DOC PREVIEW
Berkeley COMPSCI 294 - SLA Based Management of a Computing Utility

This preview shows page 1-2-3-4-5 out of 14 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 14 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 14 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 14 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 14 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 14 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 14 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

START.PDFReturn to Menu1Océano – SLA Based Management of aComputing UtilityK. Appleby, S. Fakhouri, L. Fong, G. Goldszmidt, M. Kalantar, S. Krishnakumar,D.P. Pazel, J. Pershing, and B. RochwergerIBM T.J. Watson Research CenterP.O. Box 704Yorktown Heights, New York 10598, USAContact address: [email protected]éano is a prototype of a highly available, scaleable, and manageable infrastructurefor an e-business computing utility. It enables multiple customers to be hosted on acollection of sequentially shared resources. The hosting environment is divided intosecure domains, each supporting one customer. These domains are dynamic: theresources assigned to them may be augmented when load increases and reducedwhen load dips. This dynamic resource allocation enables flexible Service LevelAgreements (SLAs) with customers in an environment where peak loads are an orderof magnitude greater than the normal steady state.KeywordsCluster Management, Service Level Agreements, Network and System Monitoring,Computing Utility1. IntroductionOcéano is a prototype of a highly available, scaleable, and manageable infrastructurefor an e-business computing utility. It enables multiple customers to be hosted on acollection of shared resources. At any point of time each resource is assigned to onlya single customer. That is, the hosting environment is divided into secure domains,each supporting one customer. These customer domains are dynamic: the resourcesassigned to them may be augmented when load increases and reduced when loaddips. This dynamic resource allocation reduces hosting costs while providing amechanism to guarantee contracted Service Level Agreements (SLAs).In e-business hosting, customers increasingly require support for peak loads thatare an order of magnitude greater than those experienced in normal steady state [16].A common commercial hosting model is colocation model, in which dedicatedresources are permanently assigned to each customer. In this model, supporting peakworkloads requires significant resource overprovisioning. Océano provides a morecost effective alternative by automatically and dynamically reassigning resources tomeet demands as they occur.0-7803-6719-7/01/$10.00 (C) 2001 IEEE2Océano manages the resources of the computing utility so that each hostedcustomer is furnished the resources necessary to provide a contracted level of serviceas specified by a Service Level Agreement (SLA) [5]. Monitoring agents issueevents when thresholds are exceeded. Events are correlated to identify root causes,which are reported to a central Resource Director responsible for planning resourceallocation and recovery actions. The main controls available are load sharing viadynamic allocation and de-allocation of servers, and throttling incoming requests.Underlying subsystems provide the mechanisms for managing resources and shiftingthem from one customer domain to another in real time (minutes) withoutcompromising security requirements.Other approaches to sharing resources among multiple hosted customers orapplications [3,7,14,15,16] focus on sharing the CPU cycles of single servers. Incontrast, Océano focuses on sequential sharing at the granularity of whole servers,and the management of a whole farm of servers. Approaches that also sequentiallyallocate whole servers allow only static server allocation. These approaches, unlikeOcéano, make no attempt to modify the computing environment to satisfy theallocation (for example, by installing an operating system or by changing thenetwork configuration). Rather, they allocate the server as is. A matchmakingalgorithm [13] must ensure that the environment is suitable. In general, theseapproaches create virtual domains without providing network isolation. IcorpMaker[15] does provide isolation via virtual private networks; Océano does so via virtualLANs. Finally, the Galaxy project [18] focuses on providing a variety of tools tobuild Windows-NT clusters for multiple different purposes. It does not provide SLAmonitoring or automatic resource reprovisioning in response to performancebottlenecks, only in response to failures. Océano provides a unique and morecomprehensive combination of technologies to address a number of issues ignoredby these approaches, including SLA driven monitoring, event correlation, networktopology discovery, and automatic network reconfiguration.AdministrativeNode............................Whale 1 Whale 2 Whale MNetworkDispatcherDolphin NDolphin 1 Dolphin 2Figure 1: Overview of one domain in Océano0-7803-6719-7/01/$10.00 (C) 2001 IEEE3In this paper, the term customer refers to a third party hosted on an “e-businesscomputing utility.” The utility is composed of a server farm that is managed by theOcéano distributed software. Customers are supported by dynamically dividing thephysical infrastructure into customer domains. The implementation of Océanodescribed here assumes that the physical infrastructure consists of 3 tiers: (1) front-end IP sprayers for load balancing (e.g. Network Dispatchers [9]), (2) a large pool ofallocable “Dolphin” servers, and (3) a pool of fixed “Whale” servers, allinterconnected by a switched network. Figure 1 shows one domain. Each domaincontains the servers currently allocated to a particular customer. The Dolphins can beallocated to a domain and later reallocated to another. On the other hand, the Whalesare permanently assigned to a domain. In practice, a customer’s database wouldreside on the fixed Whale servers. All servers are connected to an administrative, ormanagement, domain. A small set of dedicated nodes execute Océano managementapplications. Management and monitoring agents may reside on customer assignedDolphin and Whale servers.The rest of this paper provides an overview of Océano. Section 2 describes howOcéano uses SLAs to identify metrics to monitor, how they are monitored and howmonitored events are correlated. Section 3 describes the Resource Director. Themechanisms to manage bandwidth, servers, and customer data and applications aredescribed in Section 4. Section 5 describes the underlying infrastructure andconfiguration information. Section 6 reports on the status of our prototypeimplementation and reports some initial performance statistics and identifies someareas of future research. Finally, Section 7 concludes.2. SLA based ManagementKey to all resource management in Océano, is the management of customer SLAs.This is achieved by


View Full Document

Berkeley COMPSCI 294 - SLA Based Management of a Computing Utility

Documents in this Course
"Woo" MAC

"Woo" MAC

11 pages

Pangaea

Pangaea

14 pages

Load more
Download SLA Based Management of a Computing Utility
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view SLA Based Management of a Computing Utility and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view SLA Based Management of a Computing Utility 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?