SlideShare a Scribd company logo
1 of 30
Download to read offline
Sat o r i:
En ligh t e n e d Page Sh ar in g
    G r ze go r z M ilo s, D e r e k G . M u r r ay ,
     St e v e n H an d , M ic h ae l Fe t t e r m an
O u t lin e

• W h y sh ar e p age s?
• W h y u se e n ligh t e n m e n t s?
• U n iq u e se llin g p r o p o sit io n s
    – D u p lic at io n d e t e c t io n
    – Sav in gs d ist r ib u t io n
    – Fau lt h an d lin g
• Im p le m e n t at io n
• Pe r fo r m an c e r e su lt s
Page sh ar in g

• Id e n t ic al O Ss sh ar e lo t s o f id e n t ic al d at a
    – Pr o gr am b in ar ie s
    – Sh ar e d lib r ar ie s
    – C o n figu r at io n file s
• M o r e m e m o r y is alw ay s u se fu l
    – R e d u c e p agin g
    – In c r e ase p age c ac h e size
    – L e ss I/O , b e t t e r p e r fo r m an c e !
En ligh t e n m e n t

                      … o r “p ar av ir t u alisat io n ”
                          … o r “c h an gin g t h e O S”
• U se in fo r m at io n fr o m t h e O S t o gu id e sh ar in g
    – R e d u c e s t h e o v e r h e ad
• A v o id p at h o lo gic al b e h av io u r
    – D o u b le p agin g
    – H o st d e m an d p agin g
• Pr o o f-o f-c o n c e p t o n L in u x
U SP 1 : d u p lic at e d e t e c t io n

• Sat o r i m o n it o r s b lo c k r e ad s
    – H ash d at a w h e n r e ad , sh ar e p age s if id e n t ic al
    – B lo c k r e ad s fr o m c o m m o n su b st r at e d isk s
• O t h e r sy st e m s sc an m e m o r y p e r io d ic ally
    – R at e lim it e d t o p r e v e n t o v e r h e ad
    – C an m iss sh o r t -liv e d sh ar in g o p p o r t u n it ie s
U SP 2: sav in gs d ist r ib u t io n

• W an t t o e n c o u r age sh ar in g
   – T h e m o r e y o u sh ar e t h e m o r e m e m o r y y o u
     ge t
   – Eac h V M ge t s a sharing entitlement b ase d o n
     h o w m an y p age s it sh ar e s
• D e flat e t h e b allo o n t o in c r e ase allo w an c e
U SP 3: fau lt h an d lin g

• W h at h ap p e n s w h e n sh ar in g is b r o k e n ?
   – N e e d so m e m o r e m e m o r y
   – … b u t it ’s in u se
   – T y p ic ally u se h o st p agin g
• M o d ify t h e O S t o n o m in at e “v o lat ile ” p age s
   – C le an b u ffe r c ac h e p age s
   – H y p e r v iso r c an r e assign t h e m w it h o u t w ar n in g
   – “R e p ay m e n t FIFO ” b ase d o n IB M ’s C M M
Im p le m e n t at io n

• H y p e r v iso r   (5351 L O C )
    – B ase d o n X e n 3.1 w it h PV sh ad o w PT s
    – A d d e d sh ar in g-r e lat e d h y p e r c alls
• D om0       (3894 L O C )
    – M o d ifie d b lk t ap t o p e r fo r m h ash lo o k u p
• D omU       (2306 L O C )
    – Po r t e d IB M ’s C M M t o X e n /x 86
O v e r h e ad s

• W o r st -c ase : lar ge -c h u n k se q u e n t ial r e ad s
    – O v e r h e ad o f h ash in g: 0.2%
    – O v e r h e ad o f h ash in g + IPC : 34.8%
• “R e alist ic ” b e n c h m ar k (k e r n e l c o m p ile )
    – W it h o u t Sat o r i: 780 se c o n d s
    – Sat o r i e n ab le d : 779 se c o n d s
Effe c t iv e n e ss
                             Kernel Compilation (256MB)

 16000
 14000
 12000
 10000
                                                                                          Potential
  8000
                                                                                          Satori
Pages
  6000
  4000
  2000
         0
             0   2   4   6   8   10   12     14   16   18   20   22   24   26   28   30

                                           Time (mins)
Effe c t iv e n e ss
                         Kernel Compilation (512MB)

 30000

 25000

 20000
                                                                                      Potential
 15000
                                                                                      Satori
Pages
 10000

  5000

     0
         0   2   4   6   8   10   12     14   16   18   20   22   24   26   28   30

                                       Time (mins)
Effe c t iv e n e ss
                                              httperf

 120000

 100000

  80000
                                                                                           Potential
  60000
                                                                                           Satori
Pages
  40000

  20000

          0
              0   2   4   6   8   10   12     14   16   18   20   22   24   26   28   30

                                            Time (mins)
Effe c t iv e n e ss
                                          RUBiS

 30000

 25000

 20000
                                                                                      Potential
 15000
                                                                                      Satori
Pages
 10000

  5000

     0
         0   2   4   6   8   10   12     14   16   18   20   22   24   26   28   30

                                       Time (mins)
C o n c lu sio n s

• M o r e m e m o r y m e an s m o r e p e r fo r m an c e
• Page c ac h e s ar e u se fu l so u r c e s o f sh ar in g
• En ligh t e n in g t h e O S av o id s h o st p agin g

• Sat o r i is e ffic ie n t an d e ffe c t iv e !
Q u e st io n s?
Sat o r i:
En ligh t e n e d Page Sh ar in g
    G r ze go r z M ilo s, D e r e k G . M u r r ay ,
     St e v e n H an d , M ic h ae l Fe t t e r m an

                                                        1
O u t lin e

• W h y sh ar e p age s?
• W h y u se e n ligh t e n m e n t s?
• U n iq u e se llin g p r o p o sit io n s
    – D u p lic at io n d e t e c t io n
    – Sav in gs d ist r ib u t io n
    – Fau lt h an d lin g
• Im p le m e n t at io n
• Pe r fo r m an c e r e su lt s
                                              2
Page sh ar in g

• Id e n t ic al O Ss sh ar e lo t s o f id e n t ic al d at a
    – Pr o gr am b in ar ie s
    – Sh ar e d lib r ar ie s
    – C o n figu r at io n file s
• M o r e m e m o r y is alw ay s u se fu l
    – R e d u c e p agin g
    – In c r e ase p age c ac h e size
    – L e ss I/O , b e t t e r p e r fo r m an c e !
                                                                 3
En ligh t e n m e n t

                      … o r “p ar av ir t u alisat io n ”
                          … o r “c h an gin g t h e O S”
• U se in fo r m at io n fr o m t h e O S t o gu id e sh ar in g
    – R e d u c e s t h e o v e r h e ad
• A v o id p at h o lo gic al b e h av io u r
    – D o u b le p agin g
    – H o st d e m an d p agin g
• Pr o o f-o f-c o n c e p t o n L in u x

                                                                   4
U SP 1 : d u p lic at e d e t e c t io n

• Sat o r i m o n it o r s b lo c k r e ad s
    – H ash d at a w h e n r e ad , sh ar e p age s if id e n t ic al
    – B lo c k r e ad s fr o m c o m m o n su b st r at e d isk s
• O t h e r sy st e m s sc an m e m o r y p e r io d ic ally
    – R at e lim it e d t o p r e v e n t o v e r h e ad
    – C an m iss sh o r t -liv e d sh ar in g o p p o r t u n it ie s


                                                                        5
U SP 2: sav in gs d ist r ib u t io n

• W an t t o e n c o u r age sh ar in g
   – T h e m o r e y o u sh ar e t h e m o r e m e m o r y y o u
     ge t
   – Eac h V M ge t s a sharing entitlement b ase d o n
     h o w m an y p age s it sh ar e s
• D e flat e t h e b allo o n t o in c r e ase allo w an c e


                                                                   6
U SP 3: fau lt h an d lin g

• W h at h ap p e n s w h e n sh ar in g is b r o k e n ?
   – N e e d so m e m o r e m e m o r y
   – … b u t it ’s in u se
   – T y p ic ally u se h o st p agin g
• M o d ify t h e O S t o n o m in at e “v o lat ile ” p age s
   – C le an b u ffe r c ac h e p age s
   – H y p e r v iso r c an r e assign t h e m w it h o u t w ar n in g
   – “R e p ay m e n t FIFO ” b ase d o n IB M ’s C M M
                                                                          7
Im p le m e n t at io n

• H y p e r v iso r   (5351 L O C )
    – B ase d o n X e n 3.1 w it h PV sh ad o w PT s
    – A d d e d sh ar in g-r e lat e d h y p e r c alls
• D om0       (3894 L O C )
    – M o d ifie d b lk t ap t o p e r fo r m h ash lo o k u p
• D omU       (2306 L O C )
    – Po r t e d IB M ’s C M M t o X e n /x 86

                                                                 8
O v e r h e ad s

• W o r st -c ase : lar ge -c h u n k se q u e n t ial r e ad s
    – O v e r h e ad o f h ash in g: 0.2%
    – O v e r h e ad o f h ash in g + IPC : 34.8%
• “R e alist ic ” b e n c h m ar k (k e r n e l c o m p ile )
    – W it h o u t Sat o r i: 780 se c o n d s
    – Sat o r i e n ab le d : 779 se c o n d s


                                                                  9
Effe c t iv e n e ss
                             Kernel Compilation (256MB)

 16000
 14000
 12000
 10000
                                                                                          Potential
  8000
                                                                                          Satori
Pages
  6000
  4000
  2000
         0
             0   2   4   6   8   10   12     14   16   18   20   22   24   26   28   30
                                                                                          10
                                           Time (mins)
Effe c t iv e n e ss
                             Kernel Compilation (512MB)

 30000

 25000

 20000
                                                                                          Potential
 15000
                                                                                          Satori
Pages
 10000

  5000

         0
             0   2   4   6   8   10   12     14   16   18   20   22   24   26   28   30
                                                                                          11
                                           Time (mins)
Effe c t iv e n e ss
                                              httperf

 120000

 100000

  80000
                                                                                           Potential
  60000
                                                                                           Satori
Pages
  40000

  20000

          0
              0   2   4   6   8   10   12     14   16   18   20   22   24   26   28   30
                                                                                           12
                                            Time (mins)
Effe c t iv e n e ss
                                              RUBiS

 30000

 25000

 20000
                                                                                          Potential
 15000
                                                                                          Satori
Pages
 10000

  5000

         0
             0   2   4   6   8   10   12     14   16   18   20   22   24   26   28   30
                                                                                          13
                                           Time (mins)
C o n c lu sio n s

• M o r e m e m o r y m e an s m o r e p e r fo r m an c e
• Page c ac h e s ar e u se fu l so u r c e s o f sh ar in g
• En ligh t e n in g t h e O S av o id s h o st p agin g

• Sat o r i is e ffic ie n t an d e ffe c t iv e !



                                                               14
Q u e st io n s?

• C lic k t o ad d an o u t lin e




                                    15

More Related Content

What's hot

Pengabdian e jurnal
Pengabdian e jurnalPengabdian e jurnal
Pengabdian e jurnalNana Citra
 
Comandos Packet Tracer
Comandos Packet TracerComandos Packet Tracer
Comandos Packet Tracerdanists
 
Boletim Eletronico n.1
Boletim Eletronico n.1Boletim Eletronico n.1
Boletim Eletronico n.1Vânia Galvão
 
Qual a importância do Design de UX para desenvolver projetos digitais
Qual a importância do Design de UX para desenvolver projetos digitaisQual a importância do Design de UX para desenvolver projetos digitais
Qual a importância do Design de UX para desenvolver projetos digitaisFrancisco Esperança
 
campaña electoral
campaña electoralcampaña electoral
campaña electoraladricisneros
 
Mfuse - Building & Managing Mobile Solutions - Whitepaper - Oct 12
Mfuse - Building & Managing Mobile Solutions - Whitepaper - Oct 12Mfuse - Building & Managing Mobile Solutions - Whitepaper - Oct 12
Mfuse - Building & Managing Mobile Solutions - Whitepaper - Oct 12Mfuse Limited
 
I MARXA PASQÜERA 2013
I MARXA PASQÜERA 2013I MARXA PASQÜERA 2013
I MARXA PASQÜERA 2013rebentaromorir
 
2007 affordable housing development competition
2007 affordable housing development competition2007 affordable housing development competition
2007 affordable housing development competitionKelvin Xuna
 
Angel-Luis Fernandez - Intelligent Energy Management in Madrid Urban Areas
Angel-Luis Fernandez - Intelligent Energy Management in Madrid Urban AreasAngel-Luis Fernandez - Intelligent Energy Management in Madrid Urban Areas
Angel-Luis Fernandez - Intelligent Energy Management in Madrid Urban AreasShane Mitchell
 
Growing business online with Oscaro
Growing business online with OscaroGrowing business online with Oscaro
Growing business online with OscaroRocco Baldassarre
 
in the harvard news paper.
in the harvard news paper.in the harvard news paper.
in the harvard news paper.Kelvin Xuna
 

What's hot (20)

Audiwords1
Audiwords1Audiwords1
Audiwords1
 
Starbucks – leaving australia
Starbucks – leaving australiaStarbucks – leaving australia
Starbucks – leaving australia
 
MECANICA AUTOMOTRIZ
MECANICA AUTOMOTRIZMECANICA AUTOMOTRIZ
MECANICA AUTOMOTRIZ
 
XRAii#3 Adam Zygadlewicz - Netguru
XRAii#3 Adam Zygadlewicz - NetguruXRAii#3 Adam Zygadlewicz - Netguru
XRAii#3 Adam Zygadlewicz - Netguru
 
Pengabdian e jurnal
Pengabdian e jurnalPengabdian e jurnal
Pengabdian e jurnal
 
Comandos Packet Tracer
Comandos Packet TracerComandos Packet Tracer
Comandos Packet Tracer
 
Boletim Eletronico n.1
Boletim Eletronico n.1Boletim Eletronico n.1
Boletim Eletronico n.1
 
Proyecto yo
Proyecto yoProyecto yo
Proyecto yo
 
Operational
OperationalOperational
Operational
 
Qual a importância do Design de UX para desenvolver projetos digitais
Qual a importância do Design de UX para desenvolver projetos digitaisQual a importância do Design de UX para desenvolver projetos digitais
Qual a importância do Design de UX para desenvolver projetos digitais
 
recomendaciones
recomendacionesrecomendaciones
recomendaciones
 
campaña electoral
campaña electoralcampaña electoral
campaña electoral
 
Mfuse - Building & Managing Mobile Solutions - Whitepaper - Oct 12
Mfuse - Building & Managing Mobile Solutions - Whitepaper - Oct 12Mfuse - Building & Managing Mobile Solutions - Whitepaper - Oct 12
Mfuse - Building & Managing Mobile Solutions - Whitepaper - Oct 12
 
Dmitry krivitsky web
Dmitry krivitsky web Dmitry krivitsky web
Dmitry krivitsky web
 
I MARXA PASQÜERA 2013
I MARXA PASQÜERA 2013I MARXA PASQÜERA 2013
I MARXA PASQÜERA 2013
 
2007 affordable housing development competition
2007 affordable housing development competition2007 affordable housing development competition
2007 affordable housing development competition
 
Análisis de la Crisis Internacional - Versión 2
Análisis de la Crisis Internacional - Versión 2Análisis de la Crisis Internacional - Versión 2
Análisis de la Crisis Internacional - Versión 2
 
Angel-Luis Fernandez - Intelligent Energy Management in Madrid Urban Areas
Angel-Luis Fernandez - Intelligent Energy Management in Madrid Urban AreasAngel-Luis Fernandez - Intelligent Energy Management in Madrid Urban Areas
Angel-Luis Fernandez - Intelligent Energy Management in Madrid Urban Areas
 
Growing business online with Oscaro
Growing business online with OscaroGrowing business online with Oscaro
Growing business online with Oscaro
 
in the harvard news paper.
in the harvard news paper.in the harvard news paper.
in the harvard news paper.
 

More from The Linux Foundation

ELC2019: Static Partitioning Made Simple
ELC2019: Static Partitioning Made SimpleELC2019: Static Partitioning Made Simple
ELC2019: Static Partitioning Made SimpleThe Linux Foundation
 
XPDDS19: How TrenchBoot is Enabling Measured Launch for Open-Source Platform ...
XPDDS19: How TrenchBoot is Enabling Measured Launch for Open-Source Platform ...XPDDS19: How TrenchBoot is Enabling Measured Launch for Open-Source Platform ...
XPDDS19: How TrenchBoot is Enabling Measured Launch for Open-Source Platform ...The Linux Foundation
 
XPDDS19 Keynote: Xen in Automotive - Artem Mygaiev, Director, Technology Solu...
XPDDS19 Keynote: Xen in Automotive - Artem Mygaiev, Director, Technology Solu...XPDDS19 Keynote: Xen in Automotive - Artem Mygaiev, Director, Technology Solu...
XPDDS19 Keynote: Xen in Automotive - Artem Mygaiev, Director, Technology Solu...The Linux Foundation
 
XPDDS19 Keynote: Xen Project Weather Report 2019 - Lars Kurth, Director of Op...
XPDDS19 Keynote: Xen Project Weather Report 2019 - Lars Kurth, Director of Op...XPDDS19 Keynote: Xen Project Weather Report 2019 - Lars Kurth, Director of Op...
XPDDS19 Keynote: Xen Project Weather Report 2019 - Lars Kurth, Director of Op...The Linux Foundation
 
XPDDS19 Keynote: Unikraft Weather Report
XPDDS19 Keynote:  Unikraft Weather ReportXPDDS19 Keynote:  Unikraft Weather Report
XPDDS19 Keynote: Unikraft Weather ReportThe Linux Foundation
 
XPDDS19 Keynote: Secret-free Hypervisor: Now and Future - Wei Liu, Software E...
XPDDS19 Keynote: Secret-free Hypervisor: Now and Future - Wei Liu, Software E...XPDDS19 Keynote: Secret-free Hypervisor: Now and Future - Wei Liu, Software E...
XPDDS19 Keynote: Secret-free Hypervisor: Now and Future - Wei Liu, Software E...The Linux Foundation
 
XPDDS19 Keynote: Xen Dom0-less - Stefano Stabellini, Principal Engineer, Xilinx
XPDDS19 Keynote: Xen Dom0-less - Stefano Stabellini, Principal Engineer, XilinxXPDDS19 Keynote: Xen Dom0-less - Stefano Stabellini, Principal Engineer, Xilinx
XPDDS19 Keynote: Xen Dom0-less - Stefano Stabellini, Principal Engineer, XilinxThe Linux Foundation
 
XPDDS19 Keynote: Patch Review for Non-maintainers - George Dunlap, Citrix Sys...
XPDDS19 Keynote: Patch Review for Non-maintainers - George Dunlap, Citrix Sys...XPDDS19 Keynote: Patch Review for Non-maintainers - George Dunlap, Citrix Sys...
XPDDS19 Keynote: Patch Review for Non-maintainers - George Dunlap, Citrix Sys...The Linux Foundation
 
XPDDS19: Memories of a VM Funk - Mihai Donțu, Bitdefender
XPDDS19: Memories of a VM Funk - Mihai Donțu, BitdefenderXPDDS19: Memories of a VM Funk - Mihai Donțu, Bitdefender
XPDDS19: Memories of a VM Funk - Mihai Donțu, BitdefenderThe Linux Foundation
 
OSSJP/ALS19: The Road to Safety Certification: Overcoming Community Challeng...
OSSJP/ALS19:  The Road to Safety Certification: Overcoming Community Challeng...OSSJP/ALS19:  The Road to Safety Certification: Overcoming Community Challeng...
OSSJP/ALS19: The Road to Safety Certification: Overcoming Community Challeng...The Linux Foundation
 
OSSJP/ALS19: The Road to Safety Certification: How the Xen Project is Making...
 OSSJP/ALS19: The Road to Safety Certification: How the Xen Project is Making... OSSJP/ALS19: The Road to Safety Certification: How the Xen Project is Making...
OSSJP/ALS19: The Road to Safety Certification: How the Xen Project is Making...The Linux Foundation
 
XPDDS19: Speculative Sidechannels and Mitigations - Andrew Cooper, Citrix
XPDDS19: Speculative Sidechannels and Mitigations - Andrew Cooper, CitrixXPDDS19: Speculative Sidechannels and Mitigations - Andrew Cooper, Citrix
XPDDS19: Speculative Sidechannels and Mitigations - Andrew Cooper, CitrixThe Linux Foundation
 
XPDDS19: Keeping Coherency on Arm: Reborn - Julien Grall, Arm ltd
XPDDS19: Keeping Coherency on Arm: Reborn - Julien Grall, Arm ltdXPDDS19: Keeping Coherency on Arm: Reborn - Julien Grall, Arm ltd
XPDDS19: Keeping Coherency on Arm: Reborn - Julien Grall, Arm ltdThe Linux Foundation
 
XPDDS19: QEMU PV Backend 'qdevification'... What Does it Mean? - Paul Durrant...
XPDDS19: QEMU PV Backend 'qdevification'... What Does it Mean? - Paul Durrant...XPDDS19: QEMU PV Backend 'qdevification'... What Does it Mean? - Paul Durrant...
XPDDS19: QEMU PV Backend 'qdevification'... What Does it Mean? - Paul Durrant...The Linux Foundation
 
XPDDS19: Status of PCI Emulation in Xen - Roger Pau Monné, Citrix Systems R&D
XPDDS19: Status of PCI Emulation in Xen - Roger Pau Monné, Citrix Systems R&DXPDDS19: Status of PCI Emulation in Xen - Roger Pau Monné, Citrix Systems R&D
XPDDS19: Status of PCI Emulation in Xen - Roger Pau Monné, Citrix Systems R&DThe Linux Foundation
 
XPDDS19: [ARM] OP-TEE Mediator in Xen - Volodymyr Babchuk, EPAM Systems
XPDDS19: [ARM] OP-TEE Mediator in Xen - Volodymyr Babchuk, EPAM SystemsXPDDS19: [ARM] OP-TEE Mediator in Xen - Volodymyr Babchuk, EPAM Systems
XPDDS19: [ARM] OP-TEE Mediator in Xen - Volodymyr Babchuk, EPAM SystemsThe Linux Foundation
 
XPDDS19: Bringing Xen to the Masses: The Story of Building a Community-driven...
XPDDS19: Bringing Xen to the Masses: The Story of Building a Community-driven...XPDDS19: Bringing Xen to the Masses: The Story of Building a Community-driven...
XPDDS19: Bringing Xen to the Masses: The Story of Building a Community-driven...The Linux Foundation
 
XPDDS19: Will Robots Automate Your Job Away? Streamlining Xen Project Contrib...
XPDDS19: Will Robots Automate Your Job Away? Streamlining Xen Project Contrib...XPDDS19: Will Robots Automate Your Job Away? Streamlining Xen Project Contrib...
XPDDS19: Will Robots Automate Your Job Away? Streamlining Xen Project Contrib...The Linux Foundation
 
XPDDS19: Client Virtualization Toolstack in Go - Nick Rosbrook & Brendan Kerr...
XPDDS19: Client Virtualization Toolstack in Go - Nick Rosbrook & Brendan Kerr...XPDDS19: Client Virtualization Toolstack in Go - Nick Rosbrook & Brendan Kerr...
XPDDS19: Client Virtualization Toolstack in Go - Nick Rosbrook & Brendan Kerr...The Linux Foundation
 
XPDDS19: Core Scheduling in Xen - Jürgen Groß, SUSE
XPDDS19: Core Scheduling in Xen - Jürgen Groß, SUSEXPDDS19: Core Scheduling in Xen - Jürgen Groß, SUSE
XPDDS19: Core Scheduling in Xen - Jürgen Groß, SUSEThe Linux Foundation
 

More from The Linux Foundation (20)

ELC2019: Static Partitioning Made Simple
ELC2019: Static Partitioning Made SimpleELC2019: Static Partitioning Made Simple
ELC2019: Static Partitioning Made Simple
 
XPDDS19: How TrenchBoot is Enabling Measured Launch for Open-Source Platform ...
XPDDS19: How TrenchBoot is Enabling Measured Launch for Open-Source Platform ...XPDDS19: How TrenchBoot is Enabling Measured Launch for Open-Source Platform ...
XPDDS19: How TrenchBoot is Enabling Measured Launch for Open-Source Platform ...
 
XPDDS19 Keynote: Xen in Automotive - Artem Mygaiev, Director, Technology Solu...
XPDDS19 Keynote: Xen in Automotive - Artem Mygaiev, Director, Technology Solu...XPDDS19 Keynote: Xen in Automotive - Artem Mygaiev, Director, Technology Solu...
XPDDS19 Keynote: Xen in Automotive - Artem Mygaiev, Director, Technology Solu...
 
XPDDS19 Keynote: Xen Project Weather Report 2019 - Lars Kurth, Director of Op...
XPDDS19 Keynote: Xen Project Weather Report 2019 - Lars Kurth, Director of Op...XPDDS19 Keynote: Xen Project Weather Report 2019 - Lars Kurth, Director of Op...
XPDDS19 Keynote: Xen Project Weather Report 2019 - Lars Kurth, Director of Op...
 
XPDDS19 Keynote: Unikraft Weather Report
XPDDS19 Keynote:  Unikraft Weather ReportXPDDS19 Keynote:  Unikraft Weather Report
XPDDS19 Keynote: Unikraft Weather Report
 
XPDDS19 Keynote: Secret-free Hypervisor: Now and Future - Wei Liu, Software E...
XPDDS19 Keynote: Secret-free Hypervisor: Now and Future - Wei Liu, Software E...XPDDS19 Keynote: Secret-free Hypervisor: Now and Future - Wei Liu, Software E...
XPDDS19 Keynote: Secret-free Hypervisor: Now and Future - Wei Liu, Software E...
 
XPDDS19 Keynote: Xen Dom0-less - Stefano Stabellini, Principal Engineer, Xilinx
XPDDS19 Keynote: Xen Dom0-less - Stefano Stabellini, Principal Engineer, XilinxXPDDS19 Keynote: Xen Dom0-less - Stefano Stabellini, Principal Engineer, Xilinx
XPDDS19 Keynote: Xen Dom0-less - Stefano Stabellini, Principal Engineer, Xilinx
 
XPDDS19 Keynote: Patch Review for Non-maintainers - George Dunlap, Citrix Sys...
XPDDS19 Keynote: Patch Review for Non-maintainers - George Dunlap, Citrix Sys...XPDDS19 Keynote: Patch Review for Non-maintainers - George Dunlap, Citrix Sys...
XPDDS19 Keynote: Patch Review for Non-maintainers - George Dunlap, Citrix Sys...
 
XPDDS19: Memories of a VM Funk - Mihai Donțu, Bitdefender
XPDDS19: Memories of a VM Funk - Mihai Donțu, BitdefenderXPDDS19: Memories of a VM Funk - Mihai Donțu, Bitdefender
XPDDS19: Memories of a VM Funk - Mihai Donțu, Bitdefender
 
OSSJP/ALS19: The Road to Safety Certification: Overcoming Community Challeng...
OSSJP/ALS19:  The Road to Safety Certification: Overcoming Community Challeng...OSSJP/ALS19:  The Road to Safety Certification: Overcoming Community Challeng...
OSSJP/ALS19: The Road to Safety Certification: Overcoming Community Challeng...
 
OSSJP/ALS19: The Road to Safety Certification: How the Xen Project is Making...
 OSSJP/ALS19: The Road to Safety Certification: How the Xen Project is Making... OSSJP/ALS19: The Road to Safety Certification: How the Xen Project is Making...
OSSJP/ALS19: The Road to Safety Certification: How the Xen Project is Making...
 
XPDDS19: Speculative Sidechannels and Mitigations - Andrew Cooper, Citrix
XPDDS19: Speculative Sidechannels and Mitigations - Andrew Cooper, CitrixXPDDS19: Speculative Sidechannels and Mitigations - Andrew Cooper, Citrix
XPDDS19: Speculative Sidechannels and Mitigations - Andrew Cooper, Citrix
 
XPDDS19: Keeping Coherency on Arm: Reborn - Julien Grall, Arm ltd
XPDDS19: Keeping Coherency on Arm: Reborn - Julien Grall, Arm ltdXPDDS19: Keeping Coherency on Arm: Reborn - Julien Grall, Arm ltd
XPDDS19: Keeping Coherency on Arm: Reborn - Julien Grall, Arm ltd
 
XPDDS19: QEMU PV Backend 'qdevification'... What Does it Mean? - Paul Durrant...
XPDDS19: QEMU PV Backend 'qdevification'... What Does it Mean? - Paul Durrant...XPDDS19: QEMU PV Backend 'qdevification'... What Does it Mean? - Paul Durrant...
XPDDS19: QEMU PV Backend 'qdevification'... What Does it Mean? - Paul Durrant...
 
XPDDS19: Status of PCI Emulation in Xen - Roger Pau Monné, Citrix Systems R&D
XPDDS19: Status of PCI Emulation in Xen - Roger Pau Monné, Citrix Systems R&DXPDDS19: Status of PCI Emulation in Xen - Roger Pau Monné, Citrix Systems R&D
XPDDS19: Status of PCI Emulation in Xen - Roger Pau Monné, Citrix Systems R&D
 
XPDDS19: [ARM] OP-TEE Mediator in Xen - Volodymyr Babchuk, EPAM Systems
XPDDS19: [ARM] OP-TEE Mediator in Xen - Volodymyr Babchuk, EPAM SystemsXPDDS19: [ARM] OP-TEE Mediator in Xen - Volodymyr Babchuk, EPAM Systems
XPDDS19: [ARM] OP-TEE Mediator in Xen - Volodymyr Babchuk, EPAM Systems
 
XPDDS19: Bringing Xen to the Masses: The Story of Building a Community-driven...
XPDDS19: Bringing Xen to the Masses: The Story of Building a Community-driven...XPDDS19: Bringing Xen to the Masses: The Story of Building a Community-driven...
XPDDS19: Bringing Xen to the Masses: The Story of Building a Community-driven...
 
XPDDS19: Will Robots Automate Your Job Away? Streamlining Xen Project Contrib...
XPDDS19: Will Robots Automate Your Job Away? Streamlining Xen Project Contrib...XPDDS19: Will Robots Automate Your Job Away? Streamlining Xen Project Contrib...
XPDDS19: Will Robots Automate Your Job Away? Streamlining Xen Project Contrib...
 
XPDDS19: Client Virtualization Toolstack in Go - Nick Rosbrook & Brendan Kerr...
XPDDS19: Client Virtualization Toolstack in Go - Nick Rosbrook & Brendan Kerr...XPDDS19: Client Virtualization Toolstack in Go - Nick Rosbrook & Brendan Kerr...
XPDDS19: Client Virtualization Toolstack in Go - Nick Rosbrook & Brendan Kerr...
 
XPDDS19: Core Scheduling in Xen - Jürgen Groß, SUSE
XPDDS19: Core Scheduling in Xen - Jürgen Groß, SUSEXPDDS19: Core Scheduling in Xen - Jürgen Groß, SUSE
XPDDS19: Core Scheduling in Xen - Jürgen Groß, SUSE
 

XS Oracle 2009 Satori

  • 1. Sat o r i: En ligh t e n e d Page Sh ar in g G r ze go r z M ilo s, D e r e k G . M u r r ay , St e v e n H an d , M ic h ae l Fe t t e r m an
  • 2. O u t lin e • W h y sh ar e p age s? • W h y u se e n ligh t e n m e n t s? • U n iq u e se llin g p r o p o sit io n s – D u p lic at io n d e t e c t io n – Sav in gs d ist r ib u t io n – Fau lt h an d lin g • Im p le m e n t at io n • Pe r fo r m an c e r e su lt s
  • 3. Page sh ar in g • Id e n t ic al O Ss sh ar e lo t s o f id e n t ic al d at a – Pr o gr am b in ar ie s – Sh ar e d lib r ar ie s – C o n figu r at io n file s • M o r e m e m o r y is alw ay s u se fu l – R e d u c e p agin g – In c r e ase p age c ac h e size – L e ss I/O , b e t t e r p e r fo r m an c e !
  • 4. En ligh t e n m e n t … o r “p ar av ir t u alisat io n ” … o r “c h an gin g t h e O S” • U se in fo r m at io n fr o m t h e O S t o gu id e sh ar in g – R e d u c e s t h e o v e r h e ad • A v o id p at h o lo gic al b e h av io u r – D o u b le p agin g – H o st d e m an d p agin g • Pr o o f-o f-c o n c e p t o n L in u x
  • 5. U SP 1 : d u p lic at e d e t e c t io n • Sat o r i m o n it o r s b lo c k r e ad s – H ash d at a w h e n r e ad , sh ar e p age s if id e n t ic al – B lo c k r e ad s fr o m c o m m o n su b st r at e d isk s • O t h e r sy st e m s sc an m e m o r y p e r io d ic ally – R at e lim it e d t o p r e v e n t o v e r h e ad – C an m iss sh o r t -liv e d sh ar in g o p p o r t u n it ie s
  • 6. U SP 2: sav in gs d ist r ib u t io n • W an t t o e n c o u r age sh ar in g – T h e m o r e y o u sh ar e t h e m o r e m e m o r y y o u ge t – Eac h V M ge t s a sharing entitlement b ase d o n h o w m an y p age s it sh ar e s • D e flat e t h e b allo o n t o in c r e ase allo w an c e
  • 7. U SP 3: fau lt h an d lin g • W h at h ap p e n s w h e n sh ar in g is b r o k e n ? – N e e d so m e m o r e m e m o r y – … b u t it ’s in u se – T y p ic ally u se h o st p agin g • M o d ify t h e O S t o n o m in at e “v o lat ile ” p age s – C le an b u ffe r c ac h e p age s – H y p e r v iso r c an r e assign t h e m w it h o u t w ar n in g – “R e p ay m e n t FIFO ” b ase d o n IB M ’s C M M
  • 8. Im p le m e n t at io n • H y p e r v iso r (5351 L O C ) – B ase d o n X e n 3.1 w it h PV sh ad o w PT s – A d d e d sh ar in g-r e lat e d h y p e r c alls • D om0 (3894 L O C ) – M o d ifie d b lk t ap t o p e r fo r m h ash lo o k u p • D omU (2306 L O C ) – Po r t e d IB M ’s C M M t o X e n /x 86
  • 9. O v e r h e ad s • W o r st -c ase : lar ge -c h u n k se q u e n t ial r e ad s – O v e r h e ad o f h ash in g: 0.2% – O v e r h e ad o f h ash in g + IPC : 34.8% • “R e alist ic ” b e n c h m ar k (k e r n e l c o m p ile ) – W it h o u t Sat o r i: 780 se c o n d s – Sat o r i e n ab le d : 779 se c o n d s
  • 10. Effe c t iv e n e ss Kernel Compilation (256MB) 16000 14000 12000 10000 Potential 8000 Satori Pages 6000 4000 2000 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Time (mins)
  • 11. Effe c t iv e n e ss Kernel Compilation (512MB) 30000 25000 20000 Potential 15000 Satori Pages 10000 5000 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Time (mins)
  • 12. Effe c t iv e n e ss httperf 120000 100000 80000 Potential 60000 Satori Pages 40000 20000 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Time (mins)
  • 13. Effe c t iv e n e ss RUBiS 30000 25000 20000 Potential 15000 Satori Pages 10000 5000 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Time (mins)
  • 14. C o n c lu sio n s • M o r e m e m o r y m e an s m o r e p e r fo r m an c e • Page c ac h e s ar e u se fu l so u r c e s o f sh ar in g • En ligh t e n in g t h e O S av o id s h o st p agin g • Sat o r i is e ffic ie n t an d e ffe c t iv e !
  • 15. Q u e st io n s?
  • 16. Sat o r i: En ligh t e n e d Page Sh ar in g G r ze go r z M ilo s, D e r e k G . M u r r ay , St e v e n H an d , M ic h ae l Fe t t e r m an 1
  • 17. O u t lin e • W h y sh ar e p age s? • W h y u se e n ligh t e n m e n t s? • U n iq u e se llin g p r o p o sit io n s – D u p lic at io n d e t e c t io n – Sav in gs d ist r ib u t io n – Fau lt h an d lin g • Im p le m e n t at io n • Pe r fo r m an c e r e su lt s 2
  • 18. Page sh ar in g • Id e n t ic al O Ss sh ar e lo t s o f id e n t ic al d at a – Pr o gr am b in ar ie s – Sh ar e d lib r ar ie s – C o n figu r at io n file s • M o r e m e m o r y is alw ay s u se fu l – R e d u c e p agin g – In c r e ase p age c ac h e size – L e ss I/O , b e t t e r p e r fo r m an c e ! 3
  • 19. En ligh t e n m e n t … o r “p ar av ir t u alisat io n ” … o r “c h an gin g t h e O S” • U se in fo r m at io n fr o m t h e O S t o gu id e sh ar in g – R e d u c e s t h e o v e r h e ad • A v o id p at h o lo gic al b e h av io u r – D o u b le p agin g – H o st d e m an d p agin g • Pr o o f-o f-c o n c e p t o n L in u x 4
  • 20. U SP 1 : d u p lic at e d e t e c t io n • Sat o r i m o n it o r s b lo c k r e ad s – H ash d at a w h e n r e ad , sh ar e p age s if id e n t ic al – B lo c k r e ad s fr o m c o m m o n su b st r at e d isk s • O t h e r sy st e m s sc an m e m o r y p e r io d ic ally – R at e lim it e d t o p r e v e n t o v e r h e ad – C an m iss sh o r t -liv e d sh ar in g o p p o r t u n it ie s 5
  • 21. U SP 2: sav in gs d ist r ib u t io n • W an t t o e n c o u r age sh ar in g – T h e m o r e y o u sh ar e t h e m o r e m e m o r y y o u ge t – Eac h V M ge t s a sharing entitlement b ase d o n h o w m an y p age s it sh ar e s • D e flat e t h e b allo o n t o in c r e ase allo w an c e 6
  • 22. U SP 3: fau lt h an d lin g • W h at h ap p e n s w h e n sh ar in g is b r o k e n ? – N e e d so m e m o r e m e m o r y – … b u t it ’s in u se – T y p ic ally u se h o st p agin g • M o d ify t h e O S t o n o m in at e “v o lat ile ” p age s – C le an b u ffe r c ac h e p age s – H y p e r v iso r c an r e assign t h e m w it h o u t w ar n in g – “R e p ay m e n t FIFO ” b ase d o n IB M ’s C M M 7
  • 23. Im p le m e n t at io n • H y p e r v iso r (5351 L O C ) – B ase d o n X e n 3.1 w it h PV sh ad o w PT s – A d d e d sh ar in g-r e lat e d h y p e r c alls • D om0 (3894 L O C ) – M o d ifie d b lk t ap t o p e r fo r m h ash lo o k u p • D omU (2306 L O C ) – Po r t e d IB M ’s C M M t o X e n /x 86 8
  • 24. O v e r h e ad s • W o r st -c ase : lar ge -c h u n k se q u e n t ial r e ad s – O v e r h e ad o f h ash in g: 0.2% – O v e r h e ad o f h ash in g + IPC : 34.8% • “R e alist ic ” b e n c h m ar k (k e r n e l c o m p ile ) – W it h o u t Sat o r i: 780 se c o n d s – Sat o r i e n ab le d : 779 se c o n d s 9
  • 25. Effe c t iv e n e ss Kernel Compilation (256MB) 16000 14000 12000 10000 Potential 8000 Satori Pages 6000 4000 2000 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 10 Time (mins)
  • 26. Effe c t iv e n e ss Kernel Compilation (512MB) 30000 25000 20000 Potential 15000 Satori Pages 10000 5000 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 11 Time (mins)
  • 27. Effe c t iv e n e ss httperf 120000 100000 80000 Potential 60000 Satori Pages 40000 20000 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 12 Time (mins)
  • 28. Effe c t iv e n e ss RUBiS 30000 25000 20000 Potential 15000 Satori Pages 10000 5000 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 13 Time (mins)
  • 29. C o n c lu sio n s • M o r e m e m o r y m e an s m o r e p e r fo r m an c e • Page c ac h e s ar e u se fu l so u r c e s o f sh ar in g • En ligh t e n in g t h e O S av o id s h o st p agin g • Sat o r i is e ffic ie n t an d e ffe c t iv e ! 14
  • 30. Q u e st io n s? • C lic k t o ad d an o u t lin e 15