The future of DevOps Sasha Rosenbaum @DivineOps
Slide 1
Slide 2
Israeli Air Force Defense Industry R&D Cloud Consulting Microsoft GitHub DevOpsDays Chicago since 2014 Sasha Rosenbaum Red Hat @DivineOps
Slide 3
How about you?
Slide 4
Slide 5
The Past
Slide 6
Technology
Slide 7
1990s: Getting a new server for an application: 2-3 months
Slide 8
Backup
Slide 9
Slide 10
Date Release name 1990 SQL Server 1.1 (16-bit) 1992 SQL Server 4.2A 1993 SQL Server 4.21a 1995 SQL Server 6.0 1996 SQL Server 6.5 1998 SQL Server 7.0 2000 SQL Server 2000 2003 SQL Server 2000 64-bit 2005 SQL Server 2005 2008 SQL Server 2008 2010 Azure SQL database Software release cadence: 2-3-year cycle
Slide 11
Merge hell Merging the development branches and completing the test procedures could take months
Slide 12
Maintenance windows
Slide 13
Expected Availability < 99% 3.65 days / year
Slide 14
Unavailable systems were estimated to have cost American businesses $4.54 billion in 1996. Source: IBM Global Services, Improving systems availability, 1998.
Slide 15
Culture
Slide 16
Slide 17
Traditional IT dev ops wall of confusion
Slide 18
Speed Reliability
Slide 19
The problem isn’t technical. The problem isn’t people. The problem is socio-technical.
Slide 20
Darmok and Jalad at Tanagra
Slide 21
Patrick and Andrew at Agile TO 2008
Slide 22
10 deploys per day: Dev and Ops collaboration at Flickr Velocity 09: John Allspaw and Paul Hammond
Slide 23
Agile Infrastructure Velocity 09: Andrew Clay Shafer
Slide 24
Slide 25
DevOpsDays Ghent 2009: Patrick Debois
Slide 26
Slide 27
Speed Reliability
Slide 28
Charity Majors, CEO of Honeycomb
Slide 29
Slide 30
Nicole Forsgren. State of DevOps Report 2019
Slide 31
Speed Reliability
Slide 32
Software delivery is like a muscle. The more you use it, the stronger it gets. 33
Slide 33
Software Services Needs to be Operated Platform Services Needs to be Operated Infrastructure Services Needs to be Operated 34
Slide 34
In the beginning…
Slide 35
Deployment Checklists
Slide 36
Scripts
Slide 37
OS-level APIs
Slide 38
Slide 39
PowerShell (Windows) configuration management framework and scripting language Jeffrey Snover, 2006
Slide 40
Source Control for Ops
Slide 41
GitHub launch: 2008
Slide 42
Distributed version control + Pull Request system = Global collaboration
Slide 43
Infrastructure-level APIs
Slide 44
Amazon Web Services: 2002 Amazon Cloud Computing: 2006 45
Slide 45
Darwinian Pressure The new models evolved due to pressure to deliver adaptable services at scale. 46
Slide 46
Netflix, Amazon, Google, and every ‘cloud native’ company built a platform Because they had to… 47
Slide 47
‘Cloud’ evolved from lessons learned building and operating these internal services 48
Slide 48
Infrastructure as code
Slide 49
Configuration management minimizes manual toil and infrastructure configuration drift
Slide 50
“The traditional model is that you take your software to the wall that separates development and operations, and throw it over and then forget about it. Not at Amazon. You build it, you run it. This brings developers into contact with the day-to-day operation of their software. It also brings them into day-to-day contact with the customer. This customer feedback loop is essential for improving the quality of the service.” –Werner Vogels, CTO Amazon 2006
Slide 51
Jez Humble and Dave Farley: 2010
Slide 52
Continuous Integration (CI) The process of automating the build and testing of code every time a team member commits changes to version control.
Slide 53
Continuous Delivery (CD) The approach in which teams produce software in short cycles, ensuring that the software can be reliably released at any time.
Slide 54
Software Services Needs to be Operated Platform Services Needs to be Operated Infrastructure Services Needs to be Operated 55
Slide 55
Slide 56
You will automate me out of a job!
Slide 57
Toil Toil is the kind of work that tends to be manual, repetitive, automatable, tactical, devoid of enduring value, and that scales linearly as a service grows. 58
Slide 58
We would not be able to achieve the availability, reliability and speed we have today without automation
Slide 59
Slide 60
The problem isn’t technical. The problem isn’t people. The problem is socio-technical.
Slide 61
The Present
Slide 62
The future is already here. It’s just not evenly distributed ~ William Gibson
Slide 63
DevOps across Microsoft http://aka.ms/DevOps-Stories 105K 4.4M 5M 2M 500M 500K Engineers use the DevOps platform Git commits per month Builds per month Test executions per day Work items viewed per day Work items updated per day Shared with permission from Microsoft. Internal data snapshot some time in 2019 85,000 Deployments per day
Slide 64
Slide 65
Slide 66
DevOps and SRE engineers command a higher salary
Slide 67
We’ve created new jobs ¯_(ツ)_/¯
Slide 68
The new jobs got people higher salaries and more interesting work! ヽ(•‿•)ノ
Slide 69
Slide 70
We’ve created new disciplines!
Slide 71
SRE
Slide 72
SRE ≃ Google’s DevOps implementation
Slide 73
100% reliability is unattainable
Slide 74
Availability 99.999% 5.26 mins / year
Slide 75
How much does that cost?
Slide 76
Risk and Error Budgets Error Budgets An acceptable level of unreliability It’s a budget. It can be allocated.
Slide 77
SLI, SLO, and SLA Indicators Service Level Terminology Describe the metrics that matter, the values we want, and how we will react Defined measurement of an aspect of a service. Objectives Target value (or range of values) as measured by an SLI Agreements Explicit or implicit contract with users or customers, with consequences of meeting or missing objectives
Slide 78
Monitoring
Slide 79
“Monitoring is how you manage your knownunknowns, which involves checking values for predefined thresholds, creating actionable alerts and runbooks and so forth.” Charity Majors, CEO, Honeycomb
Slide 80
Without monitoring, you have no way to tell whether the service is even working
Slide 81
Observability
Slide 82
“Observability is how you handle unknown-unknowns, by instrumenting your code and capturing the right level of detail that lets you answer any question …” Charity Majors, CEO, Honeycomb
Slide 83
All of this requires collecting and analyzing massive amounts of data
Slide 84
”If we have data, let’s look at data. If all we have are opinions, let’s go with mine” - Jim Barksdale
Slide 85
Infrastructure as code ?
Slide 86
MLOps
Slide 87
MLOps •Massive amounts of data •Data model versioning •Model re-use •Model decay over time •Compliance considerations
Slide 88
Chaos Engineering
Slide 89
Chaos engineering The discipline of experimenting on a software system in production in order to build confidence in the system’s capability to withstand turbulent and unexpected conditions.
Slide 90
Slide 91
Everybody tests in production
Slide 92
Incident Response
Slide 93
Blameless postmortems
Slide 94
There is no root cause
Slide 95
DevSecOps
Slide 96
More code = more problems
Slide 97
Cyber attacks are at all time high
Slide 98
Security must be an integral part of the software development lifecycle
Slide 99
There is so much more…
Slide 100
The industry has evolved
Slide 101
Open Source
Slide 102
Open source is defining the new industry standards 1M+ projects 100M+ repositories Source: https://github.com. August, 2019. 40M developers 2.1M businesses
Slide 103
90% of IT leaders are using enterprise open source. 90% Source: Red Hat State of Open Source Report 2021
Slide 104
Cloud
Slide 105
Cloud Numbers Public Cloud ● IaaS - Dominated by 6-8 Clouds ● PaaS - Dominated by 50-100 Clouds ● SaaS - Over 4000 SaaS offerings Data Center | Private Cloud ● “Only 20% is in the Public Cloud” - IBM ● “Only 5% is in the Public Cloud “ - AWS 2020 Worldwide Public Cloud Revenues ~$235B USD 2020 Worldwide Data Center Revenues $2-4T USD
Slide 106
Kubernetes
Slide 107
85% of global IT leaders agree that Kubernetes is key to cloud-native application strategies Source: Red Hat State of Open Source Report 2021 Source: Red Hat State of Open Source Report 2021
Slide 108
The Future
Slide 109
Slide 110
The future is already here. It’s just not evenly distributed ~ William Gibson
Slide 111
seeking advantage seeking legitimacy
Slide 112
Agile DevOps SRE
Slide 113
“DevOps is a solved problem” - Someone from Google, 2019
Slide 114
Source: StackOverflow Developer Survey, 2018
Slide 115
Source: StackOverflow Developer Survey, 2018
Slide 116
If knowledge was all it took, we’d all have six pack abs.
Slide 117
We must spend time on making sure that the “standard of living” improves for everyone
Slide 118
Companies, just like people, don’t like to change
Slide 119
“Smart people don’t learn … because they have too much invested in proving what they know and avoiding being seen as not knowing.” - Chris Argyris
Slide 120
Learning requires vulnerability
Slide 121
The age of continuous updates
Slide 122
Slide 123
“In this era of becoming, everyone becomes a newbie. Worse, we will be newbies forever.” - Kevin Kelly
Slide 124
More innovation in the new DevOps disciplines
Slide 125
Managed Services
Slide 126
Spend More Time On What Matters The Cloud Native Organization DEV ARCH OPS TRADITIONAL IT CLOUD NATIVE “Artisanal Projects” “Industrial Products” ANCHORED UNINSPIRED APPS PROTECTIVE GATEKEEPING PER PROJECT INFRASTRUCTURE SELF SERVICE CREATION BRITTLE DEPLOYMENTS TOIL DRIVEN OPERATIONS ENABLING CONSTRAINTS COLLAPSE COMPLEXITY UNCHAINED DIFFERENTIATED VALUE RESPONSIVE INNOVATION PLATFORM SERVICES UBIQUITOUS AUTOMATION STANDARDIZE INFRA
Slide 127
“This is one of the innovator’s dilemmas: Blindly following the maxim that good managers should keep close to their customers can sometimes be a fatal mistake.” - Clayton Christensen
Slide 128
The DevOps evolution continues, as we solve new problems every day
Slide 129
Good DevOps copy Great DevOps steal
Slide 130
Thank you! @DivineOps
