《阳光棕榈餐饮连锁事业加盟申请表Titl6》由会员分享，可在线阅读，更多相关《阳光棕榈餐饮连锁事业加盟申请表Titl6（49页珍藏版）》请在金锄头文库上搜索。
1、Xen and Co.:Communication-Aware CPU Management in Consolidated Xen-Based Hosting Platforms報告人：倪丞頤1AuthorSriram Govindan,Jeonghwan Choi,Arjun R. Nath,Amitayu Das,Bhuvan Urgaonkar, Member, IEEE,and Anand Sivasubramaniam, Member, IEEE2Abstract (VHPs)Virtualization-based hosting platformsUsing this tech
2、nology in the design of consolidated hosting platformsVirtualization enables easier and faster application migration as well as secure co-location of antagonistic applications3Abstract (shortcomings in VMMs)Virtual Machine MonitorsTwo shortcomings in VMMsnCPU schedulers that are agnostic to the comm
3、unication behavior of modern, multi-tier applicationsnInadequate or inaccurate mechanisms for accounting the CPU overheads of I/O virtualization4Multi-tier Applications5GoalAchieve:nCommunication-aware CPU scheduling algorithmnCPU usage accounting mechanismImplement:nIn Xen VMMnA prototype VHP on a
4、cluster of 36 serversExperiment:nRealistic Internet servernTPC-W benchmark6The need for communication-aware CPU scheduling: A TPC-W benchmarkTwo tierA Jboss tiernApplication logic and interacts with the clientsA Mysql-based database tiernStores information about items7CPU Usage: Isolated ServerswEac
5、h tier of this application was run on a separate physical server8CPU Usage: Consolidated ServerwRunning this application with all its tiers consolidated on a single server running Xen along with five CPU-intensive applications.9Client Response TimeswComparing the performance experienced by the clien
6、ts of this application under the two scenarios.10Analysis of Response timeQuestion:nWhy did the performance degrade despite providing the same resource allocations?Reason:nFor applications with communicating components, providing enough CPU alone is not enough an equally important consideration is t
7、o provide CPU at the right time.11Problem 1.Can a server in a VHP schedule hosted VMs in a communication-aware manner to enable satisfactory application performance even under conditions of high consolidation, while still adhering to the high-level resource provisioning goals in a fair manner?12The
8、need for accurate accounting of the overheads of virtualizationIn VHP, the unit of resource allocation and accounting for applications changes to VMVMM is responsible for virtualizing I/O devices.Problem:nCan a server in a VHP account for the overheads of virtualization and incorporate them into its
9、 CPU scheduling to provide fair resource allocations?13Research ContributionsDevelop a CPU scheduling algorithm for a VMM that incorporates the I/O behavior of the overlying VMs into its decision making.Develop an algorithm that accurately and efficiently accounts the CPU overheads resulting from se
10、rver virtualization and attributes these to the VMs they originate from.Identify ways of implementing these algorithms in the state-of-the-art Xen VMM.14Background VMMnVirtualization of a server at the operating system level by a software layernVirtualizes the resources of a physical servernSupports
11、 the execution of multiple VMsnXen VMM1516DomainDomain0nPrivileged VMnImplements the real device driversnDoes the translation between virtual and real I/O activityNetfront drivernGuest domain implements a driver for its virtual NICNetback drivernDomain0 implementsnbetween netfront drivers and the de
12、vice driver for the physical NIC17Network I/O virtualization in Xen. (a) Reception18Network I/O virtualization in Xen. (b) transmission19Hosting Model AssumesLarge cluster of high-end serversHigh-bandwidth networkConnected to a consolidated high capacity storage20Enhanced HypervisorA modified CPU sc
13、heduler to achieve communication-aware scheduling of domainsMechanisms that measure and maintain relevant resource usage statistics pertaining to the overlying domains that are used by our scheduler21Communication-Aware Scheduling In A VHPThe VMM provides a CPU scheduler that allows applications to
14、specify guarantees on the CPU allocationsXen hypervisor implements an algorithm called Simple Earliest-Deadline-First (SEDF)22Simple Earliest-Deadline-First (SEDF)Each domain specifies a pair (slice, period)Asking for slice units of the CPU every period time unitsHypervisor ensures that the specifie
15、d reservation can be providedWe assume that a domain is not admitted if its reservation cannot be satisfiedThe residual CPU capacity is shared among the contending domains (including Domain0) in a round-robin fashion.23New CPU Scheduling AlgorithmOn top of SEDFAttempts to preferentially schedule com
16、munication-sensitive domains over othersTo reduce the aggregate scheduling-induced delay for the hosted domains while still providing guarantees on CPU allocations24Classifying Scheduling-Induced DelaysDelay at the recipientnNetwork packet is copied from the reception-I/O-ring in Domain0 into its ow
17、n address spaceDelay at the sendernBefore a domain sends a network packet, induced by the hypervisor scheduling other domains in between25Classifying Scheduling-Induced DelaysDelay associated with the scheduling of Domain0nWhen Domain0 copies a packet, reception at the physical network interface, in
18、to the address space of the domain.nBetween when a domain copies a packet into the transmit-I/O ring of Domain0 and when Domain0 gets scheduled next to actually send it over the network interface.26Preferential Scheduling of RecipientDevise a general approach that can choose between multiple recipie
19、nt domains.Pick the domain that has received the most number of packetsNot cause a domain receiving high-intensity traffic to starve other domains by the default SEDF27Implementation ConsiderationsBookkeeping PagesnEach domain is given a page that it shares with the hypervisornWe use these pages to
20、maintain various I/O relatednetwork-reception-intensitynOne for each domainnStored in the bookkeeping page of Domain0nKeeping track of the number of packets received and waiting within Domain028Algorithm: Bookkeeping at the RecipientThe netback driver figures out which domains have received packetsU
21、ses the page-flip mechanism to copy pagesUses the number of pages flipped with each domain as an indicator of the number of packets received by that domainIncrements the network-reception-intensityThe netfront driver of a recipient domain processes some packets receivedHypervisor decrements the netw
22、ork-reception-intensity29Anticipatory Scheduling of SenderUse a simple last-value-like predictionTime instants Ttx and Ttx_1The duration between the last two transmission operations Ttx_1 and TtxA predictor of the duration over which the domain is likely to indulge in a transmission again, Ttx, Ttx+
23、(Ttx-Ttx_1)Choose to schedule the one that is expected to transmit the most packets.30Implementation Considerationsactual-network-transmit-intensitynIn the bookkeeping page of each guest domainanticipated-network-transmit-intensitynFor each guest domainnIn the bookkeeping page of Domain031Algorithm:
24、 Bookkeeping at the SenderStep1:nFor each of the network packet queued (waiting to be transmitted to the NIC) inside the guest domain, the Netfront driver increments the network intensity of that domain,nTransmittingDomain-actual-network-transmit-intensity+Step2:nUpon deschedule of the guest domain,
25、 the Hypervisor increments the network intensity of the guest domain,nTransmittingDomain-anticipated-network-transmit-intensity += TransmittingDomain-actual-network-transmit-intensity.32Scheduling of Domain0By default, the Xen scheduler employs a high reservation of (15 msec, 20 msec) for domain0Two
26、 kinds of packets nPackets written by guest domains to their virtual NICsnPackets received for delivery to domains and waiting in their reception-I/O-ring within Domain033A “Greedy” AspectRespect Reservations:nScheduling D would not violate the CPU reservations of any of the domains.Minimize Delays:
27、nScheduling D will help reduce the scheduling-induced delay for the most packets.34Algorithm: Bookkeeping at Domain0Step 1:nEvery time a packet is received by the Network InterfacenCard, the network interrupt handler inside the Hypervisor increments the network intensity of Domain0,nDomain0-network-
28、reception-intensity+.Step 2:nUpon deschedule of a guest domain, the Hypervisor increments the network intensity of Domain0,nDomain0-network-transmit-intensity += TransmittingDomain-actual-network-transmit-intensity.35Communication-aware SchedulerPicks the domain with the highest network intensity wh
29、ich is the sum ofnnetwork-reception-intensity and anticipated-networktransmit-intensity for guest domainsnnetwork-reception-intensity and network-transmit-intensity for Domain036Algorithm: Communication-aware SchedulerStep1nThe Hypervisor scheduler computes the network intensity of Domain0 as,wDomai
30、n0-network-reception-intensity + Domain0-network-transmit-intensitynand for the guest domains as,wGuestDomain-network-reception-intensity + GuestDomain-anticipated-network-transmit-intensityStep2nAmong the domains eligible to run within the current period (determined by SEDF fairness limits), the sc
31、heduler chooses to schedule the domain with the highest network intensity (with most number of pending network packets).37Salient FeaturesFairness issuesnGuarantee that each domain receives at least ( slice, period)nConcerned with accounting of the CPU costsCoordinated scheduling38Accounting Of I/O
32、Virtualization OverheadsProportional-Share (PS)nClearly result in a shrinkage of the CPU capacity available to an individual guest domainReservation-Based schedulersnLike SEDF39Accounting MechanismEach guest domain to specify its CPU reservation including the virtualization costs of its I/O needsKee
33、p an accurate account of the time spent by Domain0 on behalf of each of the guest domainsI/O into four kinds:nnetwork receptions/transmissions and disk reads/writes40Implementation41Experimental SetupA rack of 36 serversnDual Xeon 3.4 GHz CPUs with 2 MB of L1 cache, 800 MHz Front Side Bus, and 2 GB
34、RAMnGigabit EthernetnBetween 8 and 12 domainsBetween 120 and 300 MB for each domain Domain 0 is given 320 MB of RAM42Server ApplicationsTPC-W-NYUnThree-tiered based on TPC-W benchmarkScalar pentadiagonal solver (SP) benchmark and Lower-upper diagonal solver (LU) benchmark from NAS Parallel benchmark
35、Streaming media serverJBoss as middle tier MySql for database tier43Performance improvement for TPC-W44Performance Improvement for NAS Parallel BenchmarkswLU benchmark is in much smaller packets (only 40 bytes) and may have to wait for additional packets. The scheduler results in an increased number
36、 of context switches.45Performance and scalability improvement for streaming media server46Fairness in CPU allocation for CPU-intensive domains consolidated with the streaming media server47Evaluation of AccountingTwo main problems with the existing mechanism in Xen:nNonreconciling the time that Dom
37、ain0 spends on behalf of each hosted domain can cause a domain without significant I/O activity to unfairly suffernThe existing mechanism requires the administrator to explicitly specify the required reservation for Domain048&s!pYmUjRfOcL9H6E3B+y(u%r#oWlTiQeNbJ8G5D1A-x*t$qYnVkSgPdMaI7F3C0z)v&s!pXmUi
38、RfOcK9H6E2B+x(u%rZoWlThQeNbJ8G4D1A-w*t$qYnVjSgPdLaI7F3C0y)v&s#pXmUiRfNcK9H5E2B+x(u$rZoWkThQeMbJ7G4D1z-w*t!qYmVjSgOdLaI6F3C0y)v%s#pXlUiRfNcK8H5E2A+x(u$rZnWkThPeMbJ7G4C1z-w&t!qYmVjRgOdL9I6F3B0y(v%s#oXlUiQfNbK8H5D2A+x*u$qZnWkShPeMaJ7G4C1z)w&t!pYmVjRgOcL9I6E3B0y(v%r#oXlTiQfNbK8G5D2A-x*u$qZnVkShPdMaJ7F4C
39、0z)w&s!pYmUjRfOcL9H6E3B+y(v%r#oWlTiQeNbK8G5D1A-x*t$qZnVkSgPdMaI7F4C0z)v&s!pXmUjRfOcK9H6E2B+y(u%rZoWlThQeNbJ8G4D1A-w*t$qYnVjSgPdLaI7F3C0z)v&s#pXmUiRfOcK9H5E2B+x(u%rZoWkThQeMbJ8G4D1z-w*t!qYnVjSgOdLaI6F3C0y)v%s#pXlUiRfNcK8H5E2A+x(u$rZoWkThPeMbJ7G4D1z-w&t!qYmVjSgOdL9I6F3B0y)v%s#oXlUiQfNcK8H5D2A+x*u$rZnW
40、kShPeMaJ7G4C1z)w&t!pYmVjRgOdL9I6E3B0y(v%s#oXlTiQfNbK8H5D2A-x*u$qZnWkShPdMaJ7F4C1z)w&s!pYmUjRgOcL9H6E3B+y(v%r#oWlTiQeNbK8G5D1A-x*t$qZnVkShPdMaI7F4C0z)w&s!pXmUjRfOcL9H6E2B+y(u%r#oWlThQeNbJ8G5D1A-w*t$qYnVkSgPdLaI7F3C0v%r#oXlTiQeNbK8G5D2A-x*t$qZnVkShPdMaI7F4C0z)w&s!pXmUjRfOcL9H6E2B+y(u%r#oWlThQeNbJ8G5D1
41、A-w*t$qYnVkSgPdLaI7F3C0z)v&s!pXmUiRfOcK9H6E2B+x(u%rZoWlThQeMbJ8G4D1A-w*t!qYnVjSgPdLaI6F3C0y)v&s#pXlUiRfNcK9H5E2A+x(u$rZoWkThQeMbJ7G4D1z-w*t!qYmVjSgOdLaI6F3B0y)v%s#pXlUiQfNcK8H5E2A+x*u$rZnWkThPeMaJ7G4C1z-w&t!pYmVjRgOdL9I6F3B0y(v%s#oXlUiQfNbK8H5D2A+x*u$qZnWkShPeMaJ7F4C1z)w&t!pYmUjRgOcL9I6E3B+y(v%r#oXl
42、TiQeNbK8G5D2A-x*t$qZnVkShPdMaJ7F4C0z)w&s!pYmUjRfOcL9H6E3B+y(u%r#oWlTiQeNbJ8G5D1A-x*t$qYnVkSgPdMaI7F3C0z)v&s!pXmUiRfOcK9H6E2B+y(u%rZoWlThQeNbJ8G4D1A-w*t$qYnVjSgPdLaI7F3C0y)v&s#pXmUiRfNcK9H5E2B+x(u$rZoWkThQeMbJ7G4D1z-w*t!qYmVjSgOdLaI6F3C0y)v%s#pXlUiRfNcK8H5E2A+x(u$rZnWkThPeMbJ7G4C1z-w&t!qYmVjRgOdL9I6F
43、3B0y(v%s#oXlUiQfNbK8H5D2A+x*u$rZnWkShPeMaJ7G4C1z)w&t!pYmVjRgOcL9I6E3B0y(v%r#oXlTiQfNbK8G5D2A-x*u$qZnVkShPdMaJ7F4C0z)w&s!pYmUjRgOcL9H6E3B+y(v%r#oWlTiQeNbK8G5D1A-x*t$qZnVkSgPdMaI7F4C0z)v&s!pXmUjRfOcK9H6E2B+y(u%rZoWlThQeNbJ8G4D1A-w*t$qYnVkSgPdLaI7F3C0z)v&s#pXmUiRfOcK9H5E2B+x(u%rZoWkThQeMbJ8G4D1z-w*tmUj
44、RfOcK9H6E2B+y(u%r#oWlThQeNbJ8G5D1A-w*t$qYnVkSgPdLaI7F3C0z)v&s#pXmUiRfOcK9H5E2B+x(u%rZoWkThQeMbJ8G4D1z-w*t!qYnVjSgOdLaI6F3C0y)v&s#pXlUiRfNcK9H5E2A+x(u$rZoWkThPeMbJ7G4D1z-w&t!qYmVjSgOdL9I6F3B0y)v%s#oXlUiQfNcK8H5D2A+x*u$rZnWkThPeMaJ7G4C1z-w&t!pYmVjRgOdL9I6E3B0y(v%s#oXlTiQfNbK8H5D2A-x*u$qZnWkShPdMaJ7F4C
45、1z)w&s!pYmUjRgOcL9I6E3B+y(v%r#oXlTiQeNbK8G5D2A-x*t$qZnVkShPdMaI7F4C0z)w&s!pXmUjRfOcL9H6E2B+y(u%r#oWlThQeNbJ8G5D1A-w*t$qYnVkSgPdMaI7F3C0z)v&s!pXmUiRfOcK9H6E2B+x(u%rZoWlThQeMbJ8G4D1A-w*t!qYnVjSgPdLaI6F3C0y)v&s#pXlUiRfNcK9H5E2B+x(u$rZoWkThQeMbJ7G4D1z-w*t!qYmVjSgOdLaI6F3B0y)v%s#pXlUiQfNcK8H5E2A+x*u$rZnW
46、kThPeMaJ7G4C1z-w&t!pYmVjRgOdH5E2B+x(u$rZoWkThQeMbJ7G4D1z-w*t!qYmVjSgOdLaI6F3B0y)v%s#pXlUiQfNcK8H5E2A+x*u$rZnWkThPeMbJ7G4C1z-w&t!qYmVjRgOdL9I6F3B0y(v%s#oXlUiQfNbK8H5D2A+x*u$qZnWkShPeMaJ7F4C1z)w&t!pYmUjRgOcL9I6E3B0y(v%r#oXlTiQfNbK8G5D2A-x*u$qZnVkShPdMaJ7F4C0z)w&s!pYmUjRfOcL9H6E3B+y(u%r#oWlTiQeNbJ8G5D1
47、A-x*t$qYnVkSgPdMaI7F4C0z)v&s!pXmUjRfOcK9H6E2B+y(u%rZoWlThQeNbJ8G4D1A-w*t$qYnVjSgPdLaI7F3C0y)v&s#pXmUiRfNcK9H5E2B+x(u%rZoWkThQeMbJ8G4D1z-w*t!qYnVjSgOdLaI6F3C0y)v%s#pXlUiRfNcK8H5E2A+x(u$rZnWkThPeMbJ7G4C1z-w&t!qYmVjRgOdL9I6F3B0y)v%s#oXlUiQfNcK8H5Dw*t!qYnVjSgOdLaI6F3C0y)v%s#pXlUiRfNcK8H5E2A+x(u$rZnWkThP
48、eMbJ7G4D1z-w&t!qYmVjSgOdL9I6F3B0y)v%s#oXlUiQfNcK8H5D2A+x*u$rZnWkShPeMaJ7G4C1z)w&t!pYmVjRgOcL9I6E3B0y(v%s#oXlTiQfNbK8H5D2A-x*u$qZnWkShPdMaJ7F4C1z)w&s!pYmUjRgOcL9H6E3B+y(v%r#oWlTiQeNbK8G5D1A-x*t$qZnVkSgPdMaI7F4C0z)w&s!pXmUjRfOcL9H6E2B+y(u%r#oWlThQeNbJ8G5D1A-w*t$qYnVkSgPdLaI7F3C0z)v&s#pXmUiRfOcK9H5E2B+
49、x(u%rZoWlThQeMbJ8G4D1A-w*t!qYnVjSgL9H6E2B+y(u%r#oWlThQeNbJ8G5D1A-w*t$qYnVkSgPdLaI7F3C0z)v&s#pXmUiRfOcK9H6E2B+x(u%rZoWlThQeMbJ8G4D1A-w*t!qYnVjSgPdLaI6F3C0y)v&s#pXlUiRfNcK9H5E2A+x(u$rZoWkThPeMbJ7G4D1z-w*t!qYmVjSgOdLaI6F3B0y)v%s#pXlUiQfNcK8H5E2A+x*u$rZnWkThPeMaJ7G4C1z-w&t!pYmVjRgOdL9I6E3B0y(v%s#oXlTiQf
50、NbK8H5D2A+x*u$qZnWkShPeMaJ7F4C1z)w&t!pYmUjRgOcL9I6E3B+y(v%r#oXlTiQeNbK8G5D2A-x*t$qZnVkShPdMaI7F4C0z)w&s!pYmUjRfOcL9H6E3B+y(u%r#oWlTiQeNbJ8G5D1A-x*t$qYnVkSgPdMaI7F3C0z)v&s!pXmUiRfOcK9H6E2B+x(u%nVkShPdMaJ7F4C0z)w&s!pYmUjRfOcL9H6E3B+y(u%r#oWlTiQeNbJ8G5D1A-x*t$qYnVkSgPdMaI7F3C0z)v&s!pXmUiRfOcK9H6E2B+y(u
51、%rZoWlThQeNbJ8G4D1A-w*t$qYnVjSgPdLaI7F3C0y)v&s#pXmUiRfNcK9H5E2B+x(u$rZoWkThQeMbJ7G4D1z-w*t!qYnVjSgOdLaI6F3C0y)v%s#pXlUiRfNcK8H5E2A+x(u$rZnWkThPeMbJ7G4C1z-w&t!qYmVjRgOdL9I6F3B0y(v%s#oXlUiQfNbK8H5D2A+x*u$rZnWkShPeMaJ7G4C1z)w&t!pYmVjRgOcL9I6E3B0y(v%r#oXlTiQfNbK8G5D2A-x*u$qZngOdL9I6F3B0y)v%s#oXlUiQfNcK8
52、H5D2A+x*u$rZnWkShPeMaJ7G4C1z)w&t!pYmVjRgOcL9I6E3B0y(v%r#oXlTiQfNbK8G5D2A-x*u$qZnWkShPdMaJ7F4C1z)w&s!pYmUjRgOcL9H6E3B+y(v%r#oWlTiQeNbK8G5D1A-x*t$qZnVkSgPdMaI7F4C0z)v&s!pXmUjRfOcK9H6E2B+y(u%r#oWlThQeNbJ8G5D1A-w*t$qYnVkSgPdLaI7F3C0z)v&s#pXmUiRfOcK9H5E2B+x(u%rZoWkThQeMbF4C0z)w&s!pXmUjRfOcL9H6E2B+y(u%r#o
53、WlThQeNbJ8G5D1A-w*t$qYnVkSgPdLaI7F3C0z)v&s#pXmUiRfOcK9H5E2B+x(u%rZoWkThQeMbJ8G4D1A-w*t!qYnVjSgPdLaI6F3C0y)v&s#pXlUiRfNcK9H5E2A+x(u$rZoWkThPeMbJ7G4D1z-w&t!qYmVjSgOdL9I6F3B0y)v%s#pXlUiQfNcK8H5E2A+x*u$rZnWkThPeMaJ7G4C1z-w&t!pYmVjRgOdL9I6E3B0y(v%s#oXlTiQfNbK8H5D2A-x*u$qZnWkShPdMaJ7F4C1z)w&t!pYmUjRgOcL9I6E3B+y(v%r#oXlTiQeNbK8G5D2A-x*t$qZnVkShPdMaI7F4C0z)w&s!pXmUjRfOcL9H6E2B+y(u%WkShPeMaJ7F4C1z)w&t!pYmUjRgOcL9I6E3B+y(v%r#oXlTiQ49
