SLA ACHIEVEMENT BY NEGOTIATION IN A CLOUD S. Jaya Nirmala Assistant Professor Dept. of C.S.E. NIT Trichy sjaya@nitt.edu Shah Maulik M.Tech Student Dept. of C.S.E. NIT Trichy maulik_k_shah@yahoo.co.in S. Mary Saira Bhanu Associate Professor Dept. of C.S.E. NIT Trichy msb@nitt.edu ABSTRACT The performance of a computing cloud depends on its agile response to the user needs and in the quality of the service provided. SLA assurance will increase user’s confidence in the system by guaranteeing the cloud resources needed to users at certain point in the future. Negotiation between the cloud provider and consumer is necessary to achieve SLA. Haizea is an open source lease manager which can act as a scheduler for OpenNebula. It uses resource leases as resource allocation abstraction and executes these leases by allocating virtual machines. Haizea does not support negotiation currently. This work implements negotiation process using Haizea to provide SLA based lease admission. The experiment results show that the proposed method increases system utilization and maximizes the number of leases accepted as compared to Haizea’s default policies. Keywords Cloud, SLA, Haizea, OpenNebula, Leases, Advance Reservation 1. INTRODUCTION Cloud computing allows a customer to host his applications over the Internet and helps to reduce the hardware cost and thereby improve productivity and reduce downtime. One of the biggest challenges cloud computing facing is providing Service Level Agreement (SLA) assurance. It offers flexibility to the clients by delivering the service according to their needs. Infrastructure as a Service (IaaS) provides resources based on predefined resource allocation policy. For example, OpenNebula’s default scheduler is designed for immediate scheduling. If a request cannot be scheduled by the OpenNebula scheduler for lack of resources it is rejected. This problem can be solved by using advanced and complex resource allocation policies at the cloud provider side. Haizea is an open source lease manager that addresses this issue. It replaces OpenNebula’s match-making scheduler with its own scheduler. It allows consumers to lease the resources using potentially complex policies instead of just allowing immediate scheduling [5]. Computational resources are requested in the form of lease through Haizea. Four types of leases are supported by Haizea: Immediate, best effort, advance reservation (AR) and deadline sensitive. It provides facility to reserve resources in advance with OpenNebula and it does not include negotiation in this process. Negotiation is done with advance reservation leases to guarantee the availability of the resources at specified time. This is done by using VMs for execution at a given time in the leases. SLA is a contract between two entities (the client and the provider), a guarantee of access to certain resources, offered by the provider to the client in order to deliver a certain quality of service. Haizea uses two new data structures to store resource lease information: Lease and Slottable. Lease is used to store all information about an accommodated lease while slottable is used to track the capacity of the physical nodes on which leases can be scheduled. It contains the resource reservation of all the leases and allows efficient access to them. The paper uses two algorithms to achieve SLA. The first generates counter offers at provider side and the second selects one among the set of counter offers generated, at the consumer side. The remainder of the paper is organized as follows. Section 2 describes the need for SLA. Section 3 lists the related work. Section 4 gives details of the algorithms used to achieve SLA. Section 5 discusses experimental setup and results. Section 6 is conclusion and future work. 2. NEED OF SLA FOR ADVANCE RESERVATION Advance reservation is necessary to use the resources at specific time. User can submit an advance reservation lease with specific start time, duration, number of resources required, etc. The provider usually checks for the availability. If the resources are available then it accepts the lease and if resources are not available it rejects the lease. The drawback of this method is that it increases the communication overhead because consumer does not know which parameter (start time, duration, and number of resources) had led to the rejection of the lease. So, the chances of acceptance of the modified lease from consumer decrease. This process also results in lower utilization of existing resources. Negotiation is necessary between the consumer and the provider to overcome this problem. Here, the provider sends the request rejection message along with the set of counter offers to the consumer. Counter offers are generated by looking into the slottable and finding resource allocation status at the current time. Consumer can use a ranking algorithm to select the best suitable counter offer from the set of counter offers sent by the provider. If both Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from Permissions@acm.org. Compute '13, August 22 - 24 2013, Vellore, Tamil Nadu, India Copyright 2013 ACM 978-1-4503-2545-5/13/08..$15.00. http://dx.doi.org/10.1145/2522548.2523135