11º Informativo Jornal o Verbo - CCMA Ministério Alfenas - Setembro - 2012
Copyright 2010 ITRI 工業技術研究院 11 ITRI Cloud OS & Virtual Resource Management Patrick Fu...
-
Upload
jarvis-bennet -
Category
Documents
-
view
221 -
download
1
Transcript of Copyright 2010 ITRI 工業技術研究院 11 ITRI Cloud OS & Virtual Resource Management Patrick Fu...
Copyright 2010 ITRI 工業技術研究院 11
ITRI Cloud OS & Virtual Resource Management
Patrick FuSystem Software Division, CCMA/[email protected]
Copyright 2010 ITRI 工業技術研究院 2
Agenda
• Cloud OS introduction• Physical resource provisioning• Virtual resource management• Adaptive provisioning and power management
Copyright 2010 ITRI 工業技術研究院 3
What is Cloud OS ?
Physical Node
Physical Node
Storage
Server
Storage
Server
Physical Node
Physical Node
Storage Server
Storage Server
Mail Virtual Cluster
Compute Nodes
Backup Virtual Cluster
HC Virtual Cluster
AppX Virtual Cluster
Data NodesService Nodes
System
Service daemons
System
Service daemons
Cloud OS agents
Cloud OS agents
• System Management Software layer– Physical Resource Provisioning– Virtual Resource Management
• Improve manageability of massive Cloud Data Center
• Enhance self-provisioning• Optimize physical resource utilization• High Availability for any single point
of failure• Energy management
– Highly Available Distributed Storage Management
– Service Load Balancing– Security– High Speed Networking
• What is it not?– It’s not Operating System– It’s not Virtualization Hypervisor
Copyright 2010 ITRI 工業技術研究院 4
Service/Technology Mapping
IaaS
PaaS
Servers Storage Arrays Power DistributionSwitches
+Scalable System Architecture System Management Cooling
Cloud Hardware Platform
Hypervisor Virtualization Mgmt Storage Mgmt Security
Backup/Replication Data Center Automation Energy Management
Cloud System Software Platform
LAMP .NET WebSphere WebLogic Google App Engine
Cloud Application Middleware Platform
SaaS Automated Cloudification TechnologyApplications
Copyright 2010 ITRI 工業技術研究院 5
Software Stack for Cloud OS
Physical ClusterDeploymentTool
Physical ClusterDeploymentTool
Virtual Machine ManagementVirtual Machine Management
Virtual Cluster ProvisioningVirtual Cluster Provisioning
PowerManagement
PowerManagement
Intra-Virtual-ClusterLoad Balancing
Intra-Virtual-ClusterLoad Balancing
System/NetworkManagement
System/NetworkManagement
SecuritySecurity
Virtual DataCenter Mgmt ConsoleVirtual DataCenter Mgmt Console
Physical Compute ServersPhysical Compute Servers
All-layer-2 Network All-layer-2 Network Distributed Main/Secondary StorageDistributed Main/Secondary Storage
Copyright 2010 ITRI 工業技術研究院 6
Cloud OS Service Model
• Provisioning & Runtime monitoring of Virtual Resources– Virtual Instance
Hypervisor construct An image of a guest OS
– Virtual Cluster A group of VM instances providing same service, front-ended by a network
load balancer Configuration
- # of virtual machines and its configuration- Storage space requirement- External network bandwidth requirement- Load balancing policy- Firewall/IDS setting- Network configuration, including DNS and DHCP- OS image and application image
– Virtual Data Center One or more virtual cluster working in coordination (multi-tier web services,
EMR’s, VDI’s, etc)
Copyright 2010 ITRI 工業技術研究院 7
CloudOS Virtualization Level
… PM
OS
AP
s
vm
OS
AP
s
vm
OS
AP
s
vm
OS
AP
s
vm
…
PM
OS
AP
s
vmO
SA
Ps
vmO
SA
Psvm
OS
AP
s
vm
…
PM
OS
AP
s
vm
OS
AP
s
vm
OS
AP
s
vm
OS
AP
s
vm
…
PM
OS
AP
s
vm
OS
AP
s
vm
OS
AP
s
vm
OS
AP
s
vm
…
VCluster VCluster VCluster
VDC VDC
CloudOS
Copyright 2010 ITRI 工業技術研究院 8
Resource Provisioning
• To prepare VMs with appropriate resources and make them ready for user applications– Allocating resources to VMs to match the workloads
• To prepare a virtual cluster with appropriate instances and make it ready for virtual cluster computation– Consolidating VMs onto physical servers
• Goals: – High resource utilization– Energy efficiency– Low performance interference
Copyright 2010 ITRI 工業技術研究院 9
Provisioning Challenges• VM size estimation
– Static SLA model/forecasting future use• Placement
– Deploy a VM onto physical servers (initially)– Policy: immediate, best effort, advance reservation, etc.
• Consolidation and Load Balancing– High consolidation ration and resource utilization – low cost of running data center– Statically
• Heuristic based• Average resource utilization
– Dynamic replacement• Measure-Forecast-Remap (MFR)• Live migration• Balancing overloaded and underloaded nodes
– Constrained bin packing problem w/ SLA• Performance isolation
– Cohosting VMs on a server creates performance interference– How to model and prevent the interference
Copyright 2010 ITRI 工業技術研究院 10
RPM Static Resource Provisioning
• Statically provision from SLA• SLA w/ historical data?
– No, conservatively allocation– Yes, forecasting joint-VM provisioning
• Immediate provisioning model (before instantiation of virtual machine)
• Placement policy– Proprietary – Virtual cluster affinity placement policy
• Performance gain from locality • Place VMs from the same virtual cluster as possible• Need experiments to support
CloudOS
Copyright 2010 ITRI 工業技術研究院 11
Our Motivation
Data Center Monthly Cost
54%
8%
21%
13% 5%
Servers
NetworkingEquipment
Power Distribution &Cooling
Power
Other Infrastructure
Source: Cost of power in Large-Scale Data Center, James Hamilton Blog, 11/28/2008
Copyright 2010 ITRI 工業技術研究院 12
Joint Provisioning via VM Multiplexing
• Dataset from a commercial data center– 15,897 VMs– 1325 physical hosts
• 94% of the hosts have more than one VM
• Joint provisioning averagely saves 40% of the capacity
Meng, X. et al. Efficient Resource Provisioning in Compute Clouds via VM Multiplexing. ICAC ‘10
Copyright 2010 ITRI 工業技術研究院 13
Joint-VM Provisioning at Runtime
Hyp
ervisor
PM
VM
Hyp
ervisor
PM
VM
100%100%
Under provisioning Over provisioning
CloudOS
CapacityCapacity
tt
tt
CapacityCapacity
Copyright 2010 ITRI 工業技術研究院 14
Load balancing and DVMM
Consolidation manager/ DVMM
PM PM PM PM PM PM
Over provisioningUnder provisioning
Joint-VM histogramVM victim histogram
Resource Provisioning Manager (RPM)
PM reconfiguration
Utilization ratio
Reach reconfiguration point
CloudOS
Copyright 2010 ITRI 工業技術研究院 15
Adaptive physical resource provisioning
PRM
Power on/off PMs
Reconfiguration map
New PM map
Utilization rate reaches threshold, sending reallocation request
Static joint-VM provisioning
DVMM
Consolidation manager
Placement
VM monitoring
Runtime joint-VM
provisioning
Performance interferenceUtilization changeVictims
RPM core
New PM map
Cloud OS RPM Software Components
Cloud OS
Copyright 2010 ITRI 工業技術研究院 16
Load balancing
Copyright 2010 ITRI 工業技術研究院 17
Consolidation plan
Copyright 2010 ITRI 工業技術研究院 18
Migration plan
Copyright 2010 ITRI 工業技術研究院 19
Runtime Reallocation
VM3
VM2
VM1
PM1
VM3
VM2
VM1
PM2
VM3
VM2
VM1
PM3
VM1
PMi
VM1
PMj
VM1
PMk
VM3
VM1
PM1
VM3
VM1
PM2
VM3VM2
VM1
PM3
VM1
PMi
VM1
PMj
VM1
PMk
VM2
VM2 VM2
Copyright 2010 ITRI 工業技術研究院 20
Adaptive Physical Resource Provisioning
Power ManagementPM Pool
Provisioned
Utilization threshold
Low utilization High utilization
Over provisioning Under provisioning
PM reallocation algorithm
PRMPower on/off PMs
Reconfiguration mapVM placement
DVMM live migrationLoad balancer
Consolidation manager Utilization monitor New PM map
Utilization rate reaches threshold, sending reallocation request
CloudOS
Copyright 2010 ITRI 工業技術研究院 21
Challenges• Triggering mechanism
– No workload consolidation “recently” (e.g. past hour)– No physical machine load balancing going on– No physical server was powered on “recently” (e.g. past hour)– Avoid oscillation
• Cost of migration– Network load– Cache effects– Domain in suspension
• Multi-dimensional bin packing– CPU, Memory, Network, Disk I/O
• Migration plan– Only 1 migration per Physical server @ a time– # of cores vs. # of VMs
Copyright 2010 ITRI 工業技術研究院 22
Backup
Copyright 2010 ITRI 工業技術研究院 23
Software architecture
Copyright 2010 ITRI 工業技術研究院 24
Procedure of Power management in Monitoring thread
Receive data from Dom0Calculate the data
WorkloadTrigger?
Yes
Perform Consolidation Plan
M*K < N ?
No
No
Receiving CCinstance data per 30 secCalculateinstMonitorThreadData->cputotalinstMonitorThreadData->memorytotalinstMonitorThreadData->count
Do Consolidation
Change instance stateinstDvmmBloc->state=doingpwminstMonitorThreadData->state=doingpwminstDvmmBloc->destHost=resource->hostNameChange external state
Call PRM to shut down Machine
Change instance stateinstDvmmBloc->pwmtime=nowinstDvmmBloc->state=pwmdoneinstMonitorThreadData->state=pwmdoneinstMonitorThreadData->pwmtime=now
Stop receiving data from Dom0
Check instMonitorThreadData->stateinstMonitorThreadData->pwmtimeinstMonitorThreadData->lbtimeinstMonitorThreadData->cputotalinstMonitorThreadData->memorytotalinstMonitorThreadData->count
Done
Yes
Do Load balancing
Change instance stateinstDvmmBloc->state=doinglbinstMonitorThreadData->state=doinglbinstDvmmBloc->destHost=resource->hostNameChange external state
Call PRM to turn on Machine
If necessary
Copyright 2010 ITRI 工業技術研究院 25
Data Mining:VM Resource Usage Patterns of each VC
• Find VM resource usage patterns for each VC
• Aid to predict the trend of resource usages
Medium
LLow
High (or unpredictable)
Time
CPU usage Monday
Copyright 2010 ITRI 工業技術研究院 26
Q&AThank you!