6/2 HPDiscover

by Gene Kim on

#hpdiscover

  • Starting podcost @RafalLos and @xthestreams: this'll be fun: #devops!
  • My understanding it is that we're bro
  • adcasting live. @Raf
  • alLos will tweet out details
  • Gene asks @RafalLos: "so we all love #devops: why is everyone so mad at you?" haha. cc @joshcorman, @wickett
  • Giving kudos to @ngalbreath on podcast w/@RafalLos & @xthestreams

Day 2

  • Sitting w/@RafalLos here at #hpdiscover, waiting for Meg Whitman keynote. 10K attendees here.
    https://p.twimg.com/AuotqbhCQAE9CR4.jpg
  • Finished fun interview with @xthestreams, @patrickdebois on CIO issues and how #devops can help. More interviews coming!
  • @cebess: RT @HP: Watch live at 8:30AM PST: HP CEO Meg Whitman and special guest speaker Jeffrey Katzenberg http://t.co/ZwFETSvV #HPDiscover
  • RT @cebess/@HP: Watch live: 8:30AM PST: HP CEO Meg Whitman & special guest Jeffrey Katzenberg http://t.co/ZwFETSvV #HPDiscover
  • @gigastacey: “@qhardy: Sign of the Times: at #HPDiscover speeches, "Industry Bloggers" tables in front of "Press" tables.” >> I say it's silly to segment
  • RT @gigastacey/@qhardy: Sign of the Times: at #HPDiscover speeches, "Industry Bloggers" tables in front of "Press" tables”
  • OH: "For keynotes (Whitman, Chris Anderson/Wired), they're using 30 projectors, each costing $30K." Wow. $1MM of projectors
  • Haha. Cloud. @rbergin: @RealGeneKim the event company uses $1MM worth of projectors: HP is using Projector-as-a-Service, rt?”
  • Whitman: "A caution: what happens in Vegas doesn't actually stay in Vegas." Haha.
  • Whitman: "HP employees are supporting lines of communications for alongside UK troops in Afghanistan"
  • Whitman: "
  • Whitman: "All CEOs view technology as fundamental to success to orgs; needs to be as ubiquitous as electricity"
  • @Wh1t3Rabbit: Meg telling how #HP employees volunteer on front lines in Afghanistan, wear body armor maintaining comms for UK Defense forces. #HPDiscover
  • Whitman: "70% of HP revenue comes from infrastructure: storage, networking, hardware; we're not afraid of commoditization"
  • Whitman: "Bank of India case study: 234% ROI in 2 years"
  • .@MegWhitman: "
  • Whitman: "Fastest data growth is unstructured data (e.g., tweets), foot traffic in stores
  • @RichDunbar: Meg Whitman: "Phone, blogs, videos and text are the backbone of human information." #HPDiscover
  • RT @RichDunbar: Meg Whitman: "Phone, blogs, videos and text are the backbone of human information." #HPDiscover
  • @johnobeto: Meg: IT is currently focused on structured data, which is only 15% of all data. We need to be able to work on unstructured data
  • RT @johnobeto: Whitman: IT is curr focused on structured data: which is 15% of data. Next: unstructured data: fastest growing
  • Next up: Jeffrey Katzenberg, CEO, Dreamworks Animation
  • Katzenberg: Showing video: 10y ago from Lion King launch: brought out lion, misbehaving: "lion drool stains tough to get out"
  • Katzenberg: "I've known Meg Whitman since 1918.. Umm, no. 1980. We worked together at Walt Disney Corp"
  • Katzenberg: "Why should you all be listening to me, a 'cartoon guy?'" Showing video: animation: hand-drawn, then sound, color
  • Katzenberg: ...then digital awakening, digital revolution: Shrek 1st to win Oscar award: Kung Fu Panda, Monsters/Aliens...
  • Katzenberg: "Animation has come a long way from artists hand-drawn each frame; we use highest end of computing technology"
  • Katzenberg: "Dreamworks: 1994: Lion King; then did Shrek: 1st Academy Award, we were a startup; we viewed CG as future"
  • Katzenberg: "we completely retooled, retrained, transformed to digital animation house; now largest animation house: 16 films"
  • Katzenberg: "After seeing Polar Express 3D, I knew we had to xform again; it would 2x compute demands: one movie for each eye"
  • Katzenberg: "We are first house to release 100% of films in 3D. Most exciting development in my 40 yrs in movie industry"
  • Katzenberg: "3D helps us tell most immersive stories possible"
  • Katzenberg: "Each movie has 130K frames, each done by 12 diff depts: on avg, goes thru 10-100 iterations in each dept" (!!)
  • Katzenberg: "That's 3 billion iterations over 4 year project to get the finished product"
  • Katzenberg: "Artists must separately animate face and model, then hair, clothing; all different depts"
  • Katzenberg: "8 hours to animate 3 seconds of film, then critiqued/reviewed; cycles begins again. Insane"
  • Katzenberg: "It used to 3 animators for 5 months of work; now we can do it in real-time; game changer or waiting 8 hrs"
  • Katzenberg: "Isn't it interesting that me, as a non-technology
  • Katzenberg: "Our artists used to work at the speed of technology; now we can work at speed of their imagination"
  • Katzenberg: "Every business wants to be better, faster, cheaper; usually can only get 2 of those; we're getting all three"
  • Katzenberg: "This will liberate our artists & totally transform our biz & production process; 12 depts -> 5 or 6; terrifying"
  • Katzenberg: "@dealarchitect: #HPDiscover Katzenberg on complexity of modern animation. A good animator generates in week 3 seconds of single character in final product
  • RT @dealarchitect:Katzenberg: A good animator generates in week 3 seconds of single character in final product
  • Katzenberg: "LCD provides much worse colors than monitor: Shrek was in different color in every scene: we hoarded old CRTs"
  • Katzenberg: "The DreamColor LCD from HP exceeded for 1st time color monitors; we could finally let CRTs go"
  • @vk_is: CEODreamWorks - "each film is made up of 130k frames, 3B iterations. 3 secs of lead character tkes 1 wk of best desiger's time." #HPDiscover
  • Katzenberg: "5 years ago, before cloud was in vocabulary, we started using HP cloud
  • Katzenberg: "Our industry plagued by digital piracy: loss of file can result in movies being avail before movie release"
  • @DavesRant: Jeffrey Katzenberg Jeffrey's law, we'll always want more than Moore. Keynote day2 #HPDISCOVER RT @DavesRant: Jeffrey Katzenberg Jeffrey's Law: Businesses will always want more than Moore. Keynote day2 #HPDISCOVER
  • Katzenberg: "Each movie project has over a half million files"
  • @MaryHWells: RT @martinmacleod: #HPDiscover - Jeffrey Katzenberg - 'we were HP's first Cloud customer before cloud existed - HP provides 20% of our processing needs'
  • RT @MaryHWells: Katzenberg: 'we were HP's first Cloud customer before cloud existed - HP provides 20% of our processing needs'
  • Katzenberg: "Each character in Madagascar has 1000s of animation control points: 6TB data for 2.5m trailer"
  • Katzenberg: "
  • @ghaff: RT @JillatHP: Puss in Boots required more than 60 million rendering hours #HPDiscover #Dreamworks
  • RT @ghaff: RT @JillatHP: Puss in Boots required more than 60 million rendering hours #HPDiscover #Dreamworks
  • Katzenberg: Dreamworks: 10 films in production at any time, 5 years for each, 400 artists worldwide, 130K frames,
  • Katzenberg: Each frame has 100+ layers, half billion files per film, 200 TB data, 60MM rendering hours, all done w/HP
  • (There's going to be a screening of Madagascar 3 here tonight: sound setup here is amazing...)
  • Katzenberg: "I'd like to do a do-over of that terrifying moment 18 years ago." (Oh, no... He's bringing out a lion?! Haha.)
  • Katzenberg: Haha. Bringing out a guy in a lion suit. That's good risk management: "I won't get on stage w/real lion again"

TB2787 The journey from SOA governance to DevOps
Lando 4302

tim durgen,

  • War room became a fixed part of the job: says that we have the problem
  • New support model needed: applications, operations, infrastrucutre
    • Unum: apps and infrastructure: both have strategy, roadmap, design, governance, SDLC/IDLC: #devops
    • Unum: "monitoring is a team sport; everyone should have access to tools & have access to same data" #devops
    • Unum: "fail fast: establish SLAs for all environments; no promotion unless SLAs are met"
    • Unum: "establish SLAs for all environments (eg, dev/test/prod); if it fails dev SLA, it'll fail in every other env, too"
    • Unum: "we're continually frustrated how our tools focused on Dev or Ops; no
    • Unum: "a server without any code provides no value to the business"
  • Eliminate the monkeys
    • Unum: "we use SOA Center (systinet) to provide ALM; integrate w/BSM, Qual Ctr to enforce SLA perf & acceptable # of defects"
    • Unum: "we use SOA Center (systinet) to provide ALM; integrate it w/ Qual Ctr to enforce SLA perf & acceptable # of defects"
    • Unum: "goal of our automation is to eliminate meetings and reviews: make it objective, no humans needed"
    • Unum: "also integ SOA Center to Layer7 SSG to govern svc deployments: prevent unauth deployments w/o SLA verif" #devops
    • Unum: "we enforce retirement: upon deployments, people have 12 months before forcible retirement; we'll stop routing traffic"
    • Unum: "we own many, many monitoring tools: somewhere between 40-100: need NDA to find out exactly. :)"

TB2658: DevOps Pane: tighter collaboration between testing and monitoring

Bryon Colaizzi, Director of App Performance, Paychex

  • Bryon Colaizzi, Director of App Performance, Paychex talking abt #devops journey. Lots of enterprise DevOps succ stories here
  • Paychex: automation of provisioning dev/prod activities reduced by a factor of 3 #devops
  • Paychex #devops goals: reduce time to value and IT cycles to decrease financial commitments
  • Paychex #devops: collocated Dev & Ops together to create tribal knowledge/collaboration/trust
  • Paychex #devops: over 100 projects in flight at any time: challenge: how do you keep environments available at any time?
  • Paychex #devops: env creation cycle time: 6mo to 3mo to 21 days: complex 26 component environment

BNY Mellon: Erik Rogers, Managing Director, Enterprise Architecture: sharing their #devops story run out of shared svcs group

  • BNY Mellon: #devops success story: team grew 20%, but project load increased by 100% (!!)
  • BNY Mellon: #devops creating standardized procedures/monitoring among 80 people spanning 20 groups (Dev, QA, Production, etc)
  • BNY Mellon: creating centralized support group to house all production readiness, common processes, #devops onramp
  • BNY Mellon: #devops was combo of top-down decree and "people in trenches"
  • BNY Mellon: #devops allows us to create library of non-functional requirements
  • BNY Mellon: #devops challenges: getting everyone to use, trust and diagnose problems with the tools
  • BNY Mellon: challenge: "it's terrible when u'r on incident call & u only have 1 person who can use tool, even tho all trained"

Ron Sitnick, World Bank, Information Officer, Lead Perf Testing, Global Monitoring Teams: #devops story

  • IFC: #devops: all performance testing and global monitoring now run by one gruop: consistent tools, practices, expertise
  • IFC: #devops: all test code and scenario data is centralized, versioned, shared
  • IFC: #devops: production data aids in building load models, increasing test accuracy
  • IFC: #devops: have diagnostics on all Dev servers: Dev & Prod using same tool sets & technologies
  • IFC: #devops: "everyone sees the same data, across all environments and silos: this elim a lot of arguing/finger pointing"

  • OH: "We created 185 questions: how fast do you want it? as fast as production. to force performance that is testable"

  • OH: "what is performance requirement for this app?" "umm... 5 seconds..."

  • Q: "is #devops only for internet companies, not enterprises?" A: "no way. Enterprises need #devops"

  • "I do believe that the larger and more complex the enterprise, #devops is more difficult"

  • "For years, monitoring is often perceived as the automated finger-pointing mechanism. No one trusted the tools."

  • "Our production monitoring tools allow us to show what functionality is never used, so we can cut the code out" Nice. #devops

  • IFC #devops: created performance metric (sum of resp times), graphed over time: biz execs loved it, seeing improving perf

  • "seeing performance over time allows you to catch defects much earlier, before catastrophic event"

  • Q: "how do you get Dev to work perf issues early?" A: "It's challenging. We put perf test into the continuous testing suit"

  • ..."we do load testing, endurance testing, try to find memory leaks for every piece of code checked in"

  • ..."these automated tests is how we get these perf issues found and fixed early in the Dev cycle."

    • Why? Is enterprises too big? It's about aligning priorities,