Performance Job Offers

In this episode, Henrik and Leandro go browsing for performance testing jobs and analyze what are the requirements that a performance engineer needs in 2024.
Not a performance tester(they are almost extinct now), but a Performance Engineer!
The requirements are very different from when we used to get performance testing jobs a few years ago.

Episode Transcript

(00:09):
Hello everybody, and welcome to anotherPerfect episode that as we promised this year,
we're going to be cranking up afew more episodes during the year because
we know that there's a lot ofperformance interests lately. Isn't that right,
Henrik? Completely? I mean Ithink with the Perfect Janse, I was

(00:30):
surprised to see that people are actuallyfollowing and very happy to have content released
quite often. So I performance isnot dead. That's great. It's a
good news for us. Yeah,it's just an industry that is going through
some changes. I'm not sure ifcalling it the performance puberty probably maturity or

(00:52):
performance modernity. Yeah, well that'san interesting concept because when do you call
it more or wasn't it modern liketen years ago? Well, I mean
we were marketing it's it's a modernperformance testing. Yeah, it's always modern.
But I don't know. I thinkI would it's interesting. I think
we should do an episode trying toput names of the Is it the Cretaceous

(01:18):
period of performance and now we areon the Cambric or I don't know,
because definitely we are seeing some changes. Like we started just in monolithical machines
that were not like accessed by multipleclients or multiple users. Then I was
a performance of servers and distributed solutions. Then we moved into service based solutions

(01:41):
that nowadays I think we're still strugglingwith those. And now it is a
cloud, the cubernetes, the microservices, elasticity and madness all over the
place. Excitement, I would say, not madness. Excitement. It's a
depending on who you ask to me, is exciting. But there's too much

(02:05):
to attack, to learn to understand. And actually that's that's a topic that
I wanted usk to talk about todaybecause I was curious, like, what
does a performance engineer nowadays needs toget a job? What jobs are out
there for performance engineers? How doesa market look like, because I'm sure

(02:27):
it's not the same as like aswe were talking in the Jurassic Age of
performance testing that it was just likeyou gotta know your load runner and to
record and replay and black use thebackend monitors. But I'm still still convinced
that there are still scriptures out therebecause the traditional model few years back or

(02:49):
maybe still now, where there isthis scripture that will just focus on scripting
and there is the senior performance guywho will are managing the rest and analyzing
and reporting and so on. ButI would be curious to see if it's
still out there because or just died. The performance need has change. I
mean, it depends obviously in whichit environment you're working on. If you're

(03:10):
working on traditional package apps, I'mpretty sure it's just going to be the
same. But if you're more workingmore on the clouditive space, I guess
the requirements and the delivery has completelychanged. For sure. I think you
made a very good point because thereare organizations which they might not be creating
software, they may just be hiringsoftware acquiring it, and due to regulations,

(03:34):
they still have to do like thisback reverse engineering to kind of be
able to automate have some performance metricsand say hey, uncle Sam or whoever
send your Sam well in Mexico noise. We don't have a name for him,
but for the government. But theseare the proofs that we acquire to
give services and it's good to go. It supports the people, they supports

(03:55):
the users, and there's this orthere are these organizations which are actively creating
software for themselves or purb or organizations. Right, yeah, but still if
they need to just check the requirementsin those expensive package apps. I would
say people are not using the defaultset up dated. Do some customizations the

(04:18):
plug it with other systems, theinclude a way of interacting with their accounting
solution or whatever. So it's notthe normal solution anymore. It's a modified
and customized version. So that whichmakes sense in that case to start benching
and making sure that it's still asgood as the previous system for sure.

(04:42):
And it depends as well on thesolution because on those situations we now have
also SaaS providers, which now thesoftware is leader, like you didn't just
acquire and install it on your serverand customized your own thing to be checking
it. It's like in the cloud, and you're sometimes even by by contract
that you won't do a load testand you want to think that you want

(05:03):
or you don't even get access tosome of those things. It's a big
debate because even if you use Salesforceexample, or of course your self was
going to be a SaaS, butI will probably connect Salesforce to my back
end or to my other applications,and Salesforce is going to be the component
that will send API calls in avery very regular base. So the load

(05:25):
obviously you don't test the UI anymore. You just test those API interactions for
sure when they have that API.Because sadly, I have seen some of
those services which is like the frontend is what you get, this is
what it is, and good luckwith your whatever you need to do performance
wise, and it's we are justcoming up with all sorts of examples that

(05:48):
are nowadays in the market. Andit's wild west because we have all those
flavors, all those variations. WhereAs you were saying, there are these
organizations that have this. They justacquire software, they just leverage it.
They don't have to be power usersor technologically rocket scientists. They just want

(06:11):
to get the software and get movingmoving on with the business. But there
are others which create their software.They are not advanced into it, but
they create their own software internally.And there are the Googles, the facebooks
that are inventing all these things thatare I would say incredibly advanced technologically speaking.
So we have all these tiers andlevels where we see positions for performance

(06:35):
engineers. Where As you say,I think there's still space for the scriptor
there's always going to be there's alwaysgoing to be this requirement that you have
the software that has to be reverseengineers scripted as we used to record check
the correlations to the good old days. But for this episode, I was

(06:58):
like wandering around the inns and Iwas looking, Hey, let's google for
performance engineer position. What are theyasking nowadays? So tell me, tell
me what is the role that youfound that would be interested to see?
So I found that I have themactually open right now. There were several
like QA performance engineer, performance testengineer, senior performance test engineer, and

(07:20):
repeating repeating QA performance tester. Firstone that I see tester, h this
is interesting performance benchmarking engineer, whichis like, huh, now it's benchmarking.
Okay, we have again perform testengineer, performance manager, senior database

(07:44):
performance engineer. Now it's specifically forthe database team lead and performance engineer.
So but still again many with performanceengineer. Funny, I didn't find any
like in the old days that justsaid performance tester because I think, I
mean performance tester was almost fifteen orsixteen years old, I think. And

(08:09):
then there is the yeah, Imean there was a term engineering that came
in the market afterwards and say,oh, now now I'm not a tester.
Anymore. I'm a great engineer andI feel that I'm an engineer,
so I'm going to put my roleas a performance engineering But in fact it's
because the work that we're delivering hasobviously changed. I mean, if you
do the one big band a gigwhere you have two weeks to test the

(08:33):
application, make and do some tunings, you are kind of testing, but
you're kind of doing some recommendation.But then if you're part of a more
a squad team where you go tothe various project teams and making sure that
they are implementing the early performance engineeringapproach and then making sure that all the

(08:56):
setup and helping and the work isnot the same anymore because you're testing less
or you're launching less tests. Butif you're more like an expert on the
overworld strategy, that's that's a veryinteresting because you you right away drew the
line performance engineer. One thing thatstarted to pop up and we started to

(09:18):
I call it that the charitisars,and before we were chairming, there's just
the performance tester. We haven't evolved, and I remember the performancessory requited as
you say, just Adam mad executegive the report, and that was it
Just like I remember the old days. I didn't even have access to the

(09:39):
back end to say other I'm pluggingthe traditional monitors that the tool came with,
or checking performan to see what wasgoing on in the back end.
I didn't get access to anything else. I couldn't. I wasn't an engineer.
I was just like testing hitting.Does it sound good? Bad?
Okay, here's a report that's whatit sounded. I don't know where it.

(10:00):
Most of the time I have noidea, I have no access to
the data. Is no permissions black. But nowadays with this performance engineer,
there are so many other tasks andof the positions that I was looking at,
they don't almost don't mention scripting atall. Tools. I found a
few positions that mentioned specific tools.There was one that was like almost mentioning

(10:24):
all the tools, like yeah,we want load runner, neo loads,
jmeter, case six, And I'mlike, do you really use all those?
I mean, it's cool, andI admire that organization if they do.
But most of what I've been seeingis that they want you to understand
and tune and be able to createframeworks for performance to be tested continuously.

(10:48):
I mean The good example is alMundo Alinda who said wapper. He said,
I don't care of the tool thatthe product will use. I define
my requirement and define how people deliver. I will provide the toolings and then
you pick whatever you want. Idon't care. But I think, yeah,
I think the market has shifted tothat. I think there are still

(11:11):
organizations saying we need one tool thatwill do everything for our organizations, and
then we acquire it and utilize itas much as possible. But then,
if I'm doing just micro services,what do I need a tool that is
smart for ACP or smart for salesforce, which I don't care. I mean,
I'm going to hit just an API, So do I need those advanced

(11:33):
feature Probably not. So that's whyI think the market change. For sure.
It's interesting as well that other thanknowing this or that low testing or
automation tool, most of them Ifound proficient in one or more programming or
scripting languages, which is something thatfor testers for a long time was like,

(11:58):
none of you know, I'm nota developer. I don't do programming.
I don't that's a weird thing.And we weren't even allowed to get
into that so many many times Iask like, hey, can I take
a look at the code. Ihave a feeling that you have something weird
in a loop, like those mistakesthat you leave something instead of a four,

(12:20):
and now is messing up your performance. Nowadays, we can hopefully the
whole team has access to the repository. You can see the code, you
can give recommendations, understand what ishappening, because I get the feeling big
time that nowadays is not so muchabout scripting but about understanding what is happening

(12:41):
in the application. Many of thisalso mention about creativity and understanding the business.
I think also one key importance isthat people say, I mean today,
if you look at more the CIICDapproach where you launch tests, if

(13:01):
you don't have the right skills.I mean, there's a lot of organizations
that says, we are automating allour tests. Oh cool, So how
do you do with performance? Wedon't do it? Okay, So at
the end you're not fully automating everything. You're just utimating the easy piece of
the cake and you forgot the completecomplicated tasks. So I think for organizations

(13:22):
you need someone who understand the challengeof the organizations and putting the right toolings,
the right processes to enable the automation. So be part of the pipeline.
Enable a way for people to trustthe output of the tests results,

(13:45):
especially if you run it from aget club or a bit back I mean
any get up actions or eat Jenkinsor tradition Jenkins. People need to be
trust the green and just focused onthe red. And I think this is
where the value is really really great. And also being able to be a
mentor where you go to those people, those teams who hasn't been doing performanceineering

(14:11):
since since ten years or five years, depend on your expertise. You are
here to give them the right information, the right mythology, so then they
are they can be independent and startworking and utilizing your best practice. And
I think I mean from an engineerperspective, it's even more motivating to work

(14:33):
in an environment where you see thatyou have unlimited possibilities in terms of tasks,
that you can influence the organization invery in the various ways. So
I think that's very exciting. Yeah, i'd see as an example, right,
it's not anymore that you just haveto understand the HTP productol and then

(14:54):
receive and that's it. That's asfar as you go. When you we're
saying right, and I'm seeing someof the job offerings that I found,
almost all of them say understand CICDand software development life cycle in modern environments,
agile methodologies, excellent analytical and problemsolving skills, which you see almost

(15:22):
everywhere, but here it applies.As you say, it's not anymore just
about scripting and bringing down the systemis still a key component on some of
the organizations. You need to knowwhat are the limits? When I always
keep bringing her up when Taylor Swiftshows up, you'd better have done some
low testing before that, or evenhaving a suitable experience in a sense where

(15:52):
you recognize problem patterns, you seesomething that oh, I think I know,
you know this is just as reactions. I think I know. And
then if you look at the rightlayer and you say, oh, yeah,
I know this is a typical problemthat I've seen of a lot of
tests or a lot of projects,I think that is a great skills that

(16:17):
yeah, could make the difference betweentwo engineers for sure. But wait,
Hendrick, how do you know it'sa lot of balance if you have not
even checked it or checked the logs. I've seen this happening before, so
yeah, trust me, I'm anengineer, a performance engineer, and I
think also the the I'm surprised thatthe job offering is not mentioning anything about

(16:37):
observability knowing. Oh yes they are. Oh yeah, I was going to
ask you because I was going togather these days you need, you have
to be aligned with a new wayof observing the system. So take advantage
of Graphana or the andres or whateverthe solution that you are in place at
home, in your in your organization. So I think this is a critical
component because at the end, ifyou know how to use those tools,

(17:00):
then you're going to be super efficientfor sure. Actually, there are a
few ones where I found that theymentioned monitoring technologies, for instrumentation, for
telemetry, for understanding all these things. The ones that have that that mention
our performance engineer don't even mention scriptingtools or automation. And that gives me

(17:21):
a feeling. Are those the typeof organizations that release into production without any
automated triggering process? Because I knowand Alma again she almost destroyed us all
with her examples and best practices,but she even mentioned. The critical things
for us is that you know howto roll back a change measure that is

(17:42):
good and that it doesn't need arollback. And this is an interesting situation
where if you have good monitoring observability, your software is instrumental and ready for
rollbacks. I don't see much needif you can gather the metrics and you
have I don't know, mixing upwith some other good practices atomation, peramit

(18:03):
having unique tests happening automatically on yourrelease pipeline. You know the performance,
you have historics, you can compareit, you don't need to automate.
And that's something that is like I'mstarting to see more here and there in
some areas. Probably you need asynthetic animation for production and have this historic
measurements, of course, but it'snot the same as what we used to

(18:26):
do, as we were just likeautomate everything and trigger your biggest low tests
and try to bring down the system. I think that the other changes that
we have many more ways and manymore tools on how to pull this off
this performance thing, because in theend, a performance engineer, you're like

(18:49):
robot cop style my private directive goodperformance for end users system not dying.
That's what you want and how youachieve it. I did put a live
stream with a perf Bye Francis,and there was one of the biggest retailer
in France who explained how they weredoing it. And they are even study

(19:12):
machine learning because they once the testis running, they extract all the results
from the low testing product and thenthey have other KPI from other environments they
feed this gigantic solution that they havebuilt from their own and this solution is
basically learning from patterns and comparing versionsand then feeling it like a big matrix

(19:34):
and then get an output in andsay okay, everything's ween or if it's
not green, it shows you whicharea is you should pay attention. So
they are faster because they're running testsevery day and they're very efficient on the
way they They are able to triageor to to take decisions on the results
that has been observed on the onthe processed and I never, I mean

(19:59):
I was very surprised when they startto mention in that I say wow,
because machine learning could be a it'sa great solution, but it could be
very expensive. You don't know whereyou're going to end at the end,
I mean be able to build somethingthat is efficient as you expect. But
I think now if you have thatskills, then you can clearly help the

(20:23):
organization to be more efficient. Yeah, that's another mix. I didn't see
any of the few job offerings thatI found mentioning AI, but definitely there
are organizations already thinking on that.I had a conversation no long ago with
another of our friends from Abstracted.Christopher and Federico were like, how would

(20:47):
we use AI in performance? Andas you say, you need to have
some previous information to teach the machinemachine learning engine and to identify some of
these things. That first step isnot easy. You you may inadvertently end
up with a data lake that youdon't know what to do with that,
and even not even the machine learningwill know, like, oh, dude,

(21:10):
that that thing that you gave me, I don't want to touch it.
But but even I was asking them, how are you able to onboard
a new project on that process becausethey were doing many one one big,
one of the biggest project, whichwhich was the main web page of the
of the organizations. But I'd say, you probably have other project who wants
to utilize this approach and say areyou can you can you easily on board

(21:30):
the other projects say ah, yeah, but the answer was not yes,
no problem, It's gonna take fewseconds. Now I think I think there
is that there's a journey behind forsure, but I would be very interested
to see what is the actual journey. Can you how does it take to
onboard new project, because this isthe type of information that people want to
have, specifically when they want tobe efficient in detecting problems in the in

(21:55):
their performance test, and as well, I think you're bringing something very into
seeing this new project because nowadays wewere used to like there was these big
bank projects and it was the release, the release, and that's the story,
and that's why we have to doall the big bang performance testing,
the bolt and everything. But nowadays, in theory, if everyone is doing

(22:18):
things the right way, they havethe MVP first, they are starting to
generate some metrics the users probably ifthey have good observability mixed up with some
machine learning, they can't quickly releasesomething you should not have that many users
like jumping at the same time atthe beginning, some better testing, good
practices, and some of these organizationsseems like they are looking forward to doing

(22:41):
things this way. But I meanit's normal. Recruiters is looking for skills
that are currently used at the organizations. But I'm thinking more in the long
term. See, if I'm enteringan organization, there's a big change that
in six in a year from now, the processes and the way we deliver

(23:02):
will have changed because the market isshifting, is moving so fast, so
you probably expect to do something atthe beginning, and then you will have
to adapt to be more efficient.And there is a way of delivering your
preor you must be aware of that. But the approach of using future flags
is taking more and more, whichI think it's a great way of saying,

(23:25):
I want just to enable this featurefor only the customer based in California,
for all the customers that are abovethirty five years old. I don't
know, liking I don't know.You can make a lot of combinations of
criteria to enable a specific feature.And once as a performance gig, the
first geek I was the first thingthat I was I had in my mind.

(23:48):
I say, okay, so nowif I need to test that system,
how I mean, I need toenable the feature flag, But how
do I test that, how canI make sure that if I enable that
future I want Andrew in a answerin a pop performance bottle like somewhere.
And this is there is no clearthis description on how you should achieve on

(24:14):
this. This is an open questionto you all you guys, by the
way, if you have any answer, drop comment below. But I think
this is something that we need torepair ourselves as the performance engineers, because
this is going to be soon anormal way of delivering software in the market,
and for us as performance dinosaurs,it could be clearly something that we

(24:37):
need to adjust ourselves to be alignedwith that new mythology. Yeah, it's
not only the methodologists are thinking howto fix this thing, because, as
you said, you think of somefeature flax and sometimes you don't even realize
and you're opening some other Pandora boxesjust by doing certain future flag and eventually,

(25:00):
oh, I shouldn't have that.But with new technologies, new methodologies.
Another one that I am seeing onmany of the job offerings is Cubernetics
Understand the Cloud, which is somethingthat as well, and I've saw a
couple with lambdas as well, andI'm like, huh, I haven't heard

(25:22):
that in a requirement in a bit, but it's potently. I mean,
if you look at the trend,I means I think seventy five percent of
the market is using community cubuneties inthe environments, which I'm not surprised to
see that skills because at the end, if people build software and deploy them

(25:44):
in a cluster, in numeran cluster, then obviously if you're a performance engineer,
you need to understand all the conceptsthat are behind that, because otherwise
tuning a permidal environment and tuning acommunity cluster is obviously completely different, and
there are some concept that comes withit. So I think it's it's for
me, it's one of the mostimportant criteria for now and probably for it's

(26:07):
going to be worse in the nextcoming month and years. For sure decade,
that would say, because my feelingis cuban neeties has come to stay
for a while. I'm not sayingit's the technology for the ages forever and
ever, but it's key because youare going to be testing everything is migrating

(26:29):
into the cloud, I'm sorry tosay, but everything, and if you
are you are not a tester anymore. I couldn't find any position that says
just performance tester. You're a performanceengineer, and you've got to be able
to say analysis to get and sillythings that we were used to kind of
I was bringing up very much onpurpose. The load balancer in cubernatis now

(26:52):
you have multiple pots that are generatingexpanding no, and you have now the
I what's the name of the Qand as element that acts now like a
load balancer among all those part servicethank you, and you can understand why

(27:14):
is it killing pots, why isit bringing more? Why is everything?
Probably and I've seen this a coupleof times going just to one part and
the other ones are just like,yeah, it's a good day. Some
say things that can happen and areperformance related, and you need to understand
what are the weakness have communities Anddoesn't matter how infinite your cloud can expand
to. If you find some ofthese problems, your performance will diminish no

(27:38):
matter how much money you throw atyour cloud. I mean, especially if
you have a cluster with lots ofservices, because yeah, there is the
you will allocate AP address, evenif it's a virtual network, there's always
a limitation of IP addresses. Ports. You will probably allocate ports on the
node because you will run different applicationsand depending on how you build your services,

(28:03):
and a physical machine has a limitednumber of worts as well. So
the traditional limitation that you know froma barnmental environment, you will have them.
But the first reaction when you startworking on communion is you just say,
oh, it's a different so let'sput those knowledge that I have side
and then focus on the new concepts. But still once you get it there,
you say, oh, now I'mgoing to bring the old concept because

(28:26):
it's going to be potentially a limitedon my coustro For sure, this is
interesting because we are bringing concepts orwe're talking to the audience. You pretty
audience watching us right now or listeningto us, who when we have the
seasoned engineer, performance engineers, performancetesters that have been like both of us

(28:48):
for a few decades in the industry, but there's also like the newcomers that
the new people that are like,yeah, I think I'm gonna become a
performance engineer, which are like,oh, so do I have to understand
how qaneties and development languages and thecloud and because it used to be script

(29:10):
execute and try to figure things outnow as a performance engineer. It's a
wider perspective, as we were seeinga few even like databased understanding, which
I haven't seen in a while,like understand the SQL full table scans,
indexes and things like that that I'mlike, huh, okay, this is
very particular. Some others got myattention big, big time. One of

(29:34):
the requirements the applicant must understand kitrepositories and version control. How was it
like topics or just understand it?Yeah, because I think testing. Putting
your test assets out of the coderepo doesn't make sense because your test is

(29:56):
something that makes the code deploy,can be a deploy so like a validation
process, So why should we storeit somewhere else where it should be living
with the actual assets of the ofthe component of the code whatever. So
I think, yeah, I'm notsurprised because I think it's it's it should
be the day to day process ofdealing with your test files and test assets

(30:22):
to yeah, to to keep trackon the changes between versions. I think
it's it's a great way of storingthose testing assets and even mixing up some
things that just popped in my mindin the past I remember, and that
was like a key plot point inthe Phoenix project. If I remember,

(30:42):
well, trying to bring up theenvironment so that you can run the tests
and start playing with the thing rightnow. If you understand get if you
understand pots or doctor or containers,which I haven't seen mentioned anywhere, but
that's also interesting or important that youunderstand, bring up pot with the service

(31:02):
that you're going to test, bringup part with your test platform script,
whatever you're going to do to thattested kill both cut your gather your results,
and move on through the pipeline.Some of those understandings are like,
wait, what, I don't havea test environment anymore, Like just bring
up a moment ten minutes and domy test and be done. What or

(31:23):
even just knowing you I mean,if you're working in a new organization that
shifted utilized platform engineering, the platformengineer will probably generate for you a virtual
cluster in the cluster, so youwill deploy whatever you have to deploy to

(31:44):
make your application runnable that in avirtual cluster. And then so you have
your own cluster running in this physicalcluster and you can do whatever you as
a test. So I think evennow, I think it's pretty exciting because
I remember how we we we aregoing to build the performance environment and waiting
to and now this is not Ithas to be as close as possible to

(32:06):
production, a multimillion dollar production oflike that. And now you have the
able ability to say, oh yeah, iron test, I destroyed, Iron
Test, I destroy Aron test.That destroy and that is exciting. So
it means that you also need toknow how you deploy things at the end,
because at the end, if youare in charge of the performance environments,
you're probably gonna have some help onthe other teams, but your your

(32:30):
pieces will be embebced close to theapp somewhere in that virtual cluster. So
you have to know how to deployit, how to manage it. And
so yeah, clearly cummunities is goingto be a requirement for sure, cubanets,
kids, pipelines, all these thingsthat we have to understand at the
very least. I don't. II'm not implying that Hendrick, you've got

(32:52):
to be an expert on everything andthe Unicorn full stack Performer blah blah blah,
Super Saiyan Level three and stuff.No, you can't understand those principles
to get into these new performance roles, trends and job positions where you have
to do a lot of these things, and hey, why don't you just

(33:15):
give me a test environment a littbit there. You know that costs per
hour or per unit of time thatyou keep it up, run your tests,
destroy it and stop wasting Money's likekeeping the microwave on all the time
without having to hit anything. Justturn it on when you need to hit
a performances or some of the otherorganizations. I remember, let's do the

(33:37):
biggest lotus and try to bring downthe system in our cloud paid elastic environment.
Right, I didn't need to clarifywhy that was stupid and you started
laughing. Well, for the onesthat do not know, your cost of
the cloud will go to the skyif you're trying to do a capacity test
in your paid cloud. So Ihad a story about that. It was

(34:01):
with my previous simployee. We hadto test the release of the latest last
season of Game of Thrones in Australiaand it was the it was the biggest
broadcaster company in Australia and we hadto hit I don't know, it was
the The throughput was insane and sowe did a low tests in the cloud,

(34:27):
and so we're utilizing all the datacenters it was in Australia, and
even with the data centers that werein Australia. First of all, one
thing is we we booked all thosemachines, We spin up those machines and
then in the middle of the night, my CTO had a phone call from
the guy on Amazon and Sydney saywhat do you what are you guys doing.

(34:50):
You've eating all our resources in ourdata center. There is no space
for any other customers anymore. Sowe were surprised first of all. So
it means that we eat it,beat up the entire day center. And
then the bandwidth that was coming outof the data center, there was no
Internet anymore on on that data center, so we have to so we we
have some guidance from those club ofOater to reduce the consumptions in in in

(35:16):
the island. So we discovered thatthere was a main internet line between Singapore
and Australia, and at the end, if you were generating the load from
Singapore or if you were generating itdirectly from from Australia, it was the
same thing or for few few millisecondsdifference, but it was almost that you

(35:37):
were in the in the Australian network. So we we basically the machine that
we were not able to get inAustralia, we were putting them in Singapore
and we're hitting a massive, massiveload test. But that that was that
was fun, just as just illustratewhat you said, if you wanted to
test something, which is to hitthat limit, you're gonna probably need so

(35:59):
many machines you don't even know thenumber of machines for sure. And then
you're facing too, yeah, bendwith limitations because you saturated network of the
cloud providers, saturated the DNS,of the local dns, the of the
of the country and so on.So yeah, don't go on that route.
And I am wondering if you tookover everylimater or inch whatever you metric

(36:22):
use of capacity on a WS everythingAustralia and in those areas, how much
was that test when the AWS billcame back. I'm very curious because I
mean we we we shared the pricewith the customers and the customers said it's
going to be There was two nightsof tests. In fact you were three

(36:43):
nights because we hit the problem withthe data center. But very stressful because
you have to be very efficient invery short time frame. And but they
were the we explained them literally,we need that amount of machines, you
need that amount of bandwidth to reachout what you want to test. And

(37:04):
so this is going to be theactual price. And it's okay, no
worries, we have the budget.No. But but on one hand is
generating the load, how much doyou want to hire to generate the load?
And the other one is how muchdo you want to hire for the
tested system? Because you're going topush that system to the limits. Probably
is going to grow, it's probablyis gonna Well this is where you end.
This is where you need a proofa role of a person that has

(37:30):
significant experience to react and be efficientdoing those those types of projects. But
yeah, but yeah, this isI mean sounds like you also need a
finance expert in the as a performanceprofile because you do you understand how much
is it going to be a profitthat I remember the roy was like,

(37:52):
hey, what is the roy forthis performance test? What is that true?
Put related or yeah? But theone thing is not to relate it
to the roles is I don't knowif for you Leandrew but I I feel
the recruiters, especially in LinkedIn aswell, Oh gosh, I mean the

(38:17):
they barely understand what we are actuallydoing and they are asking you for roles.
I mean, guys, please,I mean someone wrote a full book
based on that, just quick seguefor our friend James Bully created a book
to try to help recruiters understanding that. But that's that's a tough thing.

(38:40):
I mean, I would say youneed a performance engineer to be a good
recruiter for more performance engineers. Otherwisethe candidate can be as a recruiter the
recruiter, and you can play yourway around that. I've seen people just
throwing out passwords and trying to makeit through the interview and they do awesome,

(39:01):
and the recruits like they were incredible. They understood all the crypto AI
ledgier blathing and yeah what, ohmy god. And I don't know.
It's that the the vacancies for jobpositions are changing. I think let's just

(39:25):
ramp down and wrap up this alittle bit. I feel bad for the
recruiters because they had to understand alot. Even like the performance the candidate
has to also understand a lot.The recruiters are not in a better position.
They have to keep up or tryto. They don't, they won't
understand. They have to fake upthat they can know or pass it to

(39:46):
like the performer supreme, that theorganization has and can do the right questions
or I may I may ask,I mean, I may add to the
job description to be honest something,because I think if you want to be
a good performanceeer, you need tobe curious. Oh yeah, because if
you're not curious and you don't liketo investigate like a detective. No,

(40:09):
no, I've seen it. Andthis is I have a position right now.
Super curious and excellent skills. Idon't know what this is faired up
in that description. Strong communication skills. Yeah, that's another one. Communication
is super important. Technical creativity,empathy, humility and appreciation for understanding and

(40:32):
lean startups Okay, they closed,it's a startup. They need all that.
So I'm very happy that the artistadded this notion of curiosity and so
that's really good. Yeah, becauseit's something that I think many looking for
the job is like, yeah,what do I have to run around?
Execute by? No? Like tryto understand deeper, like why is this

(40:54):
slower? Why it suddenly got tentimes faster, which is a red light.
Well in performance, just what isgood enough? Exactly? Yeah?
Can we improve? How far dowe improve? Understand all the the sik
mass of improvement. How much isit going to be every point nine that
you add to the improvement in yourservice? How much is it going to

(41:16):
cost of time? All these concepts. It's it's very interesting, very entertaining,
and you, I think that's avery good point. Be curious.
Never never feel like you know enough, you learned enough, You're at your
peak and I'm super performer already.Why are you asking me to read about

(41:37):
cure? What? Hmm? SoHenry, to wrap up a little bit
this episode, what do you wouldyou say or give us a recommendation for
our audience that that is looking toland a new performances in job, or
to switch if they already have one, or to start in the performance testing

(42:00):
trade practice realm, if you wantto say, I mean we mentioned it's
so curiosity. Communications is a key. Knowing a tool of course, because
at the end you will have tosomehow help someone to build a scripts.
So I'm going to interject there,I would say more than knowing the tool

(42:20):
of understanding the tool, because withthat you can do like almud and I
recommend it. I don't care whatwill you use, They do more or
less the same in the bottom justunderstand it. I don't care if you
drive a Homer or an old Wolksbag and beetle. They both have a
wheel, a gas pedal and abreak right, Yeah, that's true,
that's true. Some more understanding theconcept of the tools for sure. And

(42:42):
also I think still having a backgroundof understanding the concepts of architecture that we
had a few years back HTP,the network work how which have a garbage
collector, works a lot of typeof things, but you were going to
technically be deliver things in communities,deliver things in a cloud provider, the

(43:08):
lever things in the NAMDA functions.So knowing what those terms means, how
you what are the problems and howto resolve them, how to do efficient
cow balancing, how to do networkingin those environments is going to be really
important because at the end, evenif you're not doing it today, you

(43:30):
were probably going to do it inthe next year or next month. So
it's going to be a critical knowledgethat will be required any way to deliver
properly your job. Yeah, mixinga little bit of the old stuff that
we used to know is important stillrelevant, but applied to these new technologies,

(43:51):
new environments, new methodologies, howto translate them. Well, what
things matter, as you're very wellsaid, what is java actual machine,
garbage collector and all that, butinside of a pod in an elastic environment,
in a cloud. And and alsoknowing how to observe a system with

(44:15):
a new observer stack, So understandingwhat is a trace, what's the value
of a trace? You know problemetricsbecause you've been doing performance heering, so
no need of explain metrics, buthow to take advantage of logs in a
more efficient way. So and alsohow to read a frame graph if you
have to jump into continus profiling,knowing how to read it and how to

(44:36):
utilize it properly when you need it. So I think those are very important
aspects because at the end you willbe more efficient and the organization will be
happy about the work that you're doing, and they're kind very extend the number
of performance tasks that will be requiredfor the organizations. Yeah, those are

(44:57):
very good recommendations. Everyone listening,watching or tuning up with us. I
think you got to keep that curiosity. I loved it, keep learning,
keep up with perfect because we aregoing to keep explaining those new concepts and
those things, or at least doingour best or bringing the expert because we
have invited some cool people in thepast and will continue doing that and apply

(45:22):
to these new technologies. If youare new to the performance trait, understand
them, how do they interact,and how they can get your In the
end is the end user experience,how fast, how efficient is it or
if it's still alive it stays alive, how all that will interact will affect

(45:42):
so that you can keep releasing softwarecontinuously because nowadays it's continuous repository based,
shaable and in the cloud. Andone last thing, we forgot how to
automate things mm hmm. Because it'snot only the atomission of the tests.
We're automating the automations to builds thatcompose that test, that release that create

(46:04):
your infrastructure, that create a newenvironment, close it, blah blah blah
blah, and you're gonna be able. You get to be able to control
that, observe that, understand that, and check that it's responding well.
I think those are the cornerstones fornowadays and everything that involves around understanding programming
languages and so on. So takea look, I would say to close

(46:29):
the episode. I hope these recommendations, this analysis of the job market on
performance engineering, I'm not calling testinganymore. How it is changing, How
is it going to impact you,and what you can do to keep up
with it. If we forgot something, as Hendrix said, leave it down
there in the comments, send usa note. We are on LinkedIn Twitter,

(46:55):
where else is provides Facebook. Ithink those are the three in ones
that we pay attention. No TikTokplace, no ticktok no no, I
want to do an episode on that. By the way, performance wise,
I have a very interesting personal experiencewith TikTok and performance. But to not
to dbate too much for everyone,I hope that this helped you, that

(47:21):
you enjoyed this episode. Beware ofbecause there are more coming up soon.
What do you think we should betalking about next? Hendry A few,
but you, audience, if youwant more topics, let us know.
I think we could we could startdriving topics about cubineties because we talked about
it and we didn't touch it.I think it would be very beneficial to

(47:43):
remind you concepts and that's a deeprabbit hole, yeah, or a couple
of things that you should pay attentionto when you when you have to work
on those environments. Maybe also sharein the same episode, maybe we can
cover tools that could be interesting foryou when you have to deliver that your
test in incumunis environments. So Ithink that will be very beneficial for everyone.

(48:05):
Again, if you have other ideas, drop comment for us and we
will focus on those topics as well. And with that, I think it's
time to end up this episode.Thank you very much everyone listening. Thank
you. Is this our first episodeof the year. It is. It
is so happy twenty twenty four andbest wishes for everyone watching and listening.

(48:28):
Leandro out Hendrick out see you soon. Adios.