# Performance and Sustainability in Software Architecture

**Podcast:** The InfoQ Podcast
**Published:** 2026-02-02

## Transcript

If you're the kind of senior engineer, architect, or technical leader who people look to for what's next, QCon London is probably on your radar.
Join us in London from March 16th to the 19th, where we go deep on the topics that matter, like the architectures you've always wondered about, engineering productivity, and applying AI in the real world.
This isn't about trends for their own sake.
It's about getting practical insights from senior practitioners to help you make smarter calls on where to invest your time in tech.
With Software Changing Fast, QCon London is a conference that helps you lead the change.
Learn more at QCon London.com.
Hello everybody, I'm Olimpio Pope, an InfoQ editor.
And I have in front of me Erika Pizzani, the track host in April for QCon London.
And she had very important track in my opinion.
One that is important in general, but even more important these days.
And the name of the track was Performance and Sustainability, which seems to be two opposing words.
So Erica, please introduce yourself.
Hey, nice to meet you.
My name is Erica.
I'm based out of Toronto, Canada.
I'm a senior software developer at a FinTech company called Float Financial.
And as you just mentioned, I was lucky enough to be the track host for the performance and sustainability track at QCon London 2025.
And I'm part of the programming committee for the upcoming QCon London in 2026.
Congrats.
Happy to see you on that lineup.
I think it's important you had such wonderful people and such wonderful presentations coming from them.
Oh, thank you.
And what was interesting for me is that more than just performance and sustainability, which was obvious in the track name, we had also ethical aspects.
And that's something that usually falls behind.
How did you put all these things together?
What was the heuristic or what was the gut feeling that pushed you to choose these topics?
Yeah, so when I was reached out to to put together the track, what I knew right out of the gate was I wanted to try and cover all the different angles that you could look at performance and sustainability.
And I tried my best to represent that in obviously there's only like five or six talks that you can get in a given track.
But I tried to cover as many angles as I could in those five or six talks.
What I went in thinking was I wanted someone who could speak to the efforts of trying to encourage a mindset of performance and sustainability, that they're not at odds, that they're actually quite complementary at a leadership level within an organization.
And that's why I reached out to Ludi for her talk, which was like what I wish I knew when I started with Green IT.
I think a lot of the conversations in a great way are often grassroots-driven.
A lot of individual software developers are very keen on making sure that they're working toward building a better future in whatever context that looks like for them.
But what I felt was missing in the conversation was that perspective.
So I was really excited to have Ludy on board and to have that talk featured.
Obviously, AI has been a huge thing over the past couple years.
And I knew that that needed to be a very important part of that conversation in that track.
And so that's why I was like super stoked to have Jade as part of the track, because her work is not, you know, bigger is better when it comes to AI models and LLMs.
She focuses on creating very performant smaller models, and not just from the environmental standpoint of the amount of time and energy and money involved in like building these large LLMs, but also from like uh how are you compensating the people providing that training data?
It's pretty well known that these more popular LLMs like ChatGPT and Claude leveraged the internet.
There was a pretty well-known lawsuit, I think, with writers and publishers, I think from Australia in the case of Anthropic, because they basically downloaded their works off of a site.
And so Jade's talk touched a little bit on how to make it sustainable to get the training data that you need, and not just any training data, high quality training data, and compensate the people providing it.
And so I think the the different ways that she touched on that was a perspective that like me being based in North America, I don't often see.
I think maybe the same goes for Europe as well.
And so just having that fresh take was something that I was really excited to have and something that I was looking for and was initially having a hard time finding until I came across Jade and hoped that by reaching out to her, she would say yes, unfortunately she did.
And then yeah, one of the other angles was the local first software movement.
That's gained a lot of traction in and of itself, because of its focus on increased privacy for users that use the software, not having to rely on cloud services and network reliability is also something that, again, I'm based in a big city in North America.
I take that for granted sometimes, and that's not true for a lot of people in the world.
And so I wanted the local first software movement at the table in this track because they are not only showing what performance software can look like in unreliable network scenarios, but in some ways they are more sustainable because, and I touched on this when I had a previous talk on edge computing, there's a lot of energy consumed in transmitting data over the wire.
And so if you don't have to do that, or you do that as little as possible, that translates to a very beneficial and environmental sense outcome when you build software like that.
And so again, performance and sustainability in that context aren't at odds, they're actually quite complementary.
And then the other two folks, Sarah Sue, who is well known in the green software movement, she published a book alongside Sarah Bergman and Ann Corney on building green software.
And one of the challenges with moving toward more sustainable software is that you have to measure something.
You have to see if you're like anything else in software.
You can't fix what you can't measure, and you can't prove that things are moving in the right direction depending on the goals of the business if you don't have metrics.
And Sarah, I wanted her background and expertise in this track because she knows this in and out, obviously.
And then the last person I wanted to have was to take a more systems level look at things.
So, like all these different levels I tried to touch on the track, Polly was able to talk about all the servers that were running, all the data that we're collecting and really challenging that and showing not only the environmental impacts of let's say practices that we take for granted, like just take all the data that you need because you never know when you need it.
I loved that she showed and challenged how it can be very inefficient for businesses, not only in a monetary sense, but even the environmental impact that it has, where we're just storing all this data and consuming energy to store it and it's never ending up being used for any beneficial purpose.
Nice.
Well, I didn't think of it that holistically as nicely as you put it, but it makes sense because you have pretty much your uh low-level point of view where you're just looking at how to design software and to just think about it, what you actually need, then the whole ethical aspect, and I don't know if it was one of your intentions with having Jade there, but you actually made it global because at many points of time Africa is not represented.
And Jade brought in the attention of many people, both through her presentation, the recording that is on the website that is available, but also what through the podcast, because I had a follow-up discussion with her as well.
A lot of aspects that we don't think, and one of those is the bias that the whole internet has, and that's very focused on the Western society, mainly white middle-aged male speaking English, and that's something that needs to be changed, and I think that's very important.
And one of the lessons that I learned from the the discussion with Jade is that you don't have to have the biggest model, but the most accurate one, and that's close to what you actually want.
For instance, if you're, I don't know, trying to have NLP and natural language processing for Romanian, for instance.
If I just go and have it in Bulgarian, it will be tougher rather than having another Latin language.
So she has a bunch of nuggets of wisdom that are good to be heard.
And with Holly, I had a couple of conversations over the years about this and things that we take for granted, and it they just focus on our comfort.
For instance, I don't know, we are listening to music on YouTube and we have it in HD.
All the bandwidth that we use actually to just download those things on our machine while we're actually needing just the sound, those are just things that we might shift a small switch in our brain, and then things will be a lot easier.
And given the scale that these applications have, it's very impressive because we're talking about hundreds of millions of users that are usually online.
So yeah, it's important.
Thank you for pushing all these things together.
But the question goes, especially as you put it, you're looking to being part of the program committee for next year.
What are your lessons learned?
What would you try to bring to just have a teaser for what's next for QCon?
Because I suppose you have a couple of follow-up thoughts and something that you would like to see as a follow-up conversation for what you already put together.
Yeah, so in uh I I can't recall if it was already announced.
I think it might have been, but there's a couple tracks in particular that I'll be helping champion, although the programming committee works together to help put together all the tracks.
But the ones that I'm championing right now is architecting for resilience and organizations in flux.
And this is more like uh people and the process track navigating the AI hype, basically.
And while there isn't a performance and sustainability focused track in QCon, that's not to say that there are not people that could not be part of these tracks as speakers that could not incorporate these aspects into that track.
I think architecting for resilience is a good example of the track where that is possible, where again, going back to let's say like the local first software movement, if you're designing things that are offline first, they're not assuming that you will be connected to an internet, they're very resilient.
I guess in some ways it's focusing on distributed systems and performance and sustainable software tends to be by its nature distributed.
Yeah, I'm not sure I can't give too much uh in terms of like a sneak feed for upcoming speakers or anything because it's still early days, but that was not my intention, but it was more about your takeaways after the QCon and the such a nice uh lineup.
Well, no pressure.
No, no, no worries.
I guess my biggest takeaway from the track from the people who I managed to chat with that were attendees.
Some folks, even though the green software movement has been around for a few years, still aren't aware of that.
And when they see it and the talks given in the space, they get really excited that this is something they can incorporate in their day jobs if they so choose.
They don't need necessarily approval or a blessing from a manager or a higher up.
There's things that they can do in the design decisions that they make, in the choices of vendors that they can potentially make that allow them to help move the needle in the right direction as far as green software is concerned, while still being performance, because ultimately what business care about at the end of the day is like meeting their various goals, whether they're business, whether they're corporate responsibility ones that involve tracking emissions.
That's ultimately what they care about.
But yeah, the other element was that they just didn't know it existed.
And when they realized that, they want to be a part of it.
And for me, that was one of the bigger takeaways of being able to host that track and chat with people afterwards is just how excited people are that there is something that they can do to help fight the climate crisis in various ways, big and small.
For me, what we don't realize is the impact that we have, because most of us we are working on very efficient machines, but we don't really understand that over the wire we have huge machines.
Probably most of us didn't visit a proper data center, and we don't really understand that that's actually industrial.
We don't understand that the cable is plugged in a power grid that might burn fossil fuel a lot, and depending on where it goes, in total 2024.
Adrian Cockroff spoke about it, and he just presented his perspective after working for AWS a long period of time, and just providing the way how the world is split into regions where coal is burned.
And obviously, everything that was eased from let's say European Union, so from Romania, everything was dark.
There was nothing that much in solar.
Obviously, now things are changing also in India and China, but it's still a lot to recover.
But the Green Software Foundation, because you mentioned Sarah, which is obviously involved with them, and also we have their work in and their book that is about green software.
I kept in mind a couple of metrics that I keep on my slide to just have a picture.
I know that you provided more details, so I will leave those as takeaways.
But they have a nice poster from the 2023 report, the last one that they did, and at that point of time the total consumption of the software industry, it was equal to the transportation industry altogether, meaning plane, trains, ships, everything.
And that was really striking.
A lot of people were very surprised by that because how come we are so efficient?
Yes.
But again, our comfort, our need for that number of nines after the dot 99 availability is the cost that we have to pay.
And I think that's that's quite important.
Again, the comfort and the way as you put it as well, we as software developers can take decisions.
I don't know, can be very small decisions that actually translate when you're thinking about millions of records, translate into a silicon and that consumes power and so on and so forth.
And I think that's important that we as individuals can do a lot of stuff.
Obviously, a notion is formed from multiple drops of water, but if we just join together, we'll throw a form at the one-point notion, and it's about our decisions that we make.
But still, we're talking about performance as well.
And what was your takeaway?
Is it possible or isn't it possible to have performance?
Especially that you're looking for resilience and architecting for that when you're thinking about also the ESG and everything sustainable.
Yeah, I really believe we can.
And it obviously depends on what context we're talking about it in.
The things that the various speakers in the performance and sustainability track touched on highlight different aspects of this.
One thing that is more I think a little bit more challenging is like, for instance, edge computing is something that I believe coupled alongside local first software, it can have a beneficial performance and environmental impact.
But as you just pointed out, depending on what region of the world you're in, the energy that is created in that local area might not be as clean as let's say a data center in Montreal, which is entirely hydroelectric, it's one of the cleanest data centers in the world, or in Scandinavia where it has geothermal.
Measuring that can be tricky.
And I think I would probably have to do a PhD in order to really measure that effectively.
But I think generally speaking, high performance can tie into do you need to transmit as much over the wire?
Do you need a lot of data to do what you need to do locally?
If you can get away with not needing that, obviously that's performant for the user, it's performant for the business because they maybe don't have to pull all that data together on the server before sending it over.
We saw with Jade's talk for the AI models, smaller can be very effective and highly performant.
It may not need to take as much time to think things over before giving an answer.
If it's trained with the right quality data, it's maybe a model that's very focused on a subset of problems rather than trying to be something for everyone, the way that a more general purpose model is.
But yeah, I don't think those two things are at odds.
I think that at this point there's been enough time over the years and enough great people working in different parts of the software industry that have highlighted that even like distributed systems, like we talked about, resilient systems, these things can be done in a more environmental way than maybe we have previously.
And it's just a matter of thinking more creatively about the architecture that we're starting with.
And that's something that maybe by default we think big data center as opposed to a local first piece of software that occasionally syncs with the central server.
Yeah, I spoke over the last four or five years with multiple people looking from the Cyber Resilience Act that came to Europe and put some guardrails or some rules in place for people that are writing software, foundations or organizations.
And in those conversations, we agreed upon that software is coming of age.
We are not a small industry anymore.
We are actually an industry that is all over the place.
We are powering a lot of other industries.
And that's important to be seen as well at this point because, as you said, the computing on the edge is very important.
Actually, I saw a post from somebody from MIT the other day.
It came, I think, in Forbes magazine.
Edge computing is very important, and probably a lot of the decisions also by AI will be taken on the gadget itself and not sent over the wire.
And usually at that point, I have itchy fingers or itchy tongue, and I have to say something.
What is underlining is the fact that if we manage to do that and have that kind of mentality in each and every architectural decision that we take, it would be much better because if you just open a basic application on your phone, like I don't know, Instagram, maybe, and you see that if you unplug the phone, you already don't have anything to be done.
You just have a pop-up.
And those things are just plain nonsense because you just have a lot of things cached locally, so it's pointless to not use them.
But I think it's also a matter of understanding more.
And that pushes me to think about when I was a kid, my parents were always encouraging me to think up front and think in terms of years, and not only what I like now.
And they always were saying that we are too poor to buy something cheap.
And I think that those are the decisions that we as software developers have to take again because at point it feels that we are just following the shiny object and we don't actually need it.
And I think another mind shift is needed in the industry where we just take the proper decision, we just think up front and we just bring everything together.
And on top of what you you're saying earlier about the human factor in the software development, even nowadays we know that we don't need any humans to write software.
AI is gonna do it.
The AWS outage proved it that uh pushing people out is very helpful and uh half an hour outage.
No big deal, yeah.
Yeah, you can see that.
All in all I think the social technical factor is very important.
Because if you look at everything, it becomes a lot more I was discussing the other days with Luca Metsalira, which is one of the persons that we are pushing for microfront ends.
And he was pretty much saying the same thing that Sam Newman said in his podcast when you distributed systems that when you're building a distributed system you're actually building the team for it as well.
And its architecture is resembling the structure of the teams and the way how we do it, mainly focusing on autonomy of those those teams.
And I think all these things tied together, and just we are all living under the same sky, and we should care about it because global warming is called global for one reason because it's happening everywhere.
To add to that, I I think there's a term that I I'm trying to remember if I read it in a book or I saw it in an article or something, but a term that I think about when, like for what you're talking about is ethical debt.
It's a little bit of like you just want to get something going first and you'll deal with the consequences later.
And I think that mindset of growth at all costs, because the business is trying to survive, and you're deferring these more important decisions to later, they result in this accumulation of ethical debt.
And that ethical debt can look like environmental because you you're wasting a lot of resources.
It can look like a lot of other things.
And I think that if we take the time, because it doesn't take that much time to sit and think for a few minutes or collaborate with other folks, to think about like what could this look like in seven years, let's say, rather than in the next seven days.
There might be some decisions that we can make early on that will prevent a significant accrual of ethical debt and set the software up and the business up long term for success.
So I 100% agree with you that we are at a place right now where software developers, given the role that we play in so many industries, the software that we're designing, to some extent we can't just say it's not our decision to make because we do make a lot of small decisions day in and day out from our vendors, from the decision to hold on to data, from the architectures that we're proposing that can help move the needle in a better way.
That's very true what you're saying.
And the rule of thumb that I was just thinking while listening to you is something that I think came from Amazon that was saying that you have two types of decisions.
Decisions that are easy to change in the past, or it's a one-way door that you'll never be able to come back.
And I think that's the lens that you have to look into because something can be agile, and then we can just remake it, learn from it, and then move forward, or it's something that we really have to put into experiment and understand exactly what's going on.
And the other thing that was crystallizing while listening to you was the fact that craftsmanship is very important, even though we were looking into um this era of high speed where we are discussing about AI and how fast can we do everything and then availability, and now we're not discussing about as we did a couple of years back when you had beta testing, and then you had a very small number of users that use now.
You have to go back in the open.
Just think about open AI, how they did it.
They just put it up front, and then just slowly iterated over different versions of the software, and now they reached almost a billion active users.
So that's unthinkable from other points of time.
But still, I think what makes me happy is the perspective that in important events and people like you are putting up front a lot of the more ethical aspects or the social aspects that somebody that has access to the right data can put up front a service that can be used by many people and through the internet.
We become a global village, and we can actually have an impact.
And your small pet project.
Oh my god.
Come on, I'm I'm in Romania on the other side of the world, and I really liked your how as my grocery is it a my local grocery.
And the fact that bringing together open data and open data is pushed in more and more countries, that means people have access to data.
Or if we don't, for instance, I was trying to figure out a couple of things the other day, and you have a lot of data on maps, on Google Maps, and then aggregating all those things in a nice format will allow people to take the right decisions, informed.
And we can understand as I said, that if more people are joined, then actually things might happen.
Yeah, and I think that in building that project, I was a bit inspired by um, I believe her name is Maggie Appleton, and she did this talk on home cooked software and barefoot developers and talking about how there's a certain type of software where people are making things very tailored to their communities.
They won't get a lot of VC funding because they're focusing on very specific things to their communities.
But these pieces of software can have amazing impacts.
In the case of my local grocer, for the listeners, the context of this was Canada has a number of different monopolies, grocery chains being one of them.
And I want to try and put my dollars to a local mom and pop shop that was maybe independently owned.
And funnily enough, there is a grocery chain called My Independent Grocery that belongs to one of these monopolies, and I was so frustrated by that that I wanted to see if there was a way to make a map of all the local mom and pop shops in Toronto.
I ended up picking up and putting it down over a few years, but I was so angry that this existed that I was trying to see if I could make that data available.
And obviously, there's not going to be a big VC check at the end of something like this, but I knew it was important to me.
And I know that there's other people in my community that also would want to do something similar.
That this is very much the definition of like a home-cooked piece of software.
And to add to this, I guess like tying back to this big AI boom.
I think one of the benefits of it is it makes it much easier for people with less technical backgrounds to build software like this.
And while there's some environmental considerations with training these large LLMs, this is one of the positive side effects that it can have of allowing people to tackle these very local problems that big VC companies won't because they just won't get the return on investment that they would be getting.
There was a project in Austria the other year or two years ago that I read about.
The government was complaining that they cannot do nothing about the pricing.
There was a monopoly by big chains, and what they did is they were mechanically changing the price.
They just aligned on a price, and then obviously everything was across the board in the same price range.
A good intentioned software developer got frustrated, and he put together a tool that was just scrapping all the catalogs online from all the big names in the retail space, and then he was publishing everything so that you can go to the website and just check it.
And that had a lot of traction and translated through a lot of stars in GitHub, and obviously that moved things in a whole different direction.
I think that underlies the fact that we as consumers and also software developers have something to say.
And it just pins in my head.
I mentioned it to you before, but now I'll just put it all on this as well.
Is the concept of donut economics?
Because what everybody's pushing for these days, and as you said, the VCs are always looking for the big boom and the huge multiplier, and they just want to have everything.
We are just pushing for growth regardless of what they think.
But everything has an ending, and we have to figure out what we actually want to optimize for.
As we know from distributed system, we cannot have everything.
You know, there are all these popular memes that are we don't have any planet B, so we are not able to move anywhere else.
So I think that's important.
Maybe a shift of mindset.
How we want to do it.
It's we don't need that much for just living properly.
We just good quality things that are ethical.
And the other thing that was quite interesting is that Savannah Kunovsky, if I hope I don't mess up her name, she had uh one of the keynotes you know during the the QCon this year, and I had a conversation with her as well.
And the project that stuck in my head was the talking fans.
They use generative AI to make a conversational story with a pair of pants.
When you just go and buy a pair of jeans, you have the story behind it, what kind of materials, who worked on it, and so on and so forth.
So I think we are on the two sides of the door.
We have to do the things that you said.
Take the green software principles, put them in place, but also if we as consumers to stray through to our beliefs and do something about what actually we have.
Great.
Is there anything that I should have asked you, but I didn't?
I think maybe the question that I would say is missing, maybe just like as a nice way to wrap us up, too, is if people are interested in trying to make more sustainable software.
Either this is the first time they're hearing about it, or they've heard about it, but are maybe not sure where to start.
They don't know maybe what resources are available or what they can do in their day-to-day because they're working in a legacy system and they can't make these big decisions around like how much data to store and stuff like that.
I think fortunately for us, there's been so many people that have done amazing work over the past number of years to make some resources freely available.
The Green Software Foundation being one of them.
They I think even have like a free course offered through the Linux Foundation that touches on different aspects to consider.
I think it's also worth taking a look at the local first software movement that's gaining a lot of traction.
There's a lot of really cool tools out there, especially the research lab Inc and Switch.
They have like a lot of different components of what can be used in local for software that maybe you can't necessarily apply in your day job, but if you have a fun side project idea or you see an opportunity in your day-to-day of a future thing that's coming up that might lend itself really well to such an architecture, you can take a look at it, see if it works for you, and maybe propose it as something you can do.
I would encourage folks to, like any skeptical software developer, really question the assumptions that are baked into everything that we're talking about.
AIs have their use, but we don't need to apply them to everything.
They're not a silver bullet to every single problem.
And there are some cases where just your regular programmer puts something together, maybe with the assistance of Cloud Code or OpenAI Codecs, that can work really well.
I think I just would remind people that performance and sustainability are not at odds.
They actually can be quite complementary.
And there's a lot of resources out there that can help you on that journey.
Thank you.
And to supplement what you mentioned is I remembered at some point when I was building something after we just understood that the computing power cannot double every second year or whatever was more slow.
But I started thinking a little bit out of the box.
And that's the creativity that we're looking for.
Because when you're just building the same software over and over again, at some point you feel that you're missing the train and you're just getting rusty.
But at some point, I was just looking outside of the box and I said, okay, I'll start applying embedded software principles where you have limited space, limited computing power to cloud software.
And that allows you to think outside of the box and have better algorithms and stuff that actually make the difference.
And as you put it, then you can also see the impact on there.
I know there are a couple of JavaScript libraries that also try and put the things together in terms of how much carbon do you actually use.
So I think the message would be: we are a global village, thank you so much for having me.