DevOps Theory vs. Practice: A Song of Ice and Tire Fire @ DevOps Days Charlotte 2019

00
well first of good afternoon everyone or technically so good morning and I I was
0:07
actually thinking of doing this talk alone and then I decided that who better to to join me than a thought leader a
0:16
disruptive innovator the senior sre leader at Google the senior software
0:23
engineer at Netflix the SVP of thoughts at Facebook and the guy that’s obviously
0:30
better than all of us combined ladies and gentlemen I want to welcome Luke Starsky thank you so then as you’re
0:44
obviously wondering who I am let me introduce myself quickly as well I’m an official hip tech translator so
0:50
what I do is I have a native proficiency and Dutch English and go language and I
0:55
dub it in German French and I curse in about 18 more languages which may or may
1:01
not be true and then obviously I am fluent in thought-leader gibberish so
1:07
what we’re gonna do is I have asked Baruch to to come up with with some very
1:13
innovative every thought leader ish comments for us today so Baruch thought leader away thank you
1:20
very much Liam so I’m going to spill some wisdom of you on you as as you come
1:27
on stage you are not so obviously I am better than you just by that fact so you will learn a lot today
1:32
let’s do it because the first wisdom is everybody software must be reasonable
1:39
and absolutely any time the time from your genius idea to the market this is
1:47
the key for your next startup success so it actually means by that is that
1:53
whatever he comes up with has to go into production as soon as possible
1:58
preferably before he even thought about it everyone must have 1/2 percent test
2:05
automation this is extremely important qualities the name of the game and how do we achieve it by automating all our
2:13
Quality Assurance processes so that’s an interesting one I mean we all have that app that is being used by like five
2:20
people a day and we update it maybe once every six months if we’re all lucky so
2:26
let’s spend the next 6 months making sure that everything is automated because that makes total sense no
2:33
continuous security well you heard a lot today about deficit cops and how you actually should do it and that’s the
2:40
right way you need to bake security on every stage of your pipeline you need to make sure
2:47
that everything you do is 100% secure and reliable and well he’s right but he
2:54
actually means to say is that we hired a guy just now to go around and and
2:59
validate all of our software check all of the dependencies that we have and then complain about the fact that there
3:06
are so many gaping holes in there and then obviously we fired him because he complain too much
3:12
your rate of threat is an outage not an employee you need to make sure that you
3:17
don’t have five nines you have seven nine nine nine twelve nine of of uptime
3:24
and your services are never down not even for a second good one I like that I
3:32
mean it sounds like a great marketing slogan but what you’re actually trying to say there audience is that you want
3:38
to spend the next 10 12 months figuring out how to detect and prevent that five-second outage and obviously don’t
3:45
spend too much time figuring out why that drunk employee left their laptop with five billion passwords and social
3:52
security numbers inside of a cab because that is not a threat you you hire the
3:58
best and trust them right that’s very okay now the biggest problem we have with with DevOps adoption are
4:07
virtual machines basically if you have kubernetes you have DevOps
4:13
if you have virtual machines you don’t have and cannot have DevOps that’s a
4:20
pretty bold statement coming from from Netflix Google and and Facebook so what
4:26
you’re what you’re trying to say is that we should absolutely take all of the efforts that we’ve done over the last
4:32
5-10 years throw it away start from scratch by container izing everything
4:37
makes perfect sense kind of does all right you are a beautiful unique
4:43
snowflake as are your problems no vendor could possibly understand them because
4:50
if someone that works for the vendor could possibly understand your problems
4:56
who probably will work for us by now so you’re right I mean vendors probably
5:03
don’t understand you but the thing is we we all build software to help you solve
5:08
your problems and your problems might not be as unique as you think they are
5:14
the most important you need to go for a talent and the talent exists in one
5:21
place only on this planet and that San Francisco which is obviously a very
5:29
terrifying thing to say well we’re standing here in Charlotte and obviously thinking about the fact that literally
5:35
every disaster movie that Hollywood has produced so far has a scene of San Francisco being destroyed by whatever
5:42
monster earthquake or terrifying threat there is not to mention the fact that
5:47
San Francisco is so incredibly expensive that yeah we should spend all of our money renting a top floor suite in the
5:55
highest building that exists so this is all from us thank you very much thank
6:01
you I hope you enjoyed our talk they are not sure yeah not really obviously yeah
6:13
so my name is burrow I’m a cheap sticker officer at a frog that’s my real title I
6:19
do have a business card and that what makes it real I also ahead of developer relations at
6:25
Joe frog so Joe frog does awesome t-shirts as you can see and I well I
6:31
demonstrate the front you can demonstrate the back of them search back absolutely awesome if you still didn’t grab one we have a
6:37
table right out this door so don’t forget to stop by we also do awesome
6:43
software that will also talk about on our table this is why his name is Leon
6:53
yeah I decided that you know the first time I’m gonna be a fob sighs Charlotte I probably shouldn’t lie about my name
7:00
that that wouldn’t be such a good idea so I’m a developer advocate that hei frog and I hope to make developers
7:07
better yeah so and as you already figured out we are extremely entertaining guys so you should follow
7:13
us on Twitter right now in case you are not sure and you need couple of more
7:18
slides our twitter handles are on a bell on the bottom of every slide so you
7:23
won’t forget very important actually the only slide that you need to take a
7:29
picture of is this one so we prepared a special page for you it’s a Java becomes
7:35
the show notes there is a dedicated landing page for this talk this slides are already there we are recording this
7:41
talk and the video will be out by tomorrow if I didn’t screw up with the with the a/v all the links everything we
7:48
are going to talk about is already there comments rating you can rave how great we’re right there and raffle just to
7:55
thank you for coming here and sitting here with us so I I don’t see Cory is
8:01
Cory here no okay so I wanted to thank him he’s not here screw this guy alright
8:10
so what you saw our little play is is obviously exaggerated but I think you
8:17
can a little bit relate to that when you go to conferences and you and you see all those wonderful people from those
8:25
wonderful unicorns speak about their great technology it’s kind of itching
8:31
is it really that way should we do what they do how horrible is it what we are
8:39
doing comparing to them and and to understand this phenomena we need to go
8:45
back in time you know before we do that can ask aliens a question how many of you have have been to a chat conference
8:51
like this and walked away thinking this is such awesome technology I totally need this tomorrow
8:58
now are we honest definitely me okay good yep so the vast majority that’s nice yep and to understand this
9:05
phenomenon how we how we got here we need to go back in time almost 100 years
9:10
to the late thirties the early forties of of the 20th century a to the Pacific
9:17
Ocean and there are a lot of islands there small ones there were there people
9:24
are living on those islands a hundred years ago there weren’t really advanced
9:29
and they were leaving their lives and then some strange things started happening anyone but strange things
9:37
started happening yeah I know I already gave it away okay yeah nevermind so airfields air drops supplies
9:44
buildings people money suddenly start to appear and then they kind of stopped
9:57
right because the war was over those Islands lost of their critical role in
10:05
in this universe and what does they do next where can they find about what
10:13
actually happened oh I know yeah I was gonna say hacker news good judge swear I
10:19
still have to do this is what we do right when we see in Marvel we go to hacker news I read about it and we do
10:28
what they did and we really hope that we’ll have the same marvelous effect
10:34
this is what the Islanders did they built those airplanes they built those
10:42
airfields they build those churches and then they waited for supplies to arrive
10:48
for planes to land for crates to drop this gentleman is known as the cargo
10:57
cult sorry we were asking you on yeah I tried yeah Haruko you obviously heard
11:04
the name now you know where it comes from the cargo is actually this cargo that they expected to get after they
11:12
build those Hey airplanes and and this and what happens in our conference scene
11:20
right there is a lot of cargo called out there because you see those people and
11:26
you kind of wow this is so cool and you go and you’ll build your high airplane
11:32
and you wait for the cargo that will probably never right how do we go about
11:42
it how do we battle they circled that’s that’s a good question and actually
11:47
technically they’re for questions for questions that that will help you
11:52
realize whether or not the thing that you’re looking at is more on the cargo called sighted things or as actually
11:58
that that unique thing that’s gonna help you trend and and you know be the best organization that you can be first one
12:04
and I think that’s one of the most important questions is my organization or is my team actually ready to adopt
12:11
that new technology are we are we fundamentally ready to do DevOps are we fundamentally ready to do agile are we
12:18
ready to go full-blown in to kubernetes the same question is is it even a good
12:24
tap right so there is someone on stage praised about some frameworks of the
12:30
road and they claim that it’s amazing they have those shiny slides is it
12:36
really amazing or it’s some new technology that they just invented no one ever uses it and it sucks good baby
12:43
so third question is what do I gain from adopting this technology I mean as we spoke about in in the
12:49
beginning there’s a lot of stuff that you might not have to do simply because the gains that you would come from
12:55
adopting it aren’t gonna propel you forward and you know what even if you gain from that technologically your
13:02
uptime improved from five nines to six nines your time from idea to release
13:09
moved from one hour to a half an hour is it even beneficial for the business
13:15
the investment that you are going to put into this adopting this technology is
13:20
really obviously you know what maybe containers have a smaller print
13:26
footprint that virtual machines is it really relevant to my company
13:32
business-wise to do this transition now will will be able to return our investment now the problem with the
13:40
later that the bottom two questions is that they are extremely context
13:47
sensitive right it depends on your technology that you are using today and
13:52
it depends on obviously your business and your position in the market your windows of opportunity and what’s not
13:59
but the first two we can actually talk about so let’s go and try to talk about
14:07
how we go about answering the first two questions so is my organization or team
14:14
is it ready to adopt new technology and obviously in order to to fundamentally assess that there is there’s a specific
14:22
tool and that too is called a maturity model so Martin Fowler I think has one
14:30
of the best hands down probably the best the definition of what a maturity model
14:36
is and when Georgia model is a tool that helps people assess their current effectiveness right of a person or a
14:42
group whoa that was very loud yeah that was also
14:48
loud can we kill the sound from I guess that helped when they do that to
14:54
the AV guys it helps all right and and basically yeah assess the effectiveness
15:00
and kind of where you are the problem with maturity models people hate them I
15:07
don’t know how many we try to do maturity model how many of you liked it
15:13
yeah exactly I I actually liked it yeah you are a process geek and you are a
15:18
unique snowflake thank you and I enjoy it very much because I hate process right so let’s talk about what maturity
15:26
model actually means well it’s built from so obviously they’re gonna be a
15:32
bunch of evaluation factor so the things that that are important to me why why do I want to introduce or look at new
15:39
technology to begin with second of is Korres scoring methodology because you know those evaluation factors don’t make
15:45
a lot of sense if you don’t think about how are we going to score them self assessment versus sir power to
15:51
assessment so we can think that we are great but actually we are not so much so we need this external view on how good
15:59
we are and obviously progress tracking because the maturity model is nice but
16:04
if you only use it like once then but you’re you actually measuring and maybe
16:10
the most important out there is visualization that’s critical because we want to visualize our process not only
16:18
for us but for everybody else and if you have like Excel of 2000 lines that’s
16:24
probably the information and out there but no one probably is going to ever look at it if you have a nice graph
16:30
that’s something else so now we’re going to show you a very innovative very
16:37
complicated and very groundbreaking tool for building maturity models right all
16:44
our examples are done in this tool and I’m not sure how many of you are ready
16:50
to adopt it it’s called Microsoft Excel
16:56
everything is Microsoft Excel um this is a top-level view we have two dots on it
17:06
first of all we have two axes we have the tools and process that’s kind of the
17:12
combination of what we need like for DevOps stuff and and and then we have
17:18
dots first of all initial initial is red we are like pretty good at tools
17:23
pretty good in process we are not completely awesome but not bad at all
17:28
and then there is one updated which is kind of your not as good as you think
17:33
what is this update it comes from so that is obviously talking to other teams
17:39
I mean other people within your organization if you do a video and assessment then everyone who’s ever done
17:46
an assessment of themselves including me I have to admit that will always square themselves better than they actually are
17:52
so asking other people to rate you on those questions or rate your team on how
17:58
well you are in process how well you do with tools that comes up with the the yellow balls right so that’s that’s the
18:05
self assessment versatile party another very important aspect is progress playing so this is where or today this
18:11
is where we want to be in a neck squalor this where we want to be in your target
18:17
range whatever it is a year two years or or whatever another very important aspect is a healthy competition right so
18:25
you have your team see you are kind of here and the other teams are elsewhere
18:31
and if anyone’s wondering I’m on Team B yeah so Leone leads the team B this is
18:36
why they are so awesome and everybody wants to be team B obviously I am of the chart there or maybe team B stands for
18:46
bar oh yeah I assumed you were gonna say that right alright so let’s see how those
18:52
dots appear obviously we are not like you know putting pins on honor on a
18:57
whiteboard there is some scoring methodology that ends up in those dots and the scoring methodology is actually
19:05
breaking down all the tools that you have or want to have and what they do and all the processes that you have or want to have
19:11
and how they work and then score yourself on this range so here’s an
19:16
example we’re talking about DevOps section on demand releases on demand
19:22
releases it’s kind of a group of tools and techniques we are talking about tool and the tool that we are assessing is
19:28
something that allows your bills to be configured in publish and consume
19:33
artifacts from artifact management system in a consumable format and there are some tools one of them actually
19:40
rhymes 2:02 facts and color of the factory but there are others as well obviously yeah and then you go and you
19:47
break down all the requirements I’ll deduct and be published in controlled environment here’s no partial artifacts
19:53
are published in standard consumable format for example maven docker registry etc yes or no partial etc etc you do
20:01
that for tools and you do that for process so for example here on the the
20:07
on demand releases we have the process build artifacts that are released to customers and we forgot a space there
20:14
are managed and governed things like artifacts paths all the necessary
20:19
quality checks and test prior to promotion to release which is obviously very important and yes we do that and
20:26
then you go like a yes/no or partial yeah you got that it there is a different approach which has some other
20:33
benefits and that’s listing those categories and then having all those yes
20:39
no partial summarized to very precise a number so here we have an example of
20:46
those categories and there are obviously the same categories so everything as a
20:52
code should have we have criticality must-haves should have or nice to have
20:57
benchmark what good looks like and here’s an interesting thing not all of
21:03
them and you know what only couple of them are 100% the majority of them are
21:09
not 100% this is what good look like and eighty-three how did I 183 stands for
21:17
yeah I have no clue I mean I like I know that’s that’s good at you asked so eighty-three is the result of those yes
21:26
know partially breakdowns so you have what you want to achieve and most of them will be probably yes some of them
21:33
will be partially and the total result won’t be 100 but will be 83 sometimes it
21:39
has to be 100 FedRAMP elements that’s the government regulation either you
21:45
have now all of it or it doesn’t really count you have to have an armed person next one is what we have today and you
21:52
see how grim the picture is that’s not very advanced team right so 83 was
21:59
architectural alignment 32 the next one is 24 months from now and we have
22:05
another spelling mistake here and 24 months from now is where we want to go sometimes we’re going to go all the way
22:11
so for example here I have this FedRAMP now FedRAMP is not a good example
22:17
sometimes you want to heat what you actually benchmark but most of the time you don’t most of the time you say in
22:24
two years from now if I get like close to the benchmark I’m good or if I
22:29
improved I’m good Fred ROM here is an interesting beast the benchmark has to be 100% because
22:36
either you have it or not but since it is nice to have I don’t
22:41
have anything today but I also don’t expect anything in two years because business decided that
22:48
probably we have a law lower fruits to a to collect we have slower fish to catch
22:58
and the government market it’s not something that is our priority so we
23:03
have this venom here we have 100 percent is the only benchmark that make sense but we are not going to invest in it
23:10
anything in two years so it’s good that you mentioned that the business said
23:15
criticality is nice to have because obviously for some of the others other teams will decide what the criticality
23:21
is so you want to account for the priorities of different teams I mean you want to
23:26
for the engineering perspective you want to account for ops you want to account for the company or business perspective
23:33
so for example if you look at whatever gonna pick the continuous delivery of
23:38
product in that case for engineering that might not be as relevant because
23:44
they do that anyway they have to be sure that that they can do that for ops it’s
23:49
a must-have they have to be able to take the software that the engineering teams produce move it into production at any
23:57
given time of the day and for for business perspective it might not be as
24:02
relevant so they kind of kill but don’t really if the engineering folks can
24:08
deliver their process that there are tools on horses they didn’t build horses
24:14
as long as it’s delivered on time and good quality they don’t really relevant right and then you do the combination
24:19
and all three of them and you come with this conclusion of the criticality model
24:26
definition is very important what how do we grade this yes/no partial and here we
24:33
do a breakdown of very detailed what differentiating means through the roof
24:38
what complete means what partial most partial match partial sum or no support
24:43
means and then we know how out of those one two three four five six grades
24:49
actually come up with the number of where we stand today and where we should
24:56
actually move right and we want to finish this section with a couple of
25:02
do’s and don’ts this is very important Leon green is good red is bad
25:08
yellow is okay okay don’t use other colors don’t do Jenkins don’t I’ll try
25:16
anyone got the joke the goal they were kinda nicer yeah no not really okay involve our teams in model definition
25:24
this what we want to measure and what good looks like the engineers on the team have to say
25:32
what what it is and obviously the res the business and everything but also the teams yep so let teams self assess first
25:39
I mean give your team’s to chance to figure out where it where they think they stand and then assess together so
25:45
help each other figure out where the blind spots are figure out how you can help the rest of the rest of your
25:51
organization assess themselves as well remember the B team that stands for Baro you should start with them because they
25:59
are the most successful and their position up there on the top right
26:06
corner will actually generate the friction and the activation energy that
26:12
you need to put everything in motion so M remember that’s probably the most
26:17
important thing maybe a perfect primary colors but remember that being at 100% is not the goal of a model I mean it has
26:25
to be a stretch goal but it isn’t the goal to get everything to 100% because it might not even matter to get 100%
26:31
everywhere right and the other thing the model should should involve from time to time you your priorities changed the
26:39
technology changes process changes and you need to incorporate that in the model right so for example stuff like
26:46
DevOps that happened five years ago probably affected maturity model of a lot of
26:51
people by introducing different processes and different tools and and everything else the most important
27:00
outcome of this part I would say is to remember that maturity model is a
27:05
process if you are someone or level 2 you cannot jump to level 5 you need to
27:11
go to level 3 and then to level 4 and only that you can dream of level 5 that
27:17
goes back to the cargo cult color code is being on level 2 coming to a
27:24
conference hearing from someone of level 5 and saying well I’ll go and implement it this Monday morning not gonna happen
27:33
that’s a receive that’s a recipe for disaster so second question that that we want to
27:39
talk about is is it even good technology I mean is it is it good when some
27:44
company stands here on stage and and talks about the framework to technology the language that they’ve built are is
27:52
it is it good for you and there are number of ways to learn
27:58
about the technology first and I think the most popular and most straightforward is check what the
28:04
analysts say and there are a lot of them there are Forrester Gartner HC IDC tons
28:10
of them some of them more more respectable as the other or not I guess one of the kind of the standards de
28:17
facto in our world over self a going around about technology is obviously the
28:24
government Magic Quadrant right and then you go from ok we have
28:29
ability to execute and completeness of vision basically do they have money do
28:35
they have good ideas and obviously people but companies with money and good ideas are here companies with only money
28:44
but no good ideas are here companies with ideas and no money are here and
28:50
this is what they call the niche players basically shitty tech and no money and
28:55
no money and should attack and no ideas she lets a populated quadrant here the kind that others have other means to
29:04
express the same and it only differs from the Magic Quadrant to the level
29:11
that they cannot be sued for plug arising the Magic Quadrant so this is a
29:17
Forrester wave it’s absolutely the same just looks different from current
29:24
offering from weak to strong that means do they have money by selling what they have and strategy from weak to strong it
29:31
means do they have the good ideas for the future and the attack is where
29:37
it should be that’s probably true so another way to go around it is the
29:45
gardener half-cycle Gartner hype cycle actually is a true cycle for the
29:52
majority of the technologies we start here up there was yeah sorry we know we
30:00
start here with the technology trigger and then everybody are all this is so great Doka Doka Doka and they’re like oh they
30:07
require root for everything and then they’re like well it’s we can can’t actually use it if we use it right right
30:14
that that’s the curve which kind of happens every time we go to conferences we all we all start with the technology
30:20
trigger thinking oh my god this is awesome I totally need this then on Monday morning we start implementing and
30:26
then by Tuesday or Friday we figure out this is the most terrible technology I’ve ever seen in my life
30:31
then we practice a bit more get a few updates of the tech itself and then we end up somewhere that we’re thinking hey
30:37
this might actually be a useful technology to solve so that it was a hype circle in a week basically now the
30:43
problem with those guys is that they are all paid to play right you need to be a
30:48
gardener customer to even be considered for the quadrant they eventually can put
30:54
you in the coordinator boot that you wanted you don’t want to be but until you pay you won’t even be on the counter
31:01
quadrant so you didn’t see Rd factory there we are not customers of gardener they you don’t know they don’t know
31:06
about us you don’t know about us is it true not so much there are other sources which are neutral and
31:14
respectable one of them I think the most important is felt worth technology rather any do any thought works people
31:20
in the crowd yes we have one oh no no no all right almost they usually are in
31:27
their website so we expect it to meet them here but they are not here this is amazing right because here they rate
31:33
technology by their consultants with their experience and what they have is
31:41
oh no you don’t say is not sure and we don’t have time basically go there there
31:46
are four different aspects techniques tools platforms and frameworks for each
31:52
and every one of them they have a what what does it call it adopt adopt hold
32:01
innocence Yeah right which basically means again what you should do right adopt adopt it now trial check if it
32:09
works assess expressing the restraint about it halt forget about it and if you play
32:15
with it and you kind of put their technologies that you know you can validate if what you know about the
32:21
technology is kind of aligned with the what they do and if it does it means that they are useful for you and just to
32:28
recap again the four questions this is how you battle your a cargo cult is my
32:35
organization ready to adapt a new tech maturity models is it’s even a good tech do some googling user analytics couchley
32:43
use technology rather so what do I gain from adopting this technology is it is
32:49
it gonna solve my problem am I am I gonna be better as a company by using the technology and last but certainly
32:56
not least it’s the technology good is it a good solution to the problem I have a
33:01
my that unique snowflake that may or may not have to you should use that that technology and with that thank you very
33:08
much just to remind you the show notes Jeff fobb do comes the show notes I’m Angie
33:14
bar hundred our business Lian sticks are on Twitter this is their of this Charlotte in case you forgot and Corey’s
33:21
not here so we don’t have to do that yeah yeah all right thank you very much thank you [Applause]