SlideShare una empresa de Scribd logo
1 de 30
RTLWS 2011
               October 21st, Prague




Improving Responsiveness for Virtualized
 Networking Under Intensive Computing
              Workloads


    Tommaso Cucinotta, Fabio Checconi, Dhaval Giani
            Cucinotta,
           Real-Time Systems Lab (RETIS)
        Scuola Superiore Sant'Anna, Pisa (Italy)
Context




Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy   2
What is Real-Time
    Drive assistance
       Engine control, brakes, stability,
        speed, parking
       Trajectory and set-up control
    Defence, army, space




Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore 3
What is Real-Time
    Control of chemical and nuclear plants
    Control of productive processes and
    industrial automation
    Traffic control




Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore 4
What is Real-Time
    Multimedia, videosurveillance
    Augmented virtual reality
    Telecommunications
    Environment monitoring




Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore 5
Our Focus
    We focus on systems
      With non-critical soft real-time requirements
      Where the use of a GPOS is desirable and feasible
          • As opposed to a traditional RTOS
      Where virtualization is desirable and feasible
    Application scenarios
      Multimedia
      Embedded systems
      QoS-enabled Cloud Computing
      Web servers with QoS assurance



Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore
Motivations




Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy   7
Virtualization

Virtualization is a key technology
  For IaaS providers (Cloud Computing)
  Enables server consolidation



    Physical Host
    Physical Host
          OS
          OS
          ...

    Physical Host
    Physical Host
          OS
          OS
    Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy   8
Virtualization

Virtualization is a key technology
  For IaaS providers (Cloud Computing)
  Enables server consolidation



    Physical Host
    Physical Host                                  Physical Host
                                                   Physical Host
          OS
          OS                                              VM/OS
                                                          VM/OS
          ...                                            ...

    Physical Host                                         VM/OS
                                                          VM/OS
    Physical Host
          OS
          OS
    Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy   9
Need for Performance Isolation

Resource sharing
   → Temporal interference

        Physical Host
        Physical Host
         VM
         VM

         VM
         VM

         VM
         VM

         VM
         VM

    Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy   10
Need for Performance Isolation

Resource sharing                                   ~30ms
                                                              VM Alone

   → Temporal interference

        Physical Host
        Physical Host
         VM
         VM

         VM
         VM

         VM
         VM

         VM
         VM

    Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy   11
Need for Performance Isolation

Resource sharing                                    ~30ms
                                                               VM Alone

   → Temporal interference

        Physical Host
        Physical Host
         VM
         VM
                                          τ1 = (30ms, 150ms)
                                          τ2 = (50ms, 200ms)
         VM
         VM
                                                                    ~120ms
                                                                 2 VMs

         VM
         VM

         VM
         VM

    Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy   12
Possible Solutions

Hardware replication and                                Physical Host
                                                        Physical Host
static partitioning                                      VM
                                                         VM
  Computing
    • Multi-core (1 core per VM)
    • NUMA awareness                                     VM
                                                         VM
  Networking
    • Multiple network adapters                          VM
                                                         VM
      (1 network adapter per VM)
    • Multi-queue adapters
                                                         VM
                                                         VM
Drawbacks
  Limitation of flexibility
  Under-utilization of resources
    • e.g., with real-time/interactive workloads (cloud computing)
    Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy   13
Possible Solutions

Another approach
  Let multiple VMs use the same resources
  Use proper resource scheduling strategies
For example
  Computing
     • Xen credit-based and SEDF schedulers
  Networking
     • QoS-aware protocols (IntServ, MPLS, WF2Q+)
Advantages
  Increased flexibility
  Increased resource saturation levels
  Reduced infrastructure costs
    Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy   14
Background on
         Deadline-based Scheduling




Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy   15
Real-Time Schedulers for
                         Linux
    SCHED_SPORADIC
      Additional scheduling policy within sched_rt.c
      Single-threaded applications
      API based on system call: sched_setscheduler()
    SCHED_DEADLINE
      Independent scheduling class
      Single-threaded control applications
      API based on system call: sched_setscheduler()
    IRMOS (a.k.a., EDF Throttling)
      Rewrite of RT throttling in sched_rt.c: from limits to guarantees
      Multi-threaded complex software components
          • KVM, JVM, Apache+Tomcat, …
      API based on cgroups interface
Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore
IRMOS Scheduler

          Slice the available computing power into
          reservations




  CPU 0
  CPU 1
  CPU 2
  CPU 3

Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore
Hierarchical Scheduling

                                                                               EDF
                                                                               EDF
                                          Max RT Prio
                                                …
                                                                     FP
                                                                     FP                   FP
                                                                                          FP


                                          Min RT Prio           T1
                                                                T1        T2
                                                                          T2         T3
                                                                                     T3        T4
                                                                                               T4

    Needed operations
       create & destroy reservations
       attach & detach tasks ↔ reservations
       list tasks attached to reservations (and list reservations)
       Standard operations: get & set parameters

Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore 18
Background
    Scheduling KVM VMs with
    resource reservations
    (deadline-based scheduling)
       VM1 granted Q1 time units every P1
       VM2 granted Q2 time units every P2
       VM3 granted Q3 time units every P3
      . . .


                                                      Networking Threads

                           KVM

                                                      Computing Thread

Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore 19
Problem Presentation




Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy   20
Problem Presentation
       Focus on compute-intensive VM
       Schedule with VM alone
          Unresponsiveness frame: P-Q


       Well-known worst-case schedule
          Unresponsiveness frame: 2(P-Q)




Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore 21
Problem Presentation
    Packet arrival in the unresponsiveness frame
       Its handling deferred till next activation period
    Mixed workload inside VM
                                                                   packet
       Batch compute-intensive workload
       Responsive application(s)
           • e.g., triggered by arriving packet
    Compute-intensive workload
       Period as large as possible
           • For minimizing overheads
    Responsive workload
       Period tuned on interactivity requirements
           • VM unresponsive for a time-frame as high as 2(P-Q)

Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore 22
Proposed Approach
    On packet arrival with null budget
       Attach the VM to a temporary reservation
           • With period Ps << P
           • With budget Qs sufficient to “respond” to the request
                – Including both kernel-level and user-space level operations
       Temporary reservation lasts
           • till budget exhaustion or original budget replenishment
    Worst-case latency (assump.
    of sporadic requests)
       From 2(P-Q) to 2(Ps-Qs)
    Bandwidth occupation
       From            to

Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore 23
Experimental Results
Experimental set-up
  Periodic ping every 1s
  Normal reservation: (40ms, 100ms)
  Spare reservation: (4ms, 10ms)
Batch activity
  Fake loop logging experienced loops/s




Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore 24
Achieved results
       Reservation (40,100) and no spare reservation
           • Response-time ~ 120ms (top curve)
           • Throughput ~ 1,11 cycles/us
       Reservation (40,100) with spare reservation (4,10)
           • Response-time ~ 16ms (bottom curve)
       Reservation (4,10)
           • Throughput ~ 0,56 cycles/us




Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore 25
Implementation Details




Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy   26
Modifications to the
                           Scheduler
    Proper flags into the cgroup filesystem to tag
       Spare bandwidth provider groups
       Spare bandwidth client groups
    Modifications to the networking receive path
       if packet goes to VM within client group
        (and budget is null),
       then attach VM to (any matching) provider group
    Spare bandwidth clients/providers matching
       Based on numerical identifier (clients can attach only to
        servers with the same identifier)
       Allows for realizing spare bandwidth pools


Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore 27
Future Work
    More extensive evaluation
       Presented results are very preliminary
       Need to test with user-space application, not only ping
    Servers pooling
       All VMs need the spare reservation at the same time
           • With a very low probability
           • So, leverage statistical multiplexing and use #providers < #clients
           • Lowers spare bandwidth requirements
                                                                              VM
                                                                              VM      VM
                                                                                      VM

                                                                        VM
                                                                        VM       S     S    VM
                                                                                            VM
                                                                                 S     S

                                                                              VM
                                                                              VM        VM
                                                                                        VM


Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore 28
Related Publications

      Hierarchical Multiprocessor CPU Reservations for the Linux Kernel
       F. Checconi, T. Cucinotta, D. Faggioli, G. Lipari
       OSPERT 2009, Dublin, Ireland, June 2009
      An EDF Scheduling class for the Linux kernel
       D. Faggioli, F. Checconi, M. Trimarchi, C. Scordino
       RTLWS 2009, Dresden, October 2009
      Self-tuning Schedulers for Legacy Real-Time Applications
       T. Cucinotta, F. Checconi, L. Abeni, L. Palopoli
       EuroSys 2010, Paris, April 2010
      Respecting temporal constraints in virtualised services
       T. Cucinotta, G. Anastasi, L. Abeni
       RTSOAA 2009, Seattle, Washington, July 2009




Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore
Thanks for your attention




                                   Questions ?


      http://retis.sssup.it/people/tommaso


Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna
                                                                        © 2007 Scuola Superiore 30

Más contenido relacionado

La actualidad más candente

Xo Rskillsmatrix Oct09
Xo Rskillsmatrix Oct09Xo Rskillsmatrix Oct09
Xo Rskillsmatrix Oct09martinfox
 
Hyper-V VMM ile Cloud computing
Hyper-V VMM ile Cloud computingHyper-V VMM ile Cloud computing
Hyper-V VMM ile Cloud computingAhmet Mutlu
 
Semper Continuity Suite Product Profile V6
Semper Continuity Suite   Product Profile V6Semper Continuity Suite   Product Profile V6
Semper Continuity Suite Product Profile V6mlulham
 
Atea roadshow norr
Atea roadshow norrAtea roadshow norr
Atea roadshow norrJohan Odell
 
CTI Group- Blue power technology storwize technical training for customer - p...
CTI Group- Blue power technology storwize technical training for customer - p...CTI Group- Blue power technology storwize technical training for customer - p...
CTI Group- Blue power technology storwize technical training for customer - p...Tri Susilo
 
Task-aware Virtual Machine Scheduling for I/O Performance
Task-aware Virtual Machine Scheduling for I/O PerformanceTask-aware Virtual Machine Scheduling for I/O Performance
Task-aware Virtual Machine Scheduling for I/O PerformanceHwanju Kim
 
Cooperative VM Migration for a virtualized HPC Cluster with VMM-bypass I/O de...
Cooperative VM Migration for a virtualized HPC Cluster with VMM-bypass I/O de...Cooperative VM Migration for a virtualized HPC Cluster with VMM-bypass I/O de...
Cooperative VM Migration for a virtualized HPC Cluster with VMM-bypass I/O de...Ryousei Takano
 
30ian2009 fujitsu
30ian2009 fujitsu30ian2009 fujitsu
30ian2009 fujitsuAgora Group
 
TSM 6.4 Technical updates
TSM 6.4 Technical updates TSM 6.4 Technical updates
TSM 6.4 Technical updates Solv AS
 
Introdution - Fujitsu PRIMEQUEST
Introdution - Fujitsu PRIMEQUESTIntrodution - Fujitsu PRIMEQUEST
Introdution - Fujitsu PRIMEQUESTAndrew Wong
 
IT FUTURE 2011 - Fujitsu ror orchestration
IT FUTURE 2011 - Fujitsu ror orchestrationIT FUTURE 2011 - Fujitsu ror orchestration
IT FUTURE 2011 - Fujitsu ror orchestrationFujitsu France
 
MOW2010: Under the Hood of Oracle Clusterware by Alex Gorbachev, Pythian
MOW2010: Under the Hood of Oracle Clusterware by Alex Gorbachev, PythianMOW2010: Under the Hood of Oracle Clusterware by Alex Gorbachev, Pythian
MOW2010: Under the Hood of Oracle Clusterware by Alex Gorbachev, PythianAlex Gorbachev
 
vBrownBag OpenStack Networking Talk
vBrownBag OpenStack Networking TalkvBrownBag OpenStack Networking Talk
vBrownBag OpenStack Networking Talkmestery
 
How Quantum configures Virtual Networks under the Hood?
How Quantum configures Virtual Networks under the Hood?How Quantum configures Virtual Networks under the Hood?
How Quantum configures Virtual Networks under the Hood?Etsuji Nakai
 
Il Cloud chiavi in mano | Marco Soldi (Intel) | Milano
Il Cloud chiavi in mano | Marco Soldi (Intel) | MilanoIl Cloud chiavi in mano | Marco Soldi (Intel) | Milano
Il Cloud chiavi in mano | Marco Soldi (Intel) | MilanoCA Technologies Italia
 
Windows Azure Interoperability
Windows Azure InteroperabilityWindows Azure Interoperability
Windows Azure InteroperabilityMihai Dan Nadas
 

La actualidad más candente (19)

Xo Rskillsmatrix Oct09
Xo Rskillsmatrix Oct09Xo Rskillsmatrix Oct09
Xo Rskillsmatrix Oct09
 
Hyper-V VMM ile Cloud computing
Hyper-V VMM ile Cloud computingHyper-V VMM ile Cloud computing
Hyper-V VMM ile Cloud computing
 
Semper Continuity Suite Product Profile V6
Semper Continuity Suite   Product Profile V6Semper Continuity Suite   Product Profile V6
Semper Continuity Suite Product Profile V6
 
Atea roadshow norr
Atea roadshow norrAtea roadshow norr
Atea roadshow norr
 
CTI Group- Blue power technology storwize technical training for customer - p...
CTI Group- Blue power technology storwize technical training for customer - p...CTI Group- Blue power technology storwize technical training for customer - p...
CTI Group- Blue power technology storwize technical training for customer - p...
 
OpenStack Quantum
OpenStack QuantumOpenStack Quantum
OpenStack Quantum
 
Task-aware Virtual Machine Scheduling for I/O Performance
Task-aware Virtual Machine Scheduling for I/O PerformanceTask-aware Virtual Machine Scheduling for I/O Performance
Task-aware Virtual Machine Scheduling for I/O Performance
 
Cooperative VM Migration for a virtualized HPC Cluster with VMM-bypass I/O de...
Cooperative VM Migration for a virtualized HPC Cluster with VMM-bypass I/O de...Cooperative VM Migration for a virtualized HPC Cluster with VMM-bypass I/O de...
Cooperative VM Migration for a virtualized HPC Cluster with VMM-bypass I/O de...
 
30ian2009 fujitsu
30ian2009 fujitsu30ian2009 fujitsu
30ian2009 fujitsu
 
Nuxeo 5.2 Glassfish
Nuxeo 5.2 GlassfishNuxeo 5.2 Glassfish
Nuxeo 5.2 Glassfish
 
TSM 6.4 Technical updates
TSM 6.4 Technical updates TSM 6.4 Technical updates
TSM 6.4 Technical updates
 
Introdution - Fujitsu PRIMEQUEST
Introdution - Fujitsu PRIMEQUESTIntrodution - Fujitsu PRIMEQUEST
Introdution - Fujitsu PRIMEQUEST
 
IT FUTURE 2011 - Fujitsu ror orchestration
IT FUTURE 2011 - Fujitsu ror orchestrationIT FUTURE 2011 - Fujitsu ror orchestration
IT FUTURE 2011 - Fujitsu ror orchestration
 
MOW2010: Under the Hood of Oracle Clusterware by Alex Gorbachev, Pythian
MOW2010: Under the Hood of Oracle Clusterware by Alex Gorbachev, PythianMOW2010: Under the Hood of Oracle Clusterware by Alex Gorbachev, Pythian
MOW2010: Under the Hood of Oracle Clusterware by Alex Gorbachev, Pythian
 
vBrownBag OpenStack Networking Talk
vBrownBag OpenStack Networking TalkvBrownBag OpenStack Networking Talk
vBrownBag OpenStack Networking Talk
 
How Quantum configures Virtual Networks under the Hood?
How Quantum configures Virtual Networks under the Hood?How Quantum configures Virtual Networks under the Hood?
How Quantum configures Virtual Networks under the Hood?
 
Energy efficient storage in vm
Energy efficient storage in vmEnergy efficient storage in vm
Energy efficient storage in vm
 
Il Cloud chiavi in mano | Marco Soldi (Intel) | Milano
Il Cloud chiavi in mano | Marco Soldi (Intel) | MilanoIl Cloud chiavi in mano | Marco Soldi (Intel) | Milano
Il Cloud chiavi in mano | Marco Soldi (Intel) | Milano
 
Windows Azure Interoperability
Windows Azure InteroperabilityWindows Azure Interoperability
Windows Azure Interoperability
 

Destacado

Self-tuning Schedulers for Legacy Real-Time Applications
Self-tuning Schedulers for Legacy Real-Time ApplicationsSelf-tuning Schedulers for Legacy Real-Time Applications
Self-tuning Schedulers for Legacy Real-Time Applicationstcucinotta
 
The Wizard of OS: a Heartbeat for Legacy Multimedia Applications
The Wizard of OS: a Heartbeat for Legacy Multimedia ApplicationsThe Wizard of OS: a Heartbeat for Legacy Multimedia Applications
The Wizard of OS: a Heartbeat for Legacy Multimedia Applicationstcucinotta
 
Virtualised e-Learning with Real-Time Guarantees on the IRMOS Platform
Virtualised e-Learning with Real-Time Guarantees on the IRMOS PlatformVirtualised e-Learning with Real-Time Guarantees on the IRMOS Platform
Virtualised e-Learning with Real-Time Guarantees on the IRMOS Platformtcucinotta
 
Low-Latency Audio on Linux by Means of Real-Time Scheduling
Low-Latency Audio on Linux by Means of Real-Time SchedulingLow-Latency Audio on Linux by Means of Real-Time Scheduling
Low-Latency Audio on Linux by Means of Real-Time Schedulingtcucinotta
 
Optimum Scalability Point for Parallelisable Real-Time Components
Optimum Scalability Point for Parallelisable Real-Time ComponentsOptimum Scalability Point for Parallelisable Real-Time Components
Optimum Scalability Point for Parallelisable Real-Time Componentstcucinotta
 
Help Restore Dell Place Ravine
Help Restore Dell Place RavineHelp Restore Dell Place Ravine
Help Restore Dell Place RavineCathy Grisham
 
Website design
Website designWebsite design
Website designBreezeGo
 
Love locks no more the replacements
Love locks no more   the replacementsLove locks no more   the replacements
Love locks no more the replacementsChase Rubin
 
Danu dean asmoro analisis relasi interpersonal mahasiswa fisip uajy
Danu dean asmoro   analisis relasi interpersonal mahasiswa fisip uajyDanu dean asmoro   analisis relasi interpersonal mahasiswa fisip uajy
Danu dean asmoro analisis relasi interpersonal mahasiswa fisip uajyDanu Dean Asmoro
 
「Callingood(コーリングッド)」のご案内資料です。
「Callingood(コーリングッド)」のご案内資料です。「Callingood(コーリングッド)」のご案内資料です。
「Callingood(コーリングッド)」のご案内資料です。Yuta Suenaga
 
Scripted Interface Redesign Preview
Scripted Interface Redesign PreviewScripted Interface Redesign Preview
Scripted Interface Redesign PreviewScripted.com
 
Repowering American Innovation
Repowering American InnovationRepowering American Innovation
Repowering American Innovationguest3d6941
 

Destacado (18)

Self-tuning Schedulers for Legacy Real-Time Applications
Self-tuning Schedulers for Legacy Real-Time ApplicationsSelf-tuning Schedulers for Legacy Real-Time Applications
Self-tuning Schedulers for Legacy Real-Time Applications
 
The Wizard of OS: a Heartbeat for Legacy Multimedia Applications
The Wizard of OS: a Heartbeat for Legacy Multimedia ApplicationsThe Wizard of OS: a Heartbeat for Legacy Multimedia Applications
The Wizard of OS: a Heartbeat for Legacy Multimedia Applications
 
Virtualised e-Learning with Real-Time Guarantees on the IRMOS Platform
Virtualised e-Learning with Real-Time Guarantees on the IRMOS PlatformVirtualised e-Learning with Real-Time Guarantees on the IRMOS Platform
Virtualised e-Learning with Real-Time Guarantees on the IRMOS Platform
 
Low-Latency Audio on Linux by Means of Real-Time Scheduling
Low-Latency Audio on Linux by Means of Real-Time SchedulingLow-Latency Audio on Linux by Means of Real-Time Scheduling
Low-Latency Audio on Linux by Means of Real-Time Scheduling
 
Optimum Scalability Point for Parallelisable Real-Time Components
Optimum Scalability Point for Parallelisable Real-Time ComponentsOptimum Scalability Point for Parallelisable Real-Time Components
Optimum Scalability Point for Parallelisable Real-Time Components
 
Real-Time API
Real-Time APIReal-Time API
Real-Time API
 
Help Restore Dell Place Ravine
Help Restore Dell Place RavineHelp Restore Dell Place Ravine
Help Restore Dell Place Ravine
 
Softskill kelompok 3 ppt [recovered]
Softskill kelompok 3 ppt [recovered]Softskill kelompok 3 ppt [recovered]
Softskill kelompok 3 ppt [recovered]
 
Website design
Website designWebsite design
Website design
 
Love locks no more the replacements
Love locks no more   the replacementsLove locks no more   the replacements
Love locks no more the replacements
 
Pat1
Pat1Pat1
Pat1
 
Danu dean asmoro analisis relasi interpersonal mahasiswa fisip uajy
Danu dean asmoro   analisis relasi interpersonal mahasiswa fisip uajyDanu dean asmoro   analisis relasi interpersonal mahasiswa fisip uajy
Danu dean asmoro analisis relasi interpersonal mahasiswa fisip uajy
 
Makalah huruf arab
Makalah huruf arabMakalah huruf arab
Makalah huruf arab
 
「Callingood(コーリングッド)」のご案内資料です。
「Callingood(コーリングッド)」のご案内資料です。「Callingood(コーリングッド)」のご案内資料です。
「Callingood(コーリングッド)」のご案内資料です。
 
NAS TS-453mini Presentation
NAS TS-453mini PresentationNAS TS-453mini Presentation
NAS TS-453mini Presentation
 
Scripted Interface Redesign Preview
Scripted Interface Redesign PreviewScripted Interface Redesign Preview
Scripted Interface Redesign Preview
 
Repowering American Innovation
Repowering American InnovationRepowering American Innovation
Repowering American Innovation
 
Gebeurtenis
GebeurtenisGebeurtenis
Gebeurtenis
 

Similar a Improving Responsiveness for Virtualized Networking Under Intensive Computing Workloads

Research in Soft Real-Time and Virtualized Applications on Linux
Research in Soft Real-Time and Virtualized Applications on LinuxResearch in Soft Real-Time and Virtualized Applications on Linux
Research in Soft Real-Time and Virtualized Applications on Linuxtcucinotta
 
The IRMOS Real-Time Scheduler
The IRMOS Real-Time SchedulerThe IRMOS Real-Time Scheduler
The IRMOS Real-Time Schedulertcucinotta
 
Self-tuning Schedulers for Legacy Real-Time Applications
Self-tuning Schedulers for Legacy Real-Time ApplicationsSelf-tuning Schedulers for Legacy Real-Time Applications
Self-tuning Schedulers for Legacy Real-Time Applicationsguestbbe1c83
 
Virtual Network Functions as Real-Time Containers in Private Clouds
Virtual Network Functions as Real-Time Containers in Private CloudsVirtual Network Functions as Real-Time Containers in Private Clouds
Virtual Network Functions as Real-Time Containers in Private Cloudstcucinotta
 
Mpls conference 2016-data center virtualisation-11-march
Mpls conference 2016-data center virtualisation-11-marchMpls conference 2016-data center virtualisation-11-march
Mpls conference 2016-data center virtualisation-11-marchAricent
 
5 Ways to Avoid Server and Application Downtime
5 Ways to Avoid Server and Application Downtime5 Ways to Avoid Server and Application Downtime
5 Ways to Avoid Server and Application DowntimeNeverfail Group
 
On Tune Virtualisation Monitoring
On Tune Virtualisation MonitoringOn Tune Virtualisation Monitoring
On Tune Virtualisation MonitoringTeemStone Pty Ltd
 
Netapp 1229343173196796-1
Netapp 1229343173196796-1Netapp 1229343173196796-1
Netapp 1229343173196796-1Newlink
 
Jacopo Nardiello - Monitoring Cloud-Native applications with Prometheus - Cod...
Jacopo Nardiello - Monitoring Cloud-Native applications with Prometheus - Cod...Jacopo Nardiello - Monitoring Cloud-Native applications with Prometheus - Cod...
Jacopo Nardiello - Monitoring Cloud-Native applications with Prometheus - Cod...Codemotion
 
OpenStackを利用したEnterprise Cloudを支える技術 - OpenStack最新情報セミナー 2016年5月
OpenStackを利用したEnterprise Cloudを支える技術 - OpenStack最新情報セミナー 2016年5月OpenStackを利用したEnterprise Cloudを支える技術 - OpenStack最新情報セミナー 2016年5月
OpenStackを利用したEnterprise Cloudを支える技術 - OpenStack最新情報セミナー 2016年5月VirtualTech Japan Inc.
 
An Introduction to the Emulex Network Xceleration Solution – FastStack™ Sniff...
An Introduction to the Emulex Network Xceleration Solution – FastStack™ Sniff...An Introduction to the Emulex Network Xceleration Solution – FastStack™ Sniff...
An Introduction to the Emulex Network Xceleration Solution – FastStack™ Sniff...Emulex Corporation
 
Tommaso Cucinotta - Low-latency and power-efficient audio applications on Linux
Tommaso Cucinotta - Low-latency and power-efficient audio applications on LinuxTommaso Cucinotta - Low-latency and power-efficient audio applications on Linux
Tommaso Cucinotta - Low-latency and power-efficient audio applications on Linuxlinuxlab_conf
 
Imex Research Virtualization Executive Summary On Slideshare
Imex Research Virtualization Executive Summary On SlideshareImex Research Virtualization Executive Summary On Slideshare
Imex Research Virtualization Executive Summary On SlideshareM. R. Pamidi, Ph. D.
 
Toward a practical “HPC Cloud”: Performance tuning of a virtualized HPC cluster
Toward a practical “HPC Cloud”: Performance tuning of a virtualized HPC clusterToward a practical “HPC Cloud”: Performance tuning of a virtualized HPC cluster
Toward a practical “HPC Cloud”: Performance tuning of a virtualized HPC clusterRyousei Takano
 
All about linux gaining root remote exploitation
All about linux  gaining root   remote exploitationAll about linux  gaining root   remote exploitation
All about linux gaining root remote exploitationn0rz
 
ANZTRUC - elmtree v2.2 (1)
ANZTRUC - elmtree v2.2 (1)ANZTRUC - elmtree v2.2 (1)
ANZTRUC - elmtree v2.2 (1)Mal Everett
 
Modeling and simulation of power consumption and execution times for real-tim...
Modeling and simulation of power consumption and execution times for real-tim...Modeling and simulation of power consumption and execution times for real-tim...
Modeling and simulation of power consumption and execution times for real-tim...tcucinotta
 

Similar a Improving Responsiveness for Virtualized Networking Under Intensive Computing Workloads (20)

Research in Soft Real-Time and Virtualized Applications on Linux
Research in Soft Real-Time and Virtualized Applications on LinuxResearch in Soft Real-Time and Virtualized Applications on Linux
Research in Soft Real-Time and Virtualized Applications on Linux
 
The IRMOS Real-Time Scheduler
The IRMOS Real-Time SchedulerThe IRMOS Real-Time Scheduler
The IRMOS Real-Time Scheduler
 
Self-tuning Schedulers for Legacy Real-Time Applications
Self-tuning Schedulers for Legacy Real-Time ApplicationsSelf-tuning Schedulers for Legacy Real-Time Applications
Self-tuning Schedulers for Legacy Real-Time Applications
 
Virtual Network Functions as Real-Time Containers in Private Clouds
Virtual Network Functions as Real-Time Containers in Private CloudsVirtual Network Functions as Real-Time Containers in Private Clouds
Virtual Network Functions as Real-Time Containers in Private Clouds
 
Mpls conference 2016-data center virtualisation-11-march
Mpls conference 2016-data center virtualisation-11-marchMpls conference 2016-data center virtualisation-11-march
Mpls conference 2016-data center virtualisation-11-march
 
5 Ways to Avoid Server and Application Downtime
5 Ways to Avoid Server and Application Downtime5 Ways to Avoid Server and Application Downtime
5 Ways to Avoid Server and Application Downtime
 
On Tune Virtualisation Monitoring
On Tune Virtualisation MonitoringOn Tune Virtualisation Monitoring
On Tune Virtualisation Monitoring
 
Xen Community Update 2011
Xen Community Update 2011Xen Community Update 2011
Xen Community Update 2011
 
Netapp 1229343173196796-1
Netapp 1229343173196796-1Netapp 1229343173196796-1
Netapp 1229343173196796-1
 
Jacopo Nardiello - Monitoring Cloud-Native applications with Prometheus - Cod...
Jacopo Nardiello - Monitoring Cloud-Native applications with Prometheus - Cod...Jacopo Nardiello - Monitoring Cloud-Native applications with Prometheus - Cod...
Jacopo Nardiello - Monitoring Cloud-Native applications with Prometheus - Cod...
 
OpenStackを利用したEnterprise Cloudを支える技術 - OpenStack最新情報セミナー 2016年5月
OpenStackを利用したEnterprise Cloudを支える技術 - OpenStack最新情報セミナー 2016年5月OpenStackを利用したEnterprise Cloudを支える技術 - OpenStack最新情報セミナー 2016年5月
OpenStackを利用したEnterprise Cloudを支える技術 - OpenStack最新情報セミナー 2016年5月
 
An Introduction to the Emulex Network Xceleration Solution – FastStack™ Sniff...
An Introduction to the Emulex Network Xceleration Solution – FastStack™ Sniff...An Introduction to the Emulex Network Xceleration Solution – FastStack™ Sniff...
An Introduction to the Emulex Network Xceleration Solution – FastStack™ Sniff...
 
Queues, Pools, Caches
Queues, Pools, CachesQueues, Pools, Caches
Queues, Pools, Caches
 
Tommaso Cucinotta - Low-latency and power-efficient audio applications on Linux
Tommaso Cucinotta - Low-latency and power-efficient audio applications on LinuxTommaso Cucinotta - Low-latency and power-efficient audio applications on Linux
Tommaso Cucinotta - Low-latency and power-efficient audio applications on Linux
 
Imex Research Virtualization Executive Summary On Slideshare
Imex Research Virtualization Executive Summary On SlideshareImex Research Virtualization Executive Summary On Slideshare
Imex Research Virtualization Executive Summary On Slideshare
 
Toward a practical “HPC Cloud”: Performance tuning of a virtualized HPC cluster
Toward a practical “HPC Cloud”: Performance tuning of a virtualized HPC clusterToward a practical “HPC Cloud”: Performance tuning of a virtualized HPC cluster
Toward a practical “HPC Cloud”: Performance tuning of a virtualized HPC cluster
 
All about linux gaining root remote exploitation
All about linux  gaining root   remote exploitationAll about linux  gaining root   remote exploitation
All about linux gaining root remote exploitation
 
ANZTRUC - elmtree v2.2 (1)
ANZTRUC - elmtree v2.2 (1)ANZTRUC - elmtree v2.2 (1)
ANZTRUC - elmtree v2.2 (1)
 
Modeling and simulation of power consumption and execution times for real-tim...
Modeling and simulation of power consumption and execution times for real-tim...Modeling and simulation of power consumption and execution times for real-tim...
Modeling and simulation of power consumption and execution times for real-tim...
 
Xen revisited
Xen revisitedXen revisited
Xen revisited
 

Último

Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...apidays
 
Boost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfBoost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfsudhanshuwaghmare1
 
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin WoodPolkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin WoodJuan lago vázquez
 
Artificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyArtificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyKhushali Kathiriya
 
DBX First Quarter 2024 Investor Presentation
DBX First Quarter 2024 Investor PresentationDBX First Quarter 2024 Investor Presentation
DBX First Quarter 2024 Investor PresentationDropbox
 
Spring Boot vs Quarkus the ultimate battle - DevoxxUK
Spring Boot vs Quarkus the ultimate battle - DevoxxUKSpring Boot vs Quarkus the ultimate battle - DevoxxUK
Spring Boot vs Quarkus the ultimate battle - DevoxxUKJago de Vreede
 
[BuildWithAI] Introduction to Gemini.pdf
[BuildWithAI] Introduction to Gemini.pdf[BuildWithAI] Introduction to Gemini.pdf
[BuildWithAI] Introduction to Gemini.pdfSandro Moreira
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FMESafe Software
 
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc
 
Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024
Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024
Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024Victor Rentea
 
Exploring Multimodal Embeddings with Milvus
Exploring Multimodal Embeddings with MilvusExploring Multimodal Embeddings with Milvus
Exploring Multimodal Embeddings with MilvusZilliz
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerThousandEyes
 
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ..."I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...Zilliz
 
Why Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businessWhy Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businesspanagenda
 
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingRepurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingEdi Saputra
 
Corporate and higher education May webinar.pptx
Corporate and higher education May webinar.pptxCorporate and higher education May webinar.pptx
Corporate and higher education May webinar.pptxRustici Software
 
CNIC Information System with Pakdata Cf In Pakistan
CNIC Information System with Pakdata Cf In PakistanCNIC Information System with Pakdata Cf In Pakistan
CNIC Information System with Pakdata Cf In Pakistandanishmna97
 
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, AdobeApidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobeapidays
 
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...apidays
 

Último (20)

Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
 
Boost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfBoost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdf
 
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin WoodPolkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
 
Artificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyArtificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : Uncertainty
 
DBX First Quarter 2024 Investor Presentation
DBX First Quarter 2024 Investor PresentationDBX First Quarter 2024 Investor Presentation
DBX First Quarter 2024 Investor Presentation
 
Spring Boot vs Quarkus the ultimate battle - DevoxxUK
Spring Boot vs Quarkus the ultimate battle - DevoxxUKSpring Boot vs Quarkus the ultimate battle - DevoxxUK
Spring Boot vs Quarkus the ultimate battle - DevoxxUK
 
[BuildWithAI] Introduction to Gemini.pdf
[BuildWithAI] Introduction to Gemini.pdf[BuildWithAI] Introduction to Gemini.pdf
[BuildWithAI] Introduction to Gemini.pdf
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
 
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
 
Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024
Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024
Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024
 
Exploring Multimodal Embeddings with Milvus
Exploring Multimodal Embeddings with MilvusExploring Multimodal Embeddings with Milvus
Exploring Multimodal Embeddings with Milvus
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected Worker
 
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ..."I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...
 
Why Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businessWhy Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire business
 
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingRepurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
 
Corporate and higher education May webinar.pptx
Corporate and higher education May webinar.pptxCorporate and higher education May webinar.pptx
Corporate and higher education May webinar.pptx
 
CNIC Information System with Pakdata Cf In Pakistan
CNIC Information System with Pakdata Cf In PakistanCNIC Information System with Pakdata Cf In Pakistan
CNIC Information System with Pakdata Cf In Pakistan
 
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, AdobeApidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
 
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
 
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
 

Improving Responsiveness for Virtualized Networking Under Intensive Computing Workloads

  • 1. RTLWS 2011 October 21st, Prague Improving Responsiveness for Virtualized Networking Under Intensive Computing Workloads Tommaso Cucinotta, Fabio Checconi, Dhaval Giani Cucinotta, Real-Time Systems Lab (RETIS) Scuola Superiore Sant'Anna, Pisa (Italy)
  • 2. Context Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy 2
  • 3. What is Real-Time Drive assistance  Engine control, brakes, stability, speed, parking  Trajectory and set-up control Defence, army, space Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore 3
  • 4. What is Real-Time Control of chemical and nuclear plants Control of productive processes and industrial automation Traffic control Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore 4
  • 5. What is Real-Time Multimedia, videosurveillance Augmented virtual reality Telecommunications Environment monitoring Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore 5
  • 6. Our Focus We focus on systems  With non-critical soft real-time requirements  Where the use of a GPOS is desirable and feasible • As opposed to a traditional RTOS  Where virtualization is desirable and feasible Application scenarios  Multimedia  Embedded systems  QoS-enabled Cloud Computing  Web servers with QoS assurance Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore
  • 7. Motivations Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy 7
  • 8. Virtualization Virtualization is a key technology For IaaS providers (Cloud Computing) Enables server consolidation Physical Host Physical Host OS OS ... Physical Host Physical Host OS OS Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy 8
  • 9. Virtualization Virtualization is a key technology For IaaS providers (Cloud Computing) Enables server consolidation Physical Host Physical Host Physical Host Physical Host OS OS VM/OS VM/OS ... ... Physical Host VM/OS VM/OS Physical Host OS OS Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy 9
  • 10. Need for Performance Isolation Resource sharing → Temporal interference Physical Host Physical Host VM VM VM VM VM VM VM VM Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy 10
  • 11. Need for Performance Isolation Resource sharing ~30ms VM Alone → Temporal interference Physical Host Physical Host VM VM VM VM VM VM VM VM Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy 11
  • 12. Need for Performance Isolation Resource sharing ~30ms VM Alone → Temporal interference Physical Host Physical Host VM VM τ1 = (30ms, 150ms) τ2 = (50ms, 200ms) VM VM ~120ms 2 VMs VM VM VM VM Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy 12
  • 13. Possible Solutions Hardware replication and Physical Host Physical Host static partitioning VM VM Computing • Multi-core (1 core per VM) • NUMA awareness VM VM Networking • Multiple network adapters VM VM (1 network adapter per VM) • Multi-queue adapters VM VM Drawbacks Limitation of flexibility Under-utilization of resources • e.g., with real-time/interactive workloads (cloud computing) Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy 13
  • 14. Possible Solutions Another approach Let multiple VMs use the same resources Use proper resource scheduling strategies For example Computing • Xen credit-based and SEDF schedulers Networking • QoS-aware protocols (IntServ, MPLS, WF2Q+) Advantages Increased flexibility Increased resource saturation levels Reduced infrastructure costs Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy 14
  • 15. Background on Deadline-based Scheduling Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy 15
  • 16. Real-Time Schedulers for Linux SCHED_SPORADIC  Additional scheduling policy within sched_rt.c  Single-threaded applications  API based on system call: sched_setscheduler() SCHED_DEADLINE  Independent scheduling class  Single-threaded control applications  API based on system call: sched_setscheduler() IRMOS (a.k.a., EDF Throttling)  Rewrite of RT throttling in sched_rt.c: from limits to guarantees  Multi-threaded complex software components • KVM, JVM, Apache+Tomcat, …  API based on cgroups interface Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore
  • 17. IRMOS Scheduler Slice the available computing power into reservations CPU 0 CPU 1 CPU 2 CPU 3 Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore
  • 18. Hierarchical Scheduling EDF EDF Max RT Prio … FP FP FP FP Min RT Prio T1 T1 T2 T2 T3 T3 T4 T4 Needed operations  create & destroy reservations  attach & detach tasks ↔ reservations  list tasks attached to reservations (and list reservations)  Standard operations: get & set parameters Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore 18
  • 19. Background Scheduling KVM VMs with resource reservations (deadline-based scheduling)  VM1 granted Q1 time units every P1  VM2 granted Q2 time units every P2  VM3 granted Q3 time units every P3 . . . Networking Threads KVM Computing Thread Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore 19
  • 20. Problem Presentation Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy 20
  • 21. Problem Presentation Focus on compute-intensive VM Schedule with VM alone  Unresponsiveness frame: P-Q Well-known worst-case schedule  Unresponsiveness frame: 2(P-Q) Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore 21
  • 22. Problem Presentation Packet arrival in the unresponsiveness frame  Its handling deferred till next activation period Mixed workload inside VM packet  Batch compute-intensive workload  Responsive application(s) • e.g., triggered by arriving packet Compute-intensive workload  Period as large as possible • For minimizing overheads Responsive workload  Period tuned on interactivity requirements • VM unresponsive for a time-frame as high as 2(P-Q) Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore 22
  • 23. Proposed Approach On packet arrival with null budget  Attach the VM to a temporary reservation • With period Ps << P • With budget Qs sufficient to “respond” to the request – Including both kernel-level and user-space level operations  Temporary reservation lasts • till budget exhaustion or original budget replenishment Worst-case latency (assump. of sporadic requests)  From 2(P-Q) to 2(Ps-Qs) Bandwidth occupation  From to Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore 23
  • 24. Experimental Results Experimental set-up  Periodic ping every 1s  Normal reservation: (40ms, 100ms)  Spare reservation: (4ms, 10ms) Batch activity  Fake loop logging experienced loops/s Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore 24
  • 25. Achieved results  Reservation (40,100) and no spare reservation • Response-time ~ 120ms (top curve) • Throughput ~ 1,11 cycles/us  Reservation (40,100) with spare reservation (4,10) • Response-time ~ 16ms (bottom curve)  Reservation (4,10) • Throughput ~ 0,56 cycles/us Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore 25
  • 26. Implementation Details Tommaso Cucinotta – ReTiS Lab – Scuola Superiore Sant'Anna – Pisa – Italy 26
  • 27. Modifications to the Scheduler Proper flags into the cgroup filesystem to tag  Spare bandwidth provider groups  Spare bandwidth client groups Modifications to the networking receive path  if packet goes to VM within client group (and budget is null),  then attach VM to (any matching) provider group Spare bandwidth clients/providers matching  Based on numerical identifier (clients can attach only to servers with the same identifier)  Allows for realizing spare bandwidth pools Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore 27
  • 28. Future Work More extensive evaluation  Presented results are very preliminary  Need to test with user-space application, not only ping Servers pooling  All VMs need the spare reservation at the same time • With a very low probability • So, leverage statistical multiplexing and use #providers < #clients • Lowers spare bandwidth requirements VM VM VM VM VM VM S S VM VM S S VM VM VM VM Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore 28
  • 29. Related Publications  Hierarchical Multiprocessor CPU Reservations for the Linux Kernel F. Checconi, T. Cucinotta, D. Faggioli, G. Lipari OSPERT 2009, Dublin, Ireland, June 2009  An EDF Scheduling class for the Linux kernel D. Faggioli, F. Checconi, M. Trimarchi, C. Scordino RTLWS 2009, Dresden, October 2009  Self-tuning Schedulers for Legacy Real-Time Applications T. Cucinotta, F. Checconi, L. Abeni, L. Palopoli EuroSys 2010, Paris, April 2010  Respecting temporal constraints in virtualised services T. Cucinotta, G. Anastasi, L. Abeni RTSOAA 2009, Seattle, Washington, July 2009 Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore
  • 30. Thanks for your attention Questions ? http://retis.sssup.it/people/tommaso Tommaso Cucinotta – Real Time Systems Laboratory (ReTiS) – Scuola Superiore Sant'Anna Sant’Anna © 2007 Scuola Superiore 30