Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
1:59
side and the wind and it went and
2:02
it turned into like a big mangle of
2:05
metal and broken 4 euro
2:07
pinata. Wind pinata.
2:10
I gave it one minute for each euro. That's
2:14
how long it lasted. And
2:16
then I got smart and bought a 10 euro
2:19
umbrella which was quite a bit more
2:22
sturdy and worked really well. And
2:24
feeling all smug, somebody in the hotel's speaker
2:26
said, you know, the hotel will just give you an umbrella. Thanks.
2:33
I feel smart now. Yeah, I feel smart now. Hey,
2:35
let's roll the music for better know framework. Alright.
2:38
So I practiced
2:41
the
2:42
other day. I
2:47
was asking my friend Richard Campbell, what's up with
2:49
HoloLens? I haven't heard anything about Microsoft
2:52
HoloLens in a long time and he's gone, I don't know.
2:55
So of course I went
2:56
and did what most people do. I googled
2:59
it. Binged it. Oddly enough. Google
3:01
Bing. And so it turns out there's a story from
3:06
onmsft, like
3:08
onmicrosoft.com from
3:11
October 11th, Microsoft releases Windows
3:14
holographic 23H2 update. Nice.
3:18
Here's what's new. And
3:20
so I guess they've
3:23
got their military version close
3:27
to deployment now and you can see in this article
3:29
that there are people in the army wearing it. And
3:33
it's been a big project for them. And
3:37
it's a big multi-billion
3:39
dollar contract
3:41
they have with the army. And so
3:45
hopefully that will turn into
3:47
good stuff for us in the future. Yeah,
3:49
I presume this is just Windows 10 but
3:52
the holographic edition. Yeah, Windows 11. I
3:55
think it's still 10 that they're
3:57
using for the HoloLens too. I don't
3:59
know if they've actually. moved up to the new version. Okay. Which
4:02
why you know they because Windows 10 is getting a 23H2
4:04
so this is the equivalent just for the holo-raphic edition.
4:07
Right right right right okay well
4:09
anyway so you can click on the holo lens
4:11
to release notes from October
4:13
10th 2023 and read all about it. There's no real
4:19
product as far as I know yet. You can
4:22
still buy a holo lens too. You can get a holo lens
4:24
too it's for grand and
4:26
it's been years. Yeah it's been years. And
4:28
I don't see new hardware on the immediate horizon
4:30
but you know they all they've they've basically talking
4:33
about we're waiting for the right chip
4:35
set like the right rev to come along. Yeah.
4:37
It gives us enough battery life and enough things that
4:39
it's gonna be worth making the next one. Right. Meanwhile
4:41
meta quest 3. Yeah. Is $500. Yeah.
4:44
Although it's a
4:47
VR headset not an AR. It's VR not an
4:49
AR. Yes. But and everybody
4:51
says it's a lot of fun. Yeah I've
4:53
got a quest 2. I have to decide whether I want
4:55
to upgrade it quest 3. Yeah me too. Cool.
4:58
Anyway that's what I'm thinking about today.
5:00
Who's talking to us today really? Well knowing we're going to talk
5:02
about sustainable development. Yes. I
5:04
dug back into the archives and I found a
5:06
show from 2011 episode 725. So 1200 episodes
5:09
ago in December 2011. I think I was 14
5:17
years old. I don't think that's true.
5:21
We talked to Kathy Malone who
5:23
was a tech ed speaker about
5:26
being a green developer. Great
5:28
conversation you know 45 minutes long
5:31
and she's gotten here there and
5:33
everywhere and admittedly quite
5:35
a while ago and there was comment on the show at
5:37
the time 11 years ago. This is from Timothy
5:40
Clinkie who said in our company
5:43
telling a developer to make their code more environmentally
5:46
friendly means they need to make it more green
5:48
by adding comments.
5:50
They come up green in the telephones. Oh
5:53
come on. It really does
5:56
that. I thought it was funny Tim.
5:58
I thought it was real funny. Tim, I'm going
6:00
to send you a copy of Mise de Cobaille whether Carl thought it
6:02
was funny or
6:05
not, but it made me laugh. And
6:08
if you'd like a copy of Mise de Cobaille, write a comment
6:10
on the website at dotnetrocks.com or on
6:12
the Facebooks. We publish every show there. And if
6:14
you comment there and I read it on the show, we'll send you a
6:16
copy of Mise de Cobaille. Write a green comment,
6:18
please. Very green. As long as you put the
6:20
double slash on it, it'll be green. Everything
6:23
will be fine. Everything will be fine.
6:25
And you can definitely follow us on Twitter if you want
6:27
to, but the cool kids are over at
6:29
Mastodon. I'm at CarlFranklin at
6:31
techhub.social. And I'm Rich Campbell
6:33
at mastodon.social. Send us a toot. We
6:36
will read them. And with that,
6:38
let me introduce the one, the
6:41
only, Leah Ladieno. Close.
6:44
Close? Good. Ladieno.
6:47
Yes. Let's go with that one.
6:50
Yes. There you go. Okay. Leah is a passionate
6:52
technologist that loves to get people together to solve
6:55
problems. With about 10 years of working
6:57
as a software engineer, Leah has been
6:59
working in London as a C-sharp backend
7:01
engineer, delivering value to FundApps'
7:06
clients for the last five years.
7:09
FundApps, plural. FundApps'
7:11
clients. For the last five years.
7:14
Yes. Yes. And we're talking about
7:16
sustainable development. But first, welcome
7:18
to .NET Rock. Thank you very much. Thank you for
7:20
having me. It's our pleasure to have you here. And such a cool
7:22
topic. Yes. And I
7:25
really like that we talked about this 12 years
7:28
ago because I got to hope things have progressed.
7:30
So when you say sustainable development,
7:33
what do you mean? Yes.
7:34
I think sustainable development
7:37
changed over the years.
7:38
To me, it means
7:41
I only buy a laptop every other year.
7:43
Nice. Not every
7:44
year. Yes. I mean, when we talk about sustainable,
7:47
there are like three aspects. Like if you Google
7:49
it, economical, social
7:52
and environmental. Obviously,
7:54
because we are here talking about green, we
7:56
are talking a bit more about environmental
7:59
impact. I
8:01
think, you know, like last
8:03
decade or even more, it was
8:06
how do you make developers code
8:08
more sustainably, not create
8:10
tech debt and all of that.
8:13
To me that says
8:15
maintainable rather than which
8:17
is a kind of sustainability.
8:18
Exactly. But it's a different, I would
8:20
say it's a different kind of sustainability.
8:23
It's not environmental. Right. So
8:25
like they're different aspects. Sure, without a doubt.
8:29
And can I say the obvious, like planet
8:32
is fucked. So we are more
8:34
thinking about environmental aspects
8:37
nowadays.
8:37
Don't use AI. Don't use the
8:39
cloud. And listen, the
8:42
planet's fine. It's all of
8:44
us that have got the problem. Like the planet's
8:46
going to make us. We may
8:48
be burned off of that planet. We're
8:51
an annoying rash on the planet. You know, when I do
8:54
the climate change talk for the high schoolers,
8:56
we talk about the in the Eocene, which
8:58
is like 24 million years ago, carbon
9:01
dioxide was at 1100 parts per million. Right.
9:04
As opposed to what is that right now for 18. It's like
9:07
planet was here. Heck, it was just a big old pile
9:09
of algae. That's all. You know,
9:12
Antarctica had probably had
9:14
trees growing on it. So it
9:17
is, it is, we are, the
9:19
troubles we're having are impacting
9:22
humans being able to function here.
9:24
Exactly. All of the fires
9:26
and the floods and sea levels
9:29
rising. And that's just
9:31
direct impact. Like people not being able
9:33
to live on this planet. Right. There
9:35
is also like all the health issues and what's causing
9:38
that. The subtler parts of all
9:40
these things. So you basically have
9:42
three categories, but I'm immediately
9:45
thinking there's got to be hundreds
9:47
of issues and topics that we
9:49
could talk about, right? Yeah. There's
9:52
got to be hundreds.
9:53
And I'm sure there is, but as software
9:55
engineers and software professionals.
10:00
We can talk about code and what we can do
10:03
and I learned yesterday that 20 to 25
10:07
percent of electricity usage can
10:09
be attributed to digital sector Sure
10:12
And I mean that that's why we are
10:14
talking about it even more and that's
10:16
why it is more important. Yeah,
10:19
and even Richard you had a
10:21
talk about future of electricity. Yes of
10:24
energy actually the future of energy
10:26
Yeah, and our move towards
10:28
using less carbon emitting energy sources.
10:31
Yeah, which certainly helps matters But
10:35
it is interesting in the news especially
10:37
around the large language models, which are
10:40
very compute intensive Now
10:42
there's folks talking about well just how much energy is
10:44
that is it really necessary? also
10:47
the cooling parts of it and you're complaining
10:49
the The cloud vendors
10:51
and Microsoft's among them have come
10:54
up with strategies to make very inexpensive
10:56
cooling systems for computers Yeah, but they're
10:58
fairly water intensive And
11:01
so they you know the sort of reality
11:03
of you're taking Fresh water and
11:06
you're essentially misting it into these
11:08
data centers to cool them. What could go
11:10
wrong? Well, I'll tell you what
11:12
cost of it. I built an i9 machine Yeah
11:16
in my house, and it has a giant
11:19
heat sink And when I say
11:21
giant like I can I can
11:23
barely get the case cover on it, right? That's
11:25
how big it is It's probably a half
11:27
a foot square, but it's probably a
11:30
couple of pounds of copper. Yeah, right Yeah,
11:33
but it's really quiet
11:35
and it's really cool Right and it doesn't
11:37
doesn't require water cooling and it's an i9
11:39
like it's a beefy machine Yeah, and
11:42
the graphics also. Yeah the
11:44
modern video cards you look at the family assemblies
11:46
are on those That's not true really there. The real
11:49
question
11:49
is
11:50
if you pinned it like if you had that
11:52
processor working flat out Yeah, how
11:54
warm would it get? I mean, you know the question
11:56
we know that cooler would disperse that heat
11:59
But that just means it's in the room. And
12:01
let's say that it's a client machine,
12:03
right? It's not a server machine serving up lots
12:05
of CPU. But
12:08
I wouldn't want my data
12:11
center to be completely reliant on
12:13
solar energy either because, okay, my
12:16
server's down. Yes. I
12:19
mean, they use a lot of hydroelectric power. One
12:22
sense I would think that in general, the cloud movement
12:25
has meant that a relatively few
12:27
number of people are responsible
12:30
for a tremendous amount of compute availability.
12:34
And when green requirements are put
12:36
on them, they have a better chance of implementing it
12:38
than we do if everybody was building their own
12:40
data center.
12:41
Yeah. I mean, it's here, it's numbers, right?
12:43
Like economy of scale. If
12:47
we all use cloud providers, surely
12:50
we can push them to do something better
12:52
and in a better way. Plus, we
12:54
can all reuse these resources instead
12:56
of every one of us having a server
12:59
at home.
12:59
Yeah, running 10% utilization.
13:02
I think it's also – that's true.
13:05
I think it's also making us lazy.
13:06
Yeah. Well, I don't know that we've ever
13:09
prioritized this. Like, if you think about the priorities
13:11
of a software developer, it's like, one,
13:13
make
13:14
it work. Yeah. Two,
13:17
make it work fast. Right.
13:19
Yeah. Make it work so that somebody
13:22
else could possibly fix it. Yeah.
13:25
Like, where is any other consideration at that point?
13:27
Make it available all the time.
13:29
Yeah.
13:30
Yeah. I don't remember these times because
13:32
probably I was not born or I was just
13:35
a kid. But I mean,
13:37
once upon a time, memory and CPU
13:40
and hardware were very expensive.
13:42
And constrained as a consequence.
13:44
Right. And huge.
13:46
Yeah. It was a massive expense. And
13:48
you did need to consider how you
13:50
code. It needed to
13:52
be super efficient. It needed to be for
13:55
computers. That's how we got solid
13:57
principles. Right. Like, we were writing code
13:59
from a computer. machines to be efficient.
14:01
Well, because we had to. We had to,
14:03
exactly. We also limited the feature set accordingly.
14:06
Yes. You went upon a time you could not waste
14:09
cycles on a GUI, much
14:11
less a touchscreen. Yeah. That's
14:14
a lot of extra compute. Yeah,
14:15
we move now into world of
14:18
memory is cheap, CPU is cheap.
14:20
Storage is infinite.
14:21
Storage is infinite. You can scale these
14:23
servers, you can add more nodes, you
14:25
can add more memory and CPU on fly.
14:28
It's very easy. And
14:30
everything was about performance. Like, everything
14:33
was like, I want it done immediately.
14:34
Yeah, and that's a great version
14:36
of performance. Not even the code is fast, but you got the
14:38
code written fast, like the performance of get
14:40
the feature out. Get
14:41
the feature out, but also run
14:43
as fast, but not by writing the code performance.
14:45
No, by buying the figure process. But by like, all scale
14:48
it, and like, it's going to be on a hundred nodes,
14:50
and it's going
14:50
to be super quick. I will fix this with the giant
14:53
compute button. Exactly.
14:56
So, do you see your role as
14:58
shining light on these inefficiencies and
15:00
saying, you know, hey, we have to get better at
15:02
this, or do you have solutions
15:06
to these problems that can be implemented today?
15:08
Yes, no. Well, it's
15:11
a complicated question, but I do
15:13
think we need to change something. It's
15:17
harder for the planet, because
15:19
we see what's happening. We can
15:22
get smarter at it. But
15:25
it's also, if we
15:28
don't do something, as software
15:30
engineers, our boss is going to come
15:32
after us, and it's going to say, you need to write
15:34
it in this way, because either
15:37
government told me that I need to,
15:39
or because we can't sign any more clients,
15:41
because they're all asking for
15:44
all of these checks that we are thinking about, I
15:46
mean, even in Swandos, now
15:48
we are, all of our prospects
15:50
are looking for these vendor checks, and
15:52
like, are you doing this? Are you
15:54
doing that? And the reason why we get
15:56
the signature was not only because
15:58
of that, but is because we
16:01
are thinking about these things and at the moment,
16:03
they are optional. They're not
16:06
there if you would, you
16:09
know, if companies thinking about
16:11
it and thinking about the environment. But
16:13
in my opinion, soon, this is
16:15
going to be a must. It's
16:17
going to be a mandate. Yeah.
16:19
Well, and I wonder if it just won't become table stakes.
16:21
Like Microsoft and all of these tech companies
16:24
have been talking a lot about zero carbon
16:27
and zero impact technology centers. They
16:30
help check those boxes off for you. And
16:32
one could argue that some of
16:34
this is quote unquote green washing.
16:37
Like are they really doing anything about it? But
16:39
it also makes it easy for us. Like, hey, just write
16:41
your code. Don't worry. We'll make sure it has
16:44
no impact on the planet. Right. But they
16:46
also bill us by the minute for the
16:48
compute. Like there is an angle on this. It says if
16:50
I write my code more efficiently,
16:53
it will actually cost me less. Right. You
16:55
know, when you own the computer and you don't
16:57
fully utilize it anyway, you don't care. You
17:00
have all the compute and you have enough compute for what you're doing.
17:02
But if you're charged by the transaction
17:05
and by the compute load, any
17:07
more efficient code, that's money in the bank. So
17:09
GitHub co-pilot, I'll go to say while
17:12
true and get a copilot.
17:15
Can't do that. That's a tight loop.
17:17
Yeah. I mean,
17:18
we do have serverless at the moment, like
17:20
all of these cloud providers, you can spin
17:24
up containers when
17:27
you need them, then you can scale them down and
17:30
shut them down. You can turn on
17:32
these lambdas. I
17:34
mean, even us, when we code,
17:36
we have massive problems of things taking
17:38
too long. True. And you develop
17:41
something for a kind of this very simple
17:43
things take, I don't know, five minutes
17:46
and then it scales up. You get a more complex
17:48
client. It takes two and a half hours for, let's
17:50
say, a batch process. You remove
17:52
it to strength, but it's
17:54
done tens of millions of times.
17:56
Yeah. It saves half an hour. Yeah.
17:58
No kidding.
17:59
There, you are not paying
18:02
for that half an hour of these,
18:04
I don't know, 300 containers being
18:07
gone.
18:07
Right. Or especially if there's
18:09
a big pipeline running, like you probably like fewer instances
18:11
too. Yeah, exactly. That is money you're
18:13
not spending if you do it well. I
18:15
think there is a direct
18:17
link from how much
18:19
you're paying something to how
18:22
much energy that
18:24
you use. Sure. You're basically
18:26
eliminating this infrastructure
18:28
that you don't need
18:29
to run.
18:31
So, there
18:33
are a couple of things. You can
18:36
just run these machines when
18:38
you need them, when you're running something
18:40
instead of constantly having them on.
18:42
So elasticity is cost efficient.
18:45
Exactly. That's one thing. The
18:47
other thing is write more performance code. Right.
18:50
So you don't need to have them on for an hour, maybe you
18:52
need them just for half an
18:53
hour. And that is good
18:56
for the environment, you're assuming less resources and
18:58
happens to save you money. Exactly.
19:01
But if you perform the same amount of
19:04
CPU processing that you would do in
19:06
a half an hour if you do that in five minutes, you're
19:09
jacking up the CPU, are you really saving
19:11
anything?
19:12
Maybe. I think that
19:14
really depends. I guess what I'm saying is the developers
19:16
need to think about not just going
19:18
faster but... It's really
19:20
kind of a compute cycle problem. I
19:23
like your lambda scenario
19:26
because typically when I see a spike load like
19:28
that, I've got a big old queue of calls. Like there's
19:30
a bunch of work that's just come in, we've got a whole bunch of stuff
19:32
to run. And
19:35
it is containers under the hood, we all know that,
19:37
right? We just don't have to deal with it. It lights them up
19:39
and to pump through as quickly as possible is going to light up a lot
19:42
of them. But if you've made them more efficient,
19:44
you'll literally light fewer. So
19:47
in the end, you're going to see how many
19:49
instances... It was the same workload but
19:51
how many parallel instances did it need to execute
19:54
it? And because it was finishing faster, it won't
19:56
have lit as many. The same
19:58
one will be reused more time. There's real
20:00
savings to be had there. I guess that's a
20:03
puzzle for everybody to individually figure
20:05
out. Sure. Like how to make things more efficient. Well,
20:07
and we played these map reduce games before
20:10
where it's like, hey, I can run map reduce on one machine.
20:13
Don't do that. It's dumb. But it'll take
20:15
an hour. I can run it on six
20:17
machines, and it'll take 12 minutes.
20:22
I can run it on 60 machines,
20:24
and it'll take four minutes. So
20:27
you know it's- So what's the environmental impact of running
20:29
it? You're exactly right. Like one would argue
20:31
the six machines was kind of the efficient middle point
20:33
where it took less time, and
20:36
the overhead of the mapping part and
20:38
the recombining part wasn't
20:40
as high as over distributing it, where
20:44
we may be able to cut the time in half, but we upped the
20:46
load by five times to
20:48
make it work. I don't know if you thought about this, Lab, but
20:50
I know Richard has. Do
20:52
you think quantum computing is
20:55
going to change the equation in terms of
20:58
sustainability
20:59
and
21:01
in our data centers? Can
21:04
you think of a quantum data center,
21:06
what that would look like?
21:08
I don't know that much about quantum
21:10
computing,
21:13
but there is an interesting example. My
21:16
brother, he's a physicist, and he
21:18
works with solid state physics
21:21
and kind of combination with chemistry and finding
21:23
new materials. And
21:25
this was an interesting thing. I was talking to him the
21:28
other day, and what he's trying to do
21:30
is use machine learning to find these
21:32
different- to write programs
21:35
that find these different combinations
21:37
of materials.
21:38
That
21:39
would otherwise be done with quantum
21:42
computing, which takes a lot of power,
21:44
which takes a lot of energy.
21:46
Mostly in the cooling.
21:48
Yeah. So
21:50
I can't say anything about quantum
21:52
computers, like how efficient they are, what
21:54
are they doing to their
21:55
environment? Well, they don't really exist yet, do they? I
21:57
think they kind of do, but they're within limitations.
23:59
the
24:00
my latest MacBook my M1. Yeah
24:03
the M1s and the M2s being some of the finest
24:05
machines that are made. So we used to have and
24:07
the battery lasts forever. And and well
24:10
in that chip
24:12
is an incredibly large chip it's
24:15
not actually that complex a chip because
24:18
the ARM architecture is simpler they have room
24:20
to include a GPU, include
24:23
an NPU, include more memory
24:25
all things that make it perform better. So we're
24:28
starting to see a push towards arms
24:31
in the server closets and in the cloud and
24:33
I think you're going to talk about double digit
24:35
decreases in energy consumption
24:37
for the same results. Wow.
24:39
And I think that's a good point like Pentium
24:42
is now talking like working on
24:44
chips that are low consumption.
24:47
They're no longer saying when marketing
24:49
like even Apple with their chips they're no
24:51
longer saying this is super fast in performance.
24:54
Like how fast do we need like you
24:56
know like we're
24:56
it's fast enough.
24:59
We are now in a world where it's like this
25:01
is more sustainable battery will like
25:04
it will use less energy. Right. It's
25:06
not just us like Apple like
25:08
all of these companies are also marketing
25:10
on this. When you're right that like arms pitch
25:13
has always been battery efficiency those M1s
25:15
and M2s all day battery in your laptop
25:17
and it's not a big battery. Yeah. The idea that
25:19
you put it into server because literally it'll be less
25:22
kilowatts is just another
25:24
level of the same thinking so we
25:26
should be able to get there. I mean you're
25:28
cool. Well inevitably because you're
25:30
literally moving the electrons around less.
25:33
The pipelines aren't as complicated as not as many instructions
25:36
said and in the end all that
25:38
heat comes from vibrating electrons. Well there
25:40
you go. So we don't have
25:42
to fix our code we're fine. Just you know .NET
25:46
CLR will automatically compile
25:48
onto ARM and we're gonna get the computer results.
25:51
I think there is also another lens
25:53
to it like we are using all
25:55
of these programs hopefully
25:58
to bring some value right. So what
26:00
problems are they solving? Like when we are talking about
26:03
machine learning, when we are talking about quantum
26:05
computing, when we are talking about any
26:07
software that we develop, or
26:09
even AI and crypto,
26:12
like not crypto, like blockchain,
26:15
and distributed systems, like
26:18
what problems are we solving?
26:21
How much energy they
26:23
use
26:25
might be okay if we are thinking
26:27
about the problems that we are solving, and maybe
26:30
those problems would be using
26:32
even more energy if we were not doing this.
26:35
So I think we need to kind of
26:37
look at the bigger picture, not in the small
26:39
and isolation, this is using X, Y, Z
26:42
amount of energy. It's like,
26:44
what is that replacing?
26:45
And the trick is the comparison,
26:48
because there's not going to be easy comparison. No. And
26:50
guys, I'm going to ask you to pause a moment
26:53
for these very important messages.
26:57
The era of automotive advances with
26:59
the all-electric Polestar 2. Now
27:01
with faster charging, improved EPA
27:03
estimated range of up to 320 miles, and advanced safety technology.
27:08
It's time you move up to Polestar 2.
27:10
What are you waiting for? Polestar 2.
27:13
Experience awe-inspiring performance combined
27:15
with luxury design as standard. The
27:17
time is now. The all-electric
27:20
Polestar 2. To learn more, book a test
27:22
drive and order today at PolestarColumbus.com.
27:25
Polestar 2. The time has come.
27:28
If you've been hurt by a truck, you
27:30
can call Colombo Law 24-7
27:32
and we'll be there to make sure you're taken care of. When
27:35
someone is hurt by a truck, Colombo Law
27:38
is the law firm people call to get answers.
27:40
Hurt by a truck? Call Colombo Law.
27:44
And we're back. It's .NET Rocks. I'm
27:47
Carl Franklin. That's my friend Richard Campbell. Hey. And
27:50
we're talking to Leah Mladenio. Yes,
27:53
I got it right again. I
27:55
got sustainable computing and like,
27:58
you know, there's so many. subtopics
28:01
under here but we're going for
28:30
you
28:35
know watts per transaction.
28:38
We can also be talking
28:40
about that greenwashing that you mentioned earlier
28:42
Richard. Yeah. And I think that's a
28:44
big pain. I think with anything that we do that's
28:46
a big pain like because people stop trusting
28:50
the numbers and their impact. Yeah.
28:53
If all of these companies are scamming
28:56
in a way. Yeah. Let's go over
28:58
that one. Because
29:00
then you stop trusting it. And even
29:02
if okay if we are thinking about where we
29:05
run our software, how we write our software
29:07
and how performant it is, we
29:10
can't reduce it to zero right? No. Like
29:12
we would stop living
29:15
or we would not write anything and
29:17
that's the best thing for planet.
29:19
Right. But I don't know if that's true
29:21
but yeah.
29:21
Okay. But you know like we can't
29:23
do that. So like it's also offsetting.
29:25
Yeah. There is a lot of greenwashing
29:27
there as well. So I
29:30
think our next step is as
29:32
well kind of removing
29:35
the fog around all of that or trying
29:38
to figure out.
29:39
Creating more transparency. More
29:41
transparency. Maybe standards
29:43
like it's electricity right?
29:45
Like we know how
29:47
to calculate electricity. Like. But
29:49
looking at the environmental impact of it if you have
29:51
an electrical source is challenging.
29:53
It depends on what you're using. I
29:55
mean I've had a chance to tour the Quincy Data Center
29:58
in Eastern Washington. but
30:01
you're really test sites for Azure,
30:03
but they're powered by hydroelectric power, which is about
30:05
as green as you're going to get. It's
30:08
not like hydroelectric power doesn't have environmental consequences.
30:11
You do flood a valley and a bunch of plant
30:13
life goes underwater and other things like
30:15
there are consequences. Yeah, whatever happens in that whole,
30:19
let's put a data center underwater. You
30:22
know, funny, I did a lot of... Speaking of green washing
30:24
or whatever. No, no, I talked
30:26
to Mark Resinovich about that. Okay. And
30:29
he was just a while back in the build timeframe and
30:32
we talked about Project Natick, which
30:34
was this test of putting essentially
30:36
racks in a metal cylinder under
30:39
the ocean to do the... Right.
30:42
Like take a shipping container, put all the servers in it, have
30:44
it water cooled, boom. It's a pressure
30:46
vessel, so it's cylindrical. Yeah. And
30:48
they tested it in the Orkneys and so forth. But in the end, it came down
30:51
to data
30:53
sovereignty controls, like just
30:55
being able to protect the machines. So...
30:59
Like who owns it? Well, not only who owns
31:01
it, who has access to it. Like one of the
31:03
problems with it being underwater is you don't know if people
31:05
are approaching it. Right. I
31:08
don't even notice, but we're living in a world lately where there's
31:10
a few bad state actors out there. Yeah, you're right
31:12
about that. This is exactly what Resinovich was talking
31:14
about. It's like, you know, we
31:17
like our big buildings
31:19
made of concrete with big fences
31:21
around them and
31:24
strict controls on access and
31:26
you can't do that when it's sitting at the bottom of a bay
31:28
in a city. Yeah. So,
31:31
in the end, it became impractical. The only thing they showed...
31:34
One of the things that came out of it that was really interesting was that the
31:36
hardware lasted longer when
31:39
the environment was completely undisturbed.
31:41
Wow. That the machine failure
31:43
rates were actually lower because
31:46
nobody walked by them there because
31:48
they were in a middle cylinder out of the ocean. That's interesting.
31:51
Did you ever hear about Subsea Cloud? That
31:54
was another company that planned to have
31:56
a commercially available undersea data center
32:00
operating off the US coast before the
32:02
end of 2022. Yeah, I don't think it made it.
32:04
That didn't happen apparently. No,
32:07
and I think you get back to the same old problem. It costs
32:09
more and it's questionable safety. The
32:13
ocean is a dangerous place and
32:16
a secretive place. So I think it's problematic
32:18
overall. There was a bunch of benefits to it without a
32:20
doubt. The logical
32:22
thing was you were going to be able to put them closer to cities where
32:25
the latency would be low in a space that
32:27
would be inexpensive. But
32:29
there are unusual regulations which you get
32:31
in the water that are hard to deal with. Yeah. And most
32:34
importantly, you don't have direct control over everything. Well,
32:36
probably the worst place to put a data center is above
32:38
ground, isn't it? Well, where else are you going to put it? Underground.
32:40
Yeah. Yeah,
32:43
as long as you can control the space around it. Right. And certainly
32:45
you get more heat management that way. I
32:47
think you see a lot of organizations
32:49
talking a lot about where they're getting their power from and
32:52
how they're cooling. I mean, the two big
32:55
energy consumers. And
32:57
for the most part, I want to... This
33:01
seems to be the logical thing for
33:03
us to do is we pick the provider whose
33:06
green approaches are
33:08
appropriate. I suspect
33:10
they're all going to be the same ultimately,
33:13
and they all have a certain amount
33:15
of green washing in them. And
33:19
we need to fight against it. I'm not saying it's acceptable,
33:22
just it's part of our responsibility
33:24
to be going. This part where you're buying all these carbon
33:27
offsets because you didn't actually say what you're going to do, which when
33:29
you have no carbon output on this, stop
33:31
it. Stop emitting carbon on it. But
33:33
that's the thing. As we
33:35
care more about
33:37
this, as we care more about the environment,
33:40
we are choosing those providers. Yeah. And
33:42
at the end, we'll stop choosing
33:45
those that we don't trust or that
33:47
are not doing a good thing, whatever we define
33:50
that that could think is or whatever the values are
33:52
and standards are. I think it'll get
33:54
better. Yeah. I think we just at the moment trying
33:56
a lot of different things and... people
34:00
are putting the numbers or ratings
34:03
that would make them look better.
34:04
I think that's important. It
34:07
is interesting to weigh the price
34:09
of building ultra-efficient code
34:12
that will consume as little energy as possible
34:14
versus the maintainability of that code.
34:17
Yes. My experience with making highly optimized
34:19
code is it's harder to take care of.
34:22
The simplest code is the best code
34:24
for maintainability, even if it is
34:26
going to consume more energy in the process.
34:28
There is always a balance, right? I
34:31
think we sway too much into the direction
34:33
of we want to scale, we want to fast, we don't care
34:35
about how we write the code.
34:37
I mean, honestly, Leah, I don't think people are thinking
34:40
about this for the most part.
34:40
Yeah. I get that. And
34:43
I think the first steps are to start
34:46
talking about it. Right. Get
34:48
the data on like what are you doing,
34:50
where you are. Because most
34:52
of this is we can do loads
34:55
of stuff. And I think we should
34:57
do loads of stuff now at the beginning when we
34:59
don't know what's the right thing. We
35:01
should experiment, we should learn from it. But
35:05
it's mostly to get the data and know what brings
35:07
the most value. Yeah, starting
35:08
to measure, just measure. Yeah.
35:10
I mean, data and metrics,
35:13
it's another thing to add to these
35:15
storages that don't impact
35:17
the environment in a good way. Yeah,
35:20
measuring is also a consumption of confusion. Exactly.
35:23
And storage. At least we can make an informed decision
35:26
on what we are doing. The thing
35:28
that we need to be mindful about is
35:30
that we shouldn't wait
35:33
to get something. We should all start
35:35
thinking about it and do something about it. Because
35:38
I don't think there's going to be a perfect solution. Yeah.
35:42
Nobody's perfect. No solution will be perfect. It's
35:44
a moving target. It's a moving target. But
35:46
everything is. This is a complex topic
35:49
that we need to solve. And at some
35:51
point, we will have better data. We will
35:53
have more information on what
35:56
is going to make
35:58
it all better or which direction we are. we want to go
36:00
in and there's going to be a couple of options.
36:03
But at the moment, let's
36:05
just try, let's do something.
36:07
I mean, I think the easiest thing I can
36:09
get by a senior
36:11
leadership is this. We have a SaaS
36:13
offering and I can reduce the cost of
36:15
us offering the SaaS offer for the same set of customers.
36:17
So effectively, we're going to make more money off
36:20
of existing customers by making this code more
36:22
efficient. And if you want to then
36:25
do a little green marketing on the back
36:27
of we make our code more efficient to reduce our
36:29
impact on the planet, that
36:31
fine, you know, up to you. Yes,
36:33
there is that thing
36:35
where we
36:37
need to be able to know
36:40
how to present
36:42
this work, which value this brings. There
36:45
is also that thing you want to work for the
36:47
company that wants to do good. And
36:50
if they don't understand what that
36:52
means, you want to explain
36:54
them. But if they don't get it and
36:56
if they don't want to do it, find
36:59
another job. Like we are
37:01
spending... Are
37:02
we going to get to that point where it's like, I can't work
37:04
here because you don't care about these things? I mean,
37:07
that's interesting to me. I don't know
37:09
if it's true.
37:10
Well,
37:12
there is an interesting, not interesting,
37:14
there is a phrase that
37:17
I was reminded by recently by my
37:19
CEO and it was, every
37:22
dollar you choose to spend is
37:24
a vote for the future you want to live in.
37:26
Exactly. I think
37:28
that's fair. And arguably, the pandemic
37:31
has made us more aware of the culture
37:33
we want to be in and a culture
37:36
of an organization that takes environmental
37:38
impact seriously is maybe a culture you want to
37:40
be a part of. So that does affect your
37:42
choice of where you want to work.
37:43
Yes. And I think
37:45
we shape it, right? I think
37:47
we choose and if our
37:51
employees are
37:52
not willing or don't care about
37:55
it, they need to start because people will not want
37:57
to live, they will not want to work there.
37:59
It's a signal about culture.
38:01
It's a signal. And it's a signal we want
38:03
to do something, right? And OK,
38:06
that's one goal is they
38:08
will not be able to get software engineers,
38:11
get software engineers. The other thing is these vendors,
38:13
they'll not be able to get the clients. Or the government
38:15
will say, they will start
38:18
charging you on your environmental
38:21
impact.
38:22
That's the other thing. It is interesting to get to
38:24
a point where it's like, can we get ratings
38:27
on our software for environmental
38:29
impact? I like that idea. It doesn't exist
38:31
at this point, obviously. But we certainly
38:34
talked about other rating systems
38:36
for quality. And
38:38
so could there be one that measures
38:40
this as well? Absolutely. It's
38:44
interesting just how unmeasured
38:46
our industry is. And
38:49
again, I think the cloud's a sort of workaround to this.
38:52
The byproduct of cloud is we've now concentrated
38:56
that responsibility down to a few organizations
38:59
that we can press against, for
39:01
better or worse. I think it's a client
39:03
conversation here, too. Like, I
39:06
saw the EU's talking
39:08
more about the maintainability
39:10
of smartphones. You already said
39:12
this earlier, Leo. This is all like, do you really need
39:14
more compute? It's like, do I need another camera
39:17
in my phone? I think three's enough, right?
39:21
The main reason I've replaced phones these days is because
39:23
the battery's starting to fail, because batteries
39:25
aren't replaceable. And now you see the EU coming
39:28
back to, we
39:30
used to have changeable batteries, right? And
39:34
could you tolerate a millimeter thicker on your
39:36
phone? But every few years now, you're going
39:38
to be able to take the old battery out, have it recycled, get
39:40
a new battery for your phone, and you can use it for longer.
39:42
Well, certainly, Apple doesn't want
39:45
that, do they? Because they can't sell you
39:47
a new phone. Well, I'm not
39:49
just going to lay that at the feet of Apple. All
39:51
manufacturers are in the business of selling
39:54
you new devices. But the cadence
39:56
is decreasing, right? People
39:59
aren't at it. getting every new iPhone. Maybe
40:01
they're getting every other. Certainly
40:04
over in enterprise land we were talking
40:06
about keeping workstations.
40:08
It used to be two years and that was
40:10
four years. Now we're talking five and six
40:12
years because A, they're
40:14
good enough. It's not like
40:17
the employee's productivity
40:19
is impaired. The only reason that if five
40:21
or six years we start wanting to swap out the machines
40:24
is we start having more problems with them.
40:26
I had my last workstation for over 10
40:28
years. Yeah. Over 10. This new
40:31
one is gonna last me a long time too. The
40:34
last set of servers I had when I had a server
40:36
closet because I don't have one anymore and I'm not unhappy
40:38
about that. That's greenwashing right there.
40:40
Right there. But that set of servers,
40:43
twice I changed out all the
40:45
hard drives and all the fans in it. Yeah. Like the
40:47
motherboards were good enough and
40:49
there's nothing wrong with the chassis but eventually
40:52
the bearings on the fans start to go and that will
40:54
kill a machine as it stops pumping air through well.
40:56
Well I have 32 solar panels on my roof
40:59
so I feel pretty smug right now. But
41:02
again we're thinking about in terms of when
41:05
we reduce impact, when we maintain
41:07
equipment for longer and there does seem
41:10
to be a general sense that
41:12
a lot of these devices are sufficiently performant.
41:14
We don't need to replace them to get anything from that
41:17
but they should be more repairable.
41:19
Yes. I've really hope we're going in that direction
41:21
because I mean we are a consumer.
41:24
How do you say like consumed? Yes. Yeah.
41:27
We have been trained
41:29
by a system that profits from us consuming
41:31
constantly. Exactly. They're consuming
41:33
constantly. We want everything like you
41:35
go on Amazon and you order things
41:38
immediately, clothes, gadgets.
41:40
Everything is I just want it now.
41:42
It doesn't need to be good quality because
41:45
I'm going to be dead. I'm going to get bored of
41:47
it in the month. And this is the thing
41:50
like these companies that produce
41:53
these things, they are not incentivized to make
41:55
it last
41:56
longer. But
41:58
that's what you get you end up with like regulations
42:00
say, oh no. Well, they forced iPhone
42:03
to use the USB-C port and
42:05
it's not like they couldn't do it and
42:08
there was much acclaim from everyone and now they've
42:10
also said stop shipping chargers with the phone.
42:12
Yeah. Because we've got enough chargers. I
42:14
have enough cables at home. I got a story
42:17
for you. So my second
42:19
marriage,
42:21
I inherited two step daughters and they're
42:23
wonderful bonus daughters but at
42:25
one point one of them came and said, mom are
42:28
we broke? And
42:30
she goes, no, why? Because
42:34
I haven't seen any Amazon boxes on
42:36
the front porch in like two weeks. Call!
42:43
Order something quickly.
42:48
Yeah, we
42:51
are just used to having everything
42:54
now. Yeah. And
42:57
the world is getting complicated, you know, like
42:59
all of this stuff, all of these frameworks,
43:02
all everything, you know, like run
43:04
this, ship logs there, I
43:06
want to filter on this. We talked about
43:08
observability yesterday. That's
43:10
more data. That's more, you know, like
43:13
I'm not saying any of that is bad. I'm
43:15
just saying what is the value that that's bringing?
43:18
What problems are we solving with that?
43:20
But you hit an interesting point. It's like there's
43:22
an awful lot of consumption going on without assessing
43:25
the value from it. Exactly. Like if you're gonna
43:27
do that level of logging, tell me you're looking at the
43:29
logs.
43:29
Yeah, why are you using them? Is it
43:32
just like, first of all, like do
43:34
you need it? What's the problem you're trying to solve?
43:36
The other thing is like, what's the tool for the
43:38
job? Yeah. And this is a thing like,
43:41
is that the machine learning? Is it these
43:44
AI? Is it the, can you run
43:46
this during the night when energy
43:48
consumption is lower? Can you
43:50
run it in a region where we
43:53
have greener energy?
43:54
Sure. Yeah, that's an interesting idea
43:56
that we shift workloads, time
43:58
and geography for a little bit.
43:59
lower impact. Again, I don't know
44:02
how much consideration that is right there, but it's an interesting question
44:04
to ask. Those are not
44:06
stunningly hard problems. No.
44:09
If you had the numbers in front of
44:10
you. Yeah. I think there are different ways.
44:12
It's just we do need to start talking
44:15
about them. We need to start
44:17
making it easier for people to do them.
44:19
Yeah. Make a path.
44:22
The path of least resistance is also
44:25
a path of optimal outcome. Yeah.
44:28
I buy it. It's an interesting conversation to have and
44:30
to kick upstairs. Yes. To
44:33
take to leadership and say, where does this fit into
44:35
our conversations? How are we thinking about this?
44:38
Again, as someone who's had
44:40
to convince leaders to do things for
44:43
many years, it's like, come with a dollar
44:45
figure. They understand that. Yeah.
44:48
Absolutely. You need to speak with the language
44:50
of the people you're talking to, so you need
44:53
to know your audience.
44:53
Yeah. If I can make
44:55
the CFO happy and the CEO happy
44:58
at the same time, I've got a good day. Well,
45:00
I can really tell you that in
45:03
front of us, we made this
45:05
platform greener by accident
45:08
at early stages. This was just
45:10
because we were trying to – not just because.
45:13
One of our problems was we couldn't scale
45:16
with the client sizes and the hardware
45:19
is just too expensive. By trying to reduce
45:22
the cost, we
45:24
actually made the platform greener. Nice. Scale
45:27
it up, scale it down when you're not using it, distribute
45:29
it in a different way, like those
45:32
different things. That's
45:34
a language that our CFO understands.
45:35
It's cheaper. Right. It was cheaper.
45:38
It should be cheaper. It's
45:40
awesome.
45:42
What's next for you? What's in your inbox?
45:44
What's next for me? Well, with
45:47
FundUps, I'm lucky to work in a company
45:50
that actually cares about this, that
45:52
does want to make a good impact and not just
45:54
profit. We
45:57
just re-architected our platform.
47:45
dot
48:00
com visit our website a
48:02
c o t n c r o si se
48:04
as dot com for rss feeds
48:07
downloads mobile apps comment
48:09
and access to the full archives going back
48:12
to show number one record in september
48:14
two thousand to make sure you
48:16
check out our sponsors they keep us in
48:18
business go write some code
48:21
see you next time
48:42
if you been hurt by a truck you
48:44
can call colombo twenty four seven
48:46
and will be there to make sure your taken care of
48:49
with someone's hurt by a truck colombo
48:51
law is the more from people call to get
48:53
answers for by a truck call
48:55
columbo wrong
48:57
with lucky land science think
48:59
much
48:59
just about anywhere dearly beloved
49:02
we are gathered here today to has anyone
49:04
seen the bride and groom
49:06
sorry we're here we were getting lucky
49:08
and the land l loss at the
49:09
time know lucky lane casino
49:12
with cash prizes that at a quicker than a guess
49:14
registered
49:15
in that case i pronounce you again
49:18
for free and like a landslide dot com
49:20
degree bonuses or rating no purchase
49:22
necessary void where prohibited i live in
49:24
terms and conditions were
49:26
like pretty down
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More