1、Performance attributes;C.2.4Computer-Communication Net-works:Distributed SystemsGeneral Terms:Design,Experimentation,Measurement,PerformanceAdditional Key Words and Phrases:Virtual Machines,KVM,High-Performance Computing1.INTRODUCTIONA current problem in scienti c computing is that the expertise nee
2、ded to deploy and maintain a cluster remains scarce despite recent declines in hardware costs. Smaller research groups may not be able to a ord to have their own clusters,and purchasing time on an existing cluster raises concerns such as vendor lock-in and data security.In this paper,we present a cl
3、oud computing model which de nes a minimum speci cation for a compute cluster.This clusters sole purpose is to host other com-pute clusters through virtualization.In this way,a Virtualization Service Provider (VSPcan sell compute power without having to directly maintain each end-users particular ap
4、plication.Similarly,Virtual Organizations(VOscan purchase compute power from VSPs without having to worry about hardware or software compatibility.A VO is free to develop a model cluster locally,perhaps even on a personal workstation,test it,andAuthorsaddress:School of Computing,Clemson University,C
5、lemson,SC29634-0974,USA. mfenn,mamurph,jim.martin,sebgoacs.clemson.eduThis material is based upon work supported under a National Science Foundation Graduate Re-search Fellowship.Permission to make digital/hard copy of all or part of this material without fee for personal or classroom use provided t
6、hat the copies are not made or distributed for pro t or commercial advantage,the ACM copyright/server notice,the title of the publication,and its date appear,and notice is given that copying is by permission of the ACM,Inc.To copy otherwise,to republish, to post on servers,or to redistribute to list
7、s requires prior speci c permission and/or a fee.c 20YY ACM1529-3785/20YY/0700-0001$5.00ACM Transactions on Computational Logic,Vol.V,No.N,Month20YY,Pages1 0?.2M.Fenn et al. (aType1(bType2Fig.1.Hypervisorsthen deploy it to a VSPs hardware with reasonable assurances that the operating environment wil
8、l be fully compatible.We will rst provide a brief overview of virtualization technologies,followed by a description of our model virtual cluster.Then,we will de ne the infrastructure for which an VSP would be responsible.Finally,we will present some results of a model cluster constructed at the Cybe
9、rinfrastructure Research Laboratory at Clemson University.2.VIRTUALIZATION MODELIn essence,virtualization is making one computer appear to be multiple computers. Jones2006Virtualization is accomplished with a program called a hypervisor, while systems running under a hypervisor are known as virtual
10、machines(VMs. There are two basic types of hypervisorIBM2005:Type1hypervisors directly interface with the system hardware.All operating systems run inside a virtual machine.There is usually a special,privileged virtual machine that can manage the others.Xen is an example of this type of hypervisor.
11、Type2hypervisors run as a normal program inside a normal operating system. This OS is known as the host.Each guest OS runs as a process in the host OS. These processes can be manipulated just like any other process.VMWare and KVM are examples of this type of hypervisor.See Figure1for a comparison of
12、 Type1and2hypervisors.A strict Type2hypervisor requires that all I/O devices be emulated completely in software,resulting in added overhead for I/O calls.Paravirtualization allows the virtual machine to make calls directly to the hypervisor,resulting in potentially increased e ciency.Paravirtualizat
13、ion requires modi cations to the guest kernel. IBM2005See Table I for a comparison of KVM and Xen.ACM Transactions on Computational Logic,Vol.V,No.N,Month20YY.An Evaluation of KVM for Use in Cloud Computing3Table I.KVM vs.XenKV M XenT ype1Hypervisor T ype2HypervisorHost is a privileged guest Host di
14、rectly on hardwareUnprivileged guests Guests have user privilegesx86ring abstraction UNIX process abstractionP aravirtualized guests Unmodified guests2.1Kernel-based Virtual Machine(KVMThe Kernel-based Virtual Machine(KVMis a Type2hypervisor maintained by Qumranet,IncHabib2008Qumranet2006.KVM is bas
15、ed on the QEMU emu-lator and derives all its management tools from QEMU.The main focus of KVM development is to use the x86VT extensions,which allow virtual machines to make system callsvan Doorn2006.KVM versions newer than KVM-62have support for paravirtualized Linux guests,but we did not utilize t
16、his capability in our initial prototype.KVM uses a set of Linux kernel modules to provide VT support.KVM can run on a stock Linux kernel that is:(anew enough and(bhas had the KVM modules built for it.In contrast,Xen requires a heavily patched Linux kernel,on which development lags behind the mainlin
17、e kernel.KVM supports the QEMU Copy-on-write(QCOWdisk image format,allowing it to support a snapshot mode for its disk I/O operations.In snapshot mode, all disk writes are directed to a temporary le,and changes are not persisted to the original disk image le.Multiple VMs can be run from one disk ima
18、ge, somewhat mitigating the huge storage requirements associated with hosting a grid of VMsKeahey et al.2004.Destroying a virtual cluster is as simple as sending SIGKILL to each hypervisor and deleting the image from disk.KVM supports the standard Linux TUN/TAP model for Ethernet bridging.By using t
19、his model,each VM gets its own networking resources,making it indistin-guishable from a physical machine.2.2Virtual Compute NodesCentral to the Virtual Organization Cluster Model(VOCMis the Virtual Organi-zation Cluster(VOC,which is composed of Virtual Compute Nodes(VCN.Each Virtual Organization(VOt
20、hat wishes to utilize the compute facilities provided by a Virtualization Service Provider(VSPmust provide a VM image or set of VM images,along with some general con guration parameters.Since each image will potentially be used to spawn multiple VMs,the con guration of each image must not make any a
21、ssumptions about the type of networking(hardware interface, hostname,or system-speci c con guration settings.Instead,dynamic networking con guration should be used.Once a hostname has been obtained,dynamic con- guration based upon the hostname is allowed.Our model VOC was built from two VCNs,each wi
22、th CentOS5.1installed. CentOS provides substantial out-of-the-box support for cluster computing appli-cations and,along with its cousin,Red Hat Enterprise Linux,is widely supported in the scienti c and high-performance computing communities.The two VCNs were:4(1A virtual head node,which was con gure
23、d with the Condor central managerand submit daemons(condor_collector,condor_negotiator,condor_schedd, Ganglia monitoring daemon(gmond,and Ganglia meta daemon(gmetad.(2A virtual compute element,which was con gured with the Condor job starter(condor_startd,MPICH2,ATLAS(tuned for the virtual CPU,and Ga
24、nglia monitoring daemon(gmond.Our model VOC was designed as an Open Science Grid(OSGcompute element. The virtual head node used Condor to distribute incoming OSG jobs to the virtual compute elements.3.SUPPORT MODELPreparing the physical(as opposed to virtualcluster for VOC support required con gurin
25、g the host OS,setting up support services,con guring networking ser-vices,and con guring storage services.In our prototype implementation,support services included a Lightweight Directory Access Protocol(LDAPserver for cen-tralized administration of hosts and physical user accounts,a Dynamic Host Co
26、n- guration Protocol(DHCPserver for assigning IPv4addresses to nodes,and a Domain Name Server(DNSfor host resolution.3.1Host OS con gurationWhen providing virtualization services,the host OS should be minimalist in order to reserve as many resources as possible for the VCNs.To this end,Slackware Lin
27、ux 12was chosen as the host OS.A custom kernel was compiled to enable support for KVM and additional network metrics.All unnecessary hardware drivers and other features were left out of the kernel to minimize its memory footprint.KVM driver modules also were built for the custom kernel.For maintaina
28、bility reasons,all the Slackware nodes were installed via PXE boot and a custom automated install script.This allowed the whole cluster to be re-created quickly in case of added nodes,hardware failure,or administrator error.All additional software was maintained in the form of Slackware packages to
29、allow for rapid deployment across the entire cluster.3.2Physical Support ServicesCon guration information for each VCN was stored in an LDAP database to pro-vide a centralized administration mechanism.Each VCN was represented as an LDAP entry with the hostname,IP address,and MAC address elds.The MAC address was generated as a locally-administered address as to avoid con icts with any other MACs on the network.An LDAP-aware,batch,remote administration tool
copyright@ 2008-2022 冰豆网网站版权所有
经营许可证编号:鄂ICP备2022015515号-1