning technologies. load balancing. 4. Our proposal offers a good balance between performance.1 Existing system The huge computational potential offered by large distributed systems is hindered by poor data sharing scalability. Moreover. The role of virtualization in Clouds is also emphasized by identifying it as a key component. 4. It is not possible to build a scalable. while keeping maximum portability among different hypervisor configurations. This paper proposes a distributed virtual file system specifically optimized for both the multideployment and multisnapshotting that hides the interdependencies of incremental differences and exposes standalone VM images. and network traffic consumption. fine-grain data accesses [1]. One such requirement is the need to efficiently cope with massive unstructured data (organized as huge sequences of bytes . high-performance distributed data-storage service that facilitates data sharing at large scale. raw image files (understood by most hypervisors) to the outside[6]. We addressed several major requirements related to these challenges. while handling snapshotting transparently and exposing standalone. Disk storage (cheap hard-drives with .3 Cloud infrastructure IaaS platforms are typically built on top of clusters made out of loosely-coupled commodity hardware that minimizes per unit cost and favours low power over maximum speed. Network traffic consumption also very high due to non concentrating on application status.BLOBs that can grow to TB) in very large-scale distributed systems while maintaining a very high data throughput for highly concurrent. these patterns may also generate high network traffic that interferes with the execution of applications on leased resources and generates high utilization costs for the user. provisioning.2 Disadvantages To give less performance and storage space. In addition to incurring significant delays and raising manageability issues.

store communication channels and resend lost information[5]. With the recent explosion in cloud computing demands. saving 2 GB of RAM for 1. Therefore. In order to provide persistent storage. the necessary storage space can explode to huge sizes. virtual topology. etc. RAM. the issue of capturing the global state of the communication channels is difficult and still an open problem. which is used to store only minimal information about the state. In order to avoid this issue. CPU registers.compatible between different hypervisors. since the contents of RAM. Such an approach has two important practical benefits: (i) Huge reductions in the size of the state. which is unacceptable for a single one. while the machines are interconnected with standard Ethernet links. saving the application state implies saving both the state of all VM instances and the state of all active communication channels among them. in terms of both hardware and software. and (ii) Portability since the VM can be restored on another host without having to worry about restoring the state of hardware devices that are not supported or are in.[5] This information is then later used to reboot and reinitialize the software stack running inside the VM instance. While several methods have been established in the virtualization community to capture the state of a running VM (CPU registers. Model 2 can further be simplified such that the VM state is represented only by the virtual disk attached to it (Model 3). such that they are able to host the VMs. state of devices. such as configuration files that describe the environment and temporary files that were generated by the application. 4. Even so. Any in-transit network traffic is discarded. a dedicated repository is deployed either as centralized or as distributed storage service running on dedicated storage nodes. download. in the most general case . in-transit network packets. and the like does not need to be saved. the general case is usually simplified such that the application state is reduced to the sum of states of the VM instances .).). etc. delete.point-in-time deployment checkpoint. The machines are configured with proper virtualization technology.capacities in the order of several hundred GB) is attached to each machine. for VM instances that need large amounts of memory. Thus.000 VMs consumes 2 TB of space. The repository is responsible for storing the VM images persistently in a reliable fashion and provides the means for users to manipulate them: upload.4 Application state The state of the VM deployment is defined at each moment in time by two main components: the state of each of the VM instances and the state of the communication channels between them (opened sockets. there is an acute need for scalable storage. under the assumption that a fault-tolerant networking protocol is used that is able to re. For example. and so forth. .

Our proposal offers a good balance between performance. we describe an implementation on top of Blob Seer. and network traffic consumption. 5. storage space.1 ADVANTAGE A good balance between performance. . we investigate them in conjunction. a versioning storage service specifically designed for high throughput under concurrency. storage space. We show how to realize these design principles by building a virtual file system that leverages versioningbased distributed storage services. while handling snapshotting transparently and exposing standalone. We introduce a series of design principles that optimize multi deployment and multi snapshotting patterns and describe how our design can be integrated with IaaS infrastructures. raw image files (understood by most hypervisors) to the outside[4].5. while handling snapshotting transparently and exposing standalone. and network traffic consumption. To illustrate this point. raw image files. Scope A distributed virtual file system specifically optimized for both the multi deployment and multi snapshotting patterns. Since the patterns are complementary.


The typical elements found in the cloud are illustrated with a light background. Running the application repeatedly and waiting for it to reach the point where the bug happens might be prohibitively expensive. The cloud middleware interacts directly with both the hypervisor. This approach enables snapshotting to be leveraged in interesting ways. Since all image snapshots are independent entities. Furthermore. The cloud middleware [2] in turn coordinates the compute nodes to achieve the afore-mentioned management tasks. A distributed versioning storage service that supports cloning and shadowing is deployed on the compute nodes and consolidates parts of their local disks into a common storage pool. is performed in the following fashion. Every uploaded image is automatically striped.Plan of Action The simplified architecture of a cloud that integrates our approach is depicted in Figure 1. which is responsible for ondemand mirroring and snapshotting and relies on both the local disk and the distributed versioning storage service to do so. including deploying an image on a set of compute nodes.7. CLONE and COMMIT can also be exposed by the cloud middleware at the user level through the control API for fine-grained control over snapshotting. However. Each compute node runs a hypervisor [1] that is responsible for running the VMs. and the mirroring module. while the elements that are part of our proposal are highlighted by a darker background. The first time the snap-shot is taken. telling it when to start and stop VMs.suing each mirroring module a COMMIT to its corresponding clone. CLONE is broadcast to all mirroring modules. A global snapshot of the whole application. let’s assume a scenario where a complex. when to create a new image clone (CLONE) [2]. For ex-ample. followed by COMMIT. subsequent global snapshots are performed by is. telling it what image to mirror from the repository. which involves taking a snapshot of all VM instances in parallel. The cloud client has direct access to the storage service and is allowed to upload and download images from it. dynamically adding or removing compute nodes from that set. and snapshotting individual VM instances or the whole set. the application right before the bug happens. fully independent VM image that is globally accessible through the storage service and can be deployed on other compute nodes or manipulated by the client. The reads and writes of the hypervisor are trapped by the mirroring module [1]. the cloud client interacts with the cloud middleware through a control API that enables a variety of management tasks. Once a clone is created for each VM instance. Both CLONE and COMMIT are control primitives that result in the generation of a new. and when to persistently store its local modifications (COMMIT) [2]. they can be either collectively or independently analyzed and modified in an attempt to . distributed application needs to be debugged.

Such an approach is highly useful in practice at large scale because complex synchronization bugs [3] tend to appear only in large deployments and are usually not triggered during the test phase. .fix the bug. If the attempt was not successful. the application can safely resume from the point where it left. the approach can continue iteratively until a fix is found. which is usually performed at smaller scale. Once this fix is made.

8.1 Hardware requirement : CPU type Clock speed Ram size Hard disk capacity : Dual Core : 2.2 Software requirement : Operating System: Windows Language Back End Documentation : JAVA(JDK-1. Software And Hardware Requirements 8.6) : MS-Access : Ms-Office .65 GHz : 2 GB : 40 GB 8.

[6] Hypervisor Alternative: pages 91 [5]K.pdf [3]M. Lee.I. Commun. Konwinski. B. and O. Claudel.nimbusproject. Armbrust. A view of cloud computing.A.Ameet Talwalkar (2012) Foundations of Machine Learning.USENIX Association. Katz. Taktuk. the MIT Press ISBNhttp://hal. In HPDC’09: Proceedings of the 18th ACM International Symposium on High Performance Distributed Computing. G. Afshin Rostamizadeh. Griffith. GA. Fox. Zaharia. D. G. R. .References : [1]Going Back and Forth: Efficient Multi deployment and Multi snapshotting on Clouds: www. Freeman. Joseph. Keahey and T. and M.53:50 [4]327. 2000. Stoica.4. Richard. A. Patterson. R.adaptive deployment of remote executions. Science clouds: Early experiences in cloud computing for scientific applications.inria. 2008. Huard.pdf [2]Mehryar Mohri. ACM.A. In CCA’08: Proceedings of the 1st Conference on Cloud Computing and Its Applications. Rabkin.

Left : 1. The paper should be A4 size with margins: Top: 1”.5 Paragraph spacing – 2 lines 4.Guidelines 1.5” . Subtitles Times New Roman 12.25” 5. Written matter – Times New Roman 12 (justified) Line spacing – 1. Number of copies -.02 (Project guide + Project Coordinator) . bold. 6. Headings in the chapters should have size Times New Roman 14. center. Page nos.2” . Right : 1. 2. bold. should be at the center of bottom of each page. Report should be spiral bound with white plastic cover pages only. Bottom : 1. 3. left aligned.

. Seminar report should be as per the guidelines only.7.