Slide 1

Slide 1 text

Adrian Colyer | Accel Partners @adriancolyer The Prepared Mind

Slide 2

Slide 2 text

Information overload

Slide 3

Slide 3 text

You can never bridge the gap.. Credit: Simon Wardley

Slide 4

Slide 4 text

Louis Pasteur, 1854 “Chance favours the prepared mind” "Louis Pasteur". Licensed under Public Domain via Wikimedia Commons - http://commons.wikimedia.org/wiki/File:Louis_Pasteur.j pg#/media/File:Louis_Pasteur.jpg

Slide 5

Slide 5 text

Fundamentals & Frontiers

Slide 6

Slide 6 text

No content

Slide 7

Slide 7 text

Becoming Finer-Grained

Slide 8

Slide 8 text

Finer-grained • Lifecycles • Deployment units • Organisations • Data partitions • Supporting Runtimes

Slide 9

Slide 9 text

openmirage.org

Slide 10

Slide 10 text

“Brutally hard and painful” - Asim Aslam • On designing and deploying internet scale services - Hamilton '07 • Life Beyond Distributed Transactions – Helland '07 • Building on Quicksand - Helland & Campbell '09 • Data on the Inside vs Data on the Outside – Helland '09 • Immutability Changes Everything – Helland '15 Towards Systems as Organisms

Slide 11

Slide 11 text

Distributed Snapshots • Chandy-Lamport '85

Slide 12

Slide 12 text

Thinking about Data • Goes around comes around – Stonebraker & Hellerstein '05 • Consistency, Availability, and Convergence – Mahajan '14 • The Declarative Imperative – Hellerstein '10 • Consistency analysis in Bloom, a CALM and collected approach – Alvaro et al. '11

Slide 13

Slide 13 text

Exploiting application invariants • Coordination avoidance in database systems – Bailis et al. '14 25x !

Slide 14

Slide 14 text

News you can use... $ 30% + 30% 32x !

Slide 15

Slide 15 text

Reduce Storage costs by 30%

Slide 16

Slide 16 text

Improve throughput by 30% • Cross-layer scheduling in cloud systems – Alkaff et al. '15

Slide 17

Slide 17 text

Do 32x less work! • ApproxHadoop: Bringing Approximations to MapReduce Frameworks – Goiri et al. '15

Slide 18

Slide 18 text

No content

Slide 19

Slide 19 text

x:y:z

Slide 20

Slide 20 text

mem:stg:net

Slide 21

Slide 21 text

txt:voice:img

Slide 22

Slide 22 text

No content

Slide 23

Slide 23 text

Fundamentals & Frontiers

Slide 24

Slide 24 text

the morning paper Papers announced on twitter: @adriancolyer Blog at: http://blog.acolyer.org Or subscribe to mailing list.

Slide 25

Slide 25 text

References • “The Tail at Scale” (micro-partitioning) http://blog.acolyer.org/2015/01/15/the-tail-at-scale/ • “Unikernels: Library operating systems for the cloud” http://blog.acolyer.org/2015/01/13/unikernels-library-operating-systems- for-the-cloud/ • “OSv – Optimizing the Operating System for Virtual Machines”, Kivity et al. 2014 • “On Designing and Deploying Internet Scale Services” - Hamilton 2007 • “Life Beyond Distributed Transactions” http://blog.acolyer.org/2014/11/20/life-beyond-distributed-transactions/ • “Building on Quicksand” http://blog.acolyer.org/2015/03/23/building-on-quicksand/

Slide 26

Slide 26 text

References • “Data on the outside vs Data on the inside” - Helland 2009 • “Immutability changes Everything” - Helland 2015 • “Distributed Snapshots: Determining Global States of Distributed Systems” http://blog.acolyer.org/?p=796 • “Goes Around Comes Around” - Stonebraker & Hellerstein 2005 • “The Declarative Imperative” http://blog.acolyer.org/2014/11/13/the-declarative-imperative-experience s-and-conjectures-in-distributed-logic/ • “Consistency Analysis in Bloom: A CALM and collected approach” http://blog.acolyer.org/2015/03/16/consistency-analysis-in-bloom-a-cal m-and-collected-approach/

Slide 27

Slide 27 text

References • “Consistency, Availability, and Convergence” http://blog.acolyer.org/2015/03/17/consistency-availability-and-converg ence-cops/ • “Coordination Avoidance in Database Systems” http://blog.acolyer.org/2015/03/19/coordination-avoidance-in-database- systems/ • “f4: Facebook’s warm blob storage system” http://blog.acolyer.org/2014/12/16/f4-facebooks-warm-blob-storage-sys tem/ • “A Hitch-hikers guide to fast and efficient data reconstruction in erasure coded data centers” http://blog.acolyer.org/2014/12/17/a-hitchhikers-guide-to-fast-and-effic ient-data-reconstruction-in-erasure-coded-data-centers/

Slide 28

Slide 28 text

References • “Cross-layer scheduling in Cloud Systems” http://blog.acolyer.org/2015/04/15/cross-layer-scheduling-in-cloud-syst ems/ • “ApproxHadoop: Bringing Approximations to MapReduce Frameworks” http://blog.acolyer.org/2015/04/16/approxhadoop-bringing-approximati ons-to-mapreduce-frameworks/ • “BlinkDB: Queries with bounded errors and bounded response times on very large datasets” - Agarwal et al. 2013 • “Mojim: A Reliable and Highly-Available Non-Volatile Memory System” http://blog.acolyer.org/2015/04/14/mojim-a-reliable-and-highly-availabl e-non-volatile-memory-system/

Slide 29

Slide 29 text

References • “Sirius: An open end-to-end voice and vision personal assistant and implications for future warehouse-scale computers” http://blog.acolyer.org/2015/04/13/sirius-an-open-end-to-end-voice-and- vision-personal-assistant-and-its-implications-for-future-warehouse-sca le-computers/

Slide 30

Slide 30 text

Additional Image Credits • Dust cloud - http://commons.wikimedia.org/wiki/File%3ARAF_Merlin_Helicopter_Creates_a_'Brownout'_Dust_Cloud_ Landing_in_Afghanistan_MOD_45153504.jpg • "Stonehenge Closeup". Licensed under Public Domain via Wikimedia Commons - http://commons.wikimedia.org/wiki/File:Stonehenge_Closeup.jpg#/media/File:Stonehenge_Closeup.jpg • "Stapel bakstenen - Pile of bricks 2005 Fruggo". Licensed under CC BY 1.0 via Wikimedia Commons - http://commons.wikimedia.org/wiki/File:Stapel_bakstenen_- _Pile_of_bricks_2005_Fruggo.jpg#/media/File:Stapel_bakstenen_-_Pile_of_bricks_2005_Fruggo.jpg