distorted pollyanna

Overview of Ties' thesis work. Preprint of the thesis can be found here.

Note: Yesterday my first PhD student, Ties Robroek, defended his thesis! I said some words about him after the defense. The text below was my write-up to be prepared for those words. The speech itself was of course slightly different.

I sent the email offering my first independent PhD position to Ties on June 6^th, 2021, at 11:07am. At 12:10pm, I received a reply that said “… I accept your offer …” with a level of enthusiasm that made me think “No one has ever been this excited to hear that they will work with me. That is nice.” Then, I made a mental note “He has to learn not to accept any offer before seeing the contract first.”

 Ties started on September 1^st, 2021, less than two weeks after I was done being the general co-chair of VLDB 2021 and 3.5 years after I joined ITU.

I finally had what I wanted, my own PhD student.  I was very happy, but as is often the case when you are doing something for the first time, I wasn’t sure about many of the things I was doing. Of course, I had supervised many other BSc/MSc/PhD students before, but, with Ties, for the first time I had full responsibility for someone else at work. It felt like I could make or break his career. Furthermore, the project was on a topic that was relatively new to me, ML not databases, which added to the feeling of unsureness.

But here we are today, so it all worked out.

I think there were two key reasons why it worked out.

First, we managed to have an open dialog, especially when it came to our doubts. We didn’t agree all the time, but that never hindered the open dialog.

Second, Ties took initiative early on. From the beginning, his attitude was “You know hardware, but not machine learning, so I will work to be the one providing the machine learning expertise in the team.”  Luckily, he is a data scientist at heart, so embracing that responsibility came naturally to him.

Overall, Ties has been great at combining his data scientist identity with computer systems research. One of the first things he did was to create the radT platform, on top of MLFlow. I didn’t tell him to do that, at least not explicitly. Based on what we discussed, it was clear that we needed a systematic way to do benchmarking and collecting experimental data, and radT was his solution for that, which later became the first contribution in his thesis. Since then, radT kept evolving and is used in several research (BSc/MSc/PhD-level) projects hosted by our research group.

Furthermore, while implementing radT and any other codebase that later became part of his thesis, Ties’ primary design goal was always “I want a data scientist to use this, so it needs to come with minimal impact to their code.” I appreciate this perspective, and we “computer systems” people need to remind this to ourselves regularly.

Also, some fun stuff.

When someone shared the meme below with me in a group chat, the first person I thought of was Ties, so today he will get a card with this image on.

Note: I don’t know the original source for this picture, but a quick search led me to this LinkedIn post.

Speaking of cards, once I asked Ties to get a card for a lab member’s departure. Ties kindly accepted the request, but asked “So, your generation, whichever letter you are, where do you buy cards?” (If, like Ties, you don’t know the answer to this question; supermarket works.)

Earlier this year, Ties made me realize that I had been mispronouncing the word gouda. 

If you ask Ties “What do you recommend doing in Amsterdam?”, he will likely give you the answer “You can take the train to another town.”

Finally, we all have different ways of dealing with stress. While I was writing my thesis, I took a walk almost every day near Lac Leman, usually listening to Tori Amos’ Scarlet’s Walk. In contrast, Ties did a triathlon and a DHL run within the last two weeks of his thesis submission deadline. At one point I was worried that he will kill himself before submitting his thesis. Then, one week after his thesis submission, he did something called Nordic Race. Looking at its website, Nordic Race seems like an event designed to help people enter Valhalla.

Anyway, I am glad he is still alive and is now a Dr.

The first time I attended a Dagstuhl workshop was in 2017.

From my experience, people who aren’t computer scientists, even in Germany, don’t know what or where Dagstuhl is. Thus, first an intro.

Dagstuhl is a, not-that-easy-to-reach, place in Germany that allows computer scientists to organize workshops quite cheaply. The workshops are invitation-only, and the crowd is kept to ~60 people. The goal is to focus on a research topic with experts for either half a week or a whole week. It is a big deal to get invited to one. Therefore, I was quite excited to attend one.

The first day of the workshop, we had a session, where people could sign up to give a short presentation on a topic that would potentially trigger discussion among the attendees. I did sign up. I gave my presentation. Except for a couple comments, the reactions weren’t that encouraging, and I received a somewhat personal, non-technical, negative comment as well.

For the purpose of this post, the topic of my presentation, the man who shouted out the personal comment, or the comment itself doesn’t matter. This man apologized to me the next day, which I appreciated. I have received nothing but respect from him since then, and he is someone I have high respect for as well.

The important thing is how I felt after that presentation, especially as a result of that comment.

I wished the week to be over and to get out of there. I was annoyed that Dagstuhl is out of nowhere, which is on purpose, so that people have very little distractions and can focus on thinking “big thoughts”. But if I could spend a few hours away, in a city, I could have more easily cleared my head and found focus again. I thought I would never want to attend another Dagstuhl workshop. I felt alone. I didn’t want to talk to anyone. I left the evening social activities early to lock myself in my room. I over-drank one evening. My plan was to visit my cousin and his family, who live in Stuttgart, after Dagstuhl. I looked forward to seeing my cousin, a man, whom I had a very different relationship to, one that didn’t require any armor, compared to the men who surrounded me in Dagstuhl.

The next time someone reached out to me for a Dagstuhl workshop, it was 2019. I was asked by Goetz Graefe and Caetano Sauer to co-organize one, also together with Allison Lee and my academic sister Renata Borovica. That was hard to say “no” to, and that was the workshop that healed my relationship with Dagstuhl and helped me to get rid of most of my armor in similar work events.

With the distance of the years, I can now say that most of the comments after that presentation, even though not super encouraging, helped me develop better technical arguments and realize what I should have made clearer while presenting that particular topic. That is the thing with the technical comments. Even if they may sound negative or aren’t expressed in a constructive way, they can still help you improve yourself for the next presentation, discussion, paper or grant submission, project, system design ... This is one of the reasons why I love giving presentations and receiving comments and questions on them.

However, I cannot play pretend and put a positive spin on that personal comment even today. It was not ok. But was it really “I want to get out of here” bad? Would it make the same kind of impact on me today? Why did I have such a “thin skin” back then?

Earlier this year, I read (or audibled to be more exact) Roxane Gay’s Opinions: A Decade of Arguments, Criticism, and Minding Other People's Business.

Roxane Gay is one of my favorite authors. I once met her in Copenhagen airport. We were on the same flight. I saw her sitting in the waiting area of our gate, but no one else seemed to realize that she was there. I approached her and asked, “Are you Roxane Gay?”, to be sure. She confirmed. I told her that I really loved her writing. She thanked me and asked me for my name. Then, we shook hands. That was pretty much it. I preferred to keep the interaction brief, partly because I didn’t want to bother her in her private time and partly because I thought my heart was about to find its way out of my body based on the way it was pounding.

In Opinions, Roxane Gay writes about the “thinning of the skin”. Her argument is if a person or a group of people, such as minorities, are often subjected to comments or interactions that are hurtful or painful, their skin will become thinner. Hence, they will be more sensitive to any comment and may have more triggers. This argument contradicts the popular belief, at least the one I more often heard, that the more you face harsh comments or painful events the thicker your skin gets. She also questions the glorification of thick skin pointing out that thick skin implies the lack of feeling emotions.

This year, our department had its annual retreat in September, while I was in California for the HPTS workshop. In these retreats, we can propose to run a session on a topic we deem important such as well-being of PhD students, grant writing, finding collaborators ...

One of our colleagues, Louise Meier Carlsen, proposed to run a session on Advancing Gender Equality in (everyday) Academia. She asked the female faculty, postdocs, and PhD students to anonymously share some of the things we endured that can be put under categories such as harsher comments, dismissal, isolation, sexual behavior, etc. at work. Her goal was to avoid potential dismissive comments such as “these are isolated cases” by collecting data from a variety of women in the department.

On the last day of HPTS, I took some time to write about my experiences including the one from my first Dagstuhl workshop. I was grateful to Louise for running this session. It was helpful to write but also grim to realize that I could write so many things in such a short time. Similarly, it was both helpful and grim to read what others wrote.

Do I have thicker skin now compared to my pre-30s?

While I will never be a poster child for thick skin, I think the answer to this question is yes.

Did I develop that thicker skin because of all the shit I wrote about for Louise’s session? Or did I develop it thanks to all the people who supported or acknowledged me on the way making me realize that there are people who believe in and respect me regardless of those discouraging personal comments and interactions?

Before reading Opinions, I believed, somewhat uncomfortably, the former, because that is what I was taught by society, but the latter makes a lot more sense now.

Last Thursday, I flew to Amsterdam for Tim Gubner's PhD defense. The defense was on Friday. I booked an extra day for myself in the city, since I like Amsterdam and I thought it would be a nice break after the extra-busy September.

I have been to Amsterdam many times at this point. This time, I did two things differently.

Firstly, for the stay, I picked a part of the city that I haven’t stayed before to explore something different, meaning I chose a hotel that was not 10-20mins walk to the central station. This choice came with the extra benefit of an additional m2 at a cheaper price for my stay while still being in a central location. 

Second, I didn't inform any of my friends living in the area, since I needed some time for myself, like it is expressed in that beautiful Lucinda Williams song Side of the Road. 

And I had a lovely day in Amsterdam.

Then, it was time to go back. From my hotel, the easiest route to the airport was taking bus 397 from Museumplein. My bus ride on the way to the hotel from the airport was ~30mins.

I was at the bus stop around 16:00. There seemed to be some delay in the bus schedule according to Google Maps, and I could see the traffic in front of us, but the cars still moved, and the bus came earlier than what Google predicted, so I got on it ~16:10. The bus moved slowly but steadily, so I decided to chill and turned on audible to continue listening to Kelly Bishop's memoir, The Third Gilmore Girl.

From the memoir, I learned that Kelly Bishop is a Tony-award-winning actress and dancer who has been rocking being a childless cat & dog lady since at least the 60s. As a Gilmore Girls nerd, I knew her as one of the pinnacles of the Amy Sherman-Palladino TV-universe, most notably for her role as Emily Gilmore, the mother to Lorelai Gilmore and grandmother to Rory Gilmore. The first time I watched the show, the mother-daughter relationship I focused on was the one between Lorelai and Rory. As I got older, in my Gilmore Girls rewatches, I became more drawn to the relationship between Emily and Lorelai and highly appreciated the existence of a matriarch like Emily on TV.  

Emily Gilmore is also part of my TV battle crew, which, in addition to Emily, includes Sister Michael from Derry Girls, Olenna Tyrell from Game of Thrones, Xena from Xena: The Warrior Princess, Buffy from Buffy the Vampire Slayer, and Villanelle from Killing Eve.

Let's get back to the bus, which stopped moving somewhere after two stops. It was 16:45. I had been in the bus for more than 30mins at this point.

My flight was at 18:45. I prefer being at the airport at least two hours before for within-EU-flights, since there are no passport and visa checks. I aim for three hours, otherwise. In other words, I was already late based on what I was aiming for. While there was no reason to panic yet, it was time to put Kelly and my battle crew on pause and start thinking about alternatives.

I checked Google Maps again, but it wasn't able to show me a solution. There was a tram stuck in front of us and heavy traffic left and right. It was too far to walk to the Central Station at this point (~1hour), and I wasn't sure which other station I could check. I didn't know how to call a taxi in Amsterdam, but even if I did, I wasn't sure how a taxi could enter that mess or where to walk to in order to call one. In the meantime, we managed to move a little bit, but not by much. There were no announcements by the driver. Some asked the driver to open the doors so they could leave, but all the people with luggage remained in the bus, so it wasn't only me who didn't know a better way to get to the airport.

I decided that it was time to ask for help. I approached the closest person with luggage, who was a young lady (will call her Helper Sister from now on): "Do you know any other way to get to the airport?" She shook her head and said: "No." She looked worried.

More passengers became audible asking similar questions. Some started to move toward the front of the bus, and Helper Sister and I followed. Helper Sister started chatting with another lady in Dutch. Afterwards, she told me we would get off the bus and call a taxi.

Getting off the bus took some time, since the check-in/out system of the bus was disabled and people wanted to check-out before getting out. Eventually, people simply accepted that they would have to get out of the bus without checking out.

As soon as we got out, the tram in front of us started moving. The bus door was still open, so we got back in, and the bus started moving, but only to get stuck again in a short-while. 

At this point, I started thinking "They probably have another flight to Copenhagen in the evening. It is Saturday night, it shouldn't be difficult to find space in it. In the worst case, I'll stay one more night in Amsterdam." Helper Sister was on the phone talking to someone. The driver got off the bus to talk to some other drivers who were stuck in their cars behind us and came back with no news about what was going on.

Then, Helper Sister ended her phone call and told me: "I will leave and try something else, you can come with me." I replied: "OK."

It was ~17:15. We got off the bus. She mentioned something that I didn't fully get, but I remember hearing "I think it will be faster. But we need to walk." I said, "I am fine with walking." I blindly followed her as we alternated between fast walking and running. When we were at a spot where our side of the road was empty, meaning the cars could move, she started hitchhiking. It was the second time in my life I was hitchhiking. The first car we saw stopped. It was a mother and daughter in one of these small cars that requires the people in the front to exit the car in order for people to sit in the back. We did that at the side of the road and took our luggage on our lap. My luggage was a carry-on, but Helper Sister's wasn't.

Helper Sister told them directions in Dutch while in parallel explaining the situation. At least that is what I assume, as I can't really comprehend Dutch. We started to approach an Amsterdam Zuid sign, and I finally got the plan.

Where we got stuck in the bus was south of the city, which was closer to the Amsterdam Zuid station, where one can take a train to the airport. I don't know if hitchhiking was always part of the plan or if Helper Sister assumed we can walk/run to Amsterdam Zuid and improvised on the way.

We thanked the mother and daughter, they were lovely, made it to the 17:35 train, and were at the airport at 17:45. I was at my gate at 18:05 and ate a gevulde for dinner.

Helper Sister and I went our separate ways once at the airport since we were on different flights, but I thanked her a bunch before that. I don't know her name, she doesn't know mine either. The only thing I know about her is that she had a flight to Heraklion at 18:40 on that Saturday and she can speak Dutch. I don't know why she picked me from that bus full of passengers. Was it simply because I reached out to her for help? In any case, I am very grateful. I hope she had or have been having the time of her life in Heraklion.

Breakfast has always been my favorite meal of the day.

When I was a kid, my dad prepared the breakfast table for us, unless he was extremely ill or away. He is still the one who prepares breakfast at home. This is simply because he wakes up really early (~5am) as opposed to my mom who prefers sleeping longer.

After I moved away from the home my parents gave me, and lost the privilege of being served breakfast, I very rarely skipped it even if I had to rush it on some weekdays. In contrast, I don’t mind skipping lunch or dinner on a busy day.

After I moved abroad, I realized that not every country gives as much importance to breakfast as my home country. Then, breakfast became the main meal that brought me back together with family and friends.

Since the beginning of this year, I have been experimenting with writing a play that takes place around a breakfast table that brings back old friends. I have worked on it one evening each month on average. It has been evolving, but it is still a baby, and it will keep evolving. A part of the current draft will be read in public on August 28th at LiteraturHaus. The event is free, starts at 8pm, and also includes the works of other creatives that I had the chance to meet during this year in a writing course led by Paul Gordon. I cannot vouch for the quality of my work, but I heard excerpts from the works of others, and I can vouch for them. 

The Facebook link for the event is here.

The ticket link is here (it is free, this is just to keep the count).

Early in the movie Oppenheimer, Oppenheimer is asked whether he was happier in Cambridge, UK compared to in his home country, USA. He answers, “No. I was homesick and emotionally immature ...”

I watched Oppenheimer at delphi LUX during my sabbatical in Berlin to complete my Barbenheimer, more than a month after I watched Barbie at Cinema Victoria in Cluj after VLDB Summer School. (I watched Barbie two more times afterward.) As soon as I heard Oppenheimer’s answer, I searched for my notepad and pen in my bag in the dark of the movie theater. I scribbled the answer in my notepad without seeing what I was writing. It was an answer I could deeply relate to when I think of the years I lived in Lausanne, Switzerland.

Whenever I visit Turkey, it is common for me to receive a question that starts with “Do you miss …”

Do I miss Turkey / Istanbul? No. I miss my parents, close family members, and friends, but not the place. 

Do I miss Zonguldak (my hometown)? No. I visited Zonguldak once in 15 years. Most of my family and friends don’t live there anymore.

Do I miss the food in Turkey? No. My eating habits used to be mocked by the fellow Turkish PhD students at EPFL for not being Turkish-like. This topic even found its way into my PhD defense ceremony video prepared by close Turkish friends. Also, I don’t need to be in Turkey to prepare the food I like for myself, since I am not picky about having the exact ingredients. Finally, as my dad likes to say, “The important thing is not the food on the table but the people around it!”

This doesn’t mean that I don’t love and care about Turkey or I don’t like visiting it and eating the food there. At this point in my life, being able to speak with people who can comprehend what I say in my native language is a luxury, and, sure, the tomatoes taste much better in Turkey. The difference is I don’t long for places or food anymore the way I did back when I lived in Switzerland. I only feel homesick for the people, and my people are scattered around many places.

I wrote about Home back when I lived in San Jose. It was my time there that made me redefine homesickness. First, San Jose became a generous home for me over time; it made me realize that I can have homes at other countries. Second, its distance to Turkey, compared to Switzerland, and the longer stretch of time I spent there without visiting Turkey forced me to look at the bigger picture; I can bare not visiting Turkey if I have to, but I need to find ways to visit the people that are part of my home or to have them over for visits.

These days, I am lucky that I get to feel like I arrived home in Copenhagen, Istanbul, Chicago, Paris, London, Stuttgart, Munich, and maybe even Zurich. Except for Copenhagen, this is because of the people that welcome me in these locations, value my presence regardless of my work affiliation, and give me a place to crash and let go of my armor.

During my sabbatical last year, I visited Copenhagen twice. The first visit was roughly after a month of stay in Berlin. It wasn’t because I missed Copenhagen. I wanted to attend ITU’s Teaching Award ceremony, since I was the recipient of the award that year. It was also a good excuse to see Sister in Movies and other buddies in Copenhagen. After the Hamburg-Copenhagen train passed the Danish border, I started to hear more and more Danish around me compared to German. This triggered an unexpected feeling of calm. I was back in the land where I built another home.

When I lived in Lausanne, whenever I arrived in Geneva airport, I felt disappointment.

When I lived in San Jose, whenever I was coming back to US from Turkey, I was upset at the airport in Istanbul but was fine by the time I landed in San Francisco.

Now, I look forward to arriving in both Istanbul and Copenhagen.

I recently re-read A Game of You; the 5^th volume of The Sandman series. It is hard to pick a favorite from The Sandman, but if I absolutely had to, I would pick A Game of You, which also has a Barbie character. It was that Barbie that I quoted in the Home post: “I don’t think home is a place anymore. I think it’s a state of mind.” I am grateful for all the people and the places that help me enter that state of mind.

My (rad) research team at EuroSys 2024; from left to right, Ehsan, me, Robert, Ties. After I introduced them to Gustavo Alonso, he asked if I had a height requirement for my PhD students.

I have already written about the rejections in academia once (in a post titled Rejections). This is a never-ending story for most academics. Recently, two papers from my group got published and presented at conference workshops after 5 and 4 rejections from conferences, respectively. Hence, I decided to revisit the topic of rejections by focusing on these two papers.

Before I start with the individual papers, I would like to acknowledge a few things.

First, I have bias with respect to my own work. The main reason I work on what I work on is because I find the topic exciting and important. Otherwise, I wouldn’t be working on it.

Second, like everyone I know, I have a drive to share what I find exciting and important with others. That is why I welcome any opportunity to present our work and enjoy writing about it in the form of academic papers. Through this dissemination process, we share and build knowledge, get constructive criticism from our peers to improve the work, and start collaborations and new research directions. Not everyone shares the same level of excitement on the same research topics, though, and there is always room for improvement in a work. As a result, sometimes the feedback you receive sounds discouraging. This discouragement combined with the high dependency of one’s career on publications makes rejections difficult even though we all know that they are inevitable in our profession.

Third, I am aware that my health is the most important thing, and nothing I do or accomplish at work makes me as happy as the time I spend with the people that make me feel at home or at a movie theater or the beach. But I don’t want to diminish people’s career ambitions, especially in a world where women’s career ambitions are still under-supported, by over-emphasizing these cliché-but-true health and happiness statements.

An Analysis of Collocation on GPUs for Deep Learning Training

Ties Robroek, Ehsan Yousefzadeh-Asl-Miandoab, Pınar Tözün

EuroMLSys 2024 - https://dl.acm.org/doi/10.1145/3642970.3655827

This paper characterizes the performance of the different task collocation methods available on NVIDIA GPUs for deep learning training. The motivation came after realizing that not everyone that trains deep learning models is <insert your favorite big tech company here>. Thus, not every model training needs many GPUs or even the entire resources of a single GPU. This means that if we always train one model at a time on a GPU, that GPU is likely a wasted resource. Wasting hardware resources is a waste of money and energy inefficient. Studying how deep learning tasks can effectively share the resources of a GPU, therefore, made sense and was a relatively under-researched subject at the time.

We started back in September 2021 when my first PhD student[1] Ties Robroek joined my 1-person team. A couple of MSc students, Anders Friis Kaas and Stilyan Petrov Paleykov, were also interested in the topic for their MSc thesis project. The initial team was formed. 

We started with an investigation into the MIG (multi-instance GPU) technology, since it was the newest thing offered by NVIDIA GPUs at the time. MIG allows a GPU to be split into smaller units enabling task collocation with isolation guarantees.

1^st reject: The MSc students finished their thesis in June 2022. SoCC (ACM Symposium on Cloud Computing) paper submission deadline was around that time. We submitted the work from their thesis there. It got rejected with overall constructive and encouraging reviews. The main issue was people thought the paper didn’t have enough lessons-learned to warrant a SoCC publication. However, we got clear ideas for improving the paper for a possible resubmission. The key suggestion was to expand the study beyond MIG and add a comparison to other collocation methods on NVIDIA GPUs namely multi-streams and multi-process service (MPS).

2^nd reject: For this submission, we included EhsanYousefzadeh-Asl-Miandoab, my second PhD student, in the study. The entire paper was almost redone. We submitted the outcome to MLSys (Conference on Machine Learning and Systems) in fall 2022. It got rejected again. While the reviews were slightly less encouraging than SoCC, they were overall constructive. The reviewers didn’t find the results surprising enough, asked for deeper analysis on some experiments, suggested adding more diverse deep learning models to the study, and asked for scenarios that involve multiple GPUs.

3^rd reject: Following the MLSys reviews, for the next resubmission, we added more diverse models, dug deeper into certain results, changed the metrics we report to give a finer-grained picture for the GPU resource utilization, and wrote clearer guidelines for when it makes sense to use each collocation mechanism. The last point was to address the “no surprising result” comment. Since we cannot create surprising results out of nowhere, wrapping them up in clearer “take-away messages” made more sense. Finally, we deemed multi-GPU case out-of-scope for this study, since I strongly believe in the importance of optimizing things for the smaller scale as much as the big scale.

The resulting paper was submitted to ASPLOS (ACM International Conference on Architectural Support for Programming Languages and Operating Systems) in Spring of 2023 and was rejected once again. The workload diversity was praised by some reviewers, while some others asked for alternative workloads. However, the unsurprising results and the lack of deeper insights were once again the main issues.

4^th reject: I was overall optimistic after the first two rejects, because we had clear ideas for improving the paper for a resubmission. I think the paper indeed got substantially better as a result of those resubmissions. However, after the 3^rd reject, I didn’t know how to improve the paper anymore. We couldn’t mockup unstraightforward results. Further in-depth analysis wasn’t easy due to not every GPU hardware detail being openly shared by the vendor. Of course, one can always apply extra analysis through more profiling and add more workloads to the study if one has infinite time. However, I thought it would be better for the students to move onto the next stage/work in their PhD at this point. They also had the desire to move on. So, we decided to resubmit the paper to HPCA (IEEE International Symposium on High-Performance Computer Architecture) during the summer of 2023 without making extensive changes to it this time around. It got rejected with similar reviews to ASPLOS.

5^th reject: In one last attempt, we resubmitted the paper to SIGMOD (ACM International Conference on Management of Data) in fall of 2024 with extra results but not substantial changes. I wasn’t sure if SIGMOD was the right venue for this type of work, but as a SIGMOD reviewer I have seen papers on utilizing GPU resources better for deep learning being welcomed by some, if not most, of the program committee. I also thought that the insights we deliver on GPUs may be interesting to the data systems community. We got rejected again mainly due to straightforward lessons-learned and the topic being a borderline fit for SIGMOD.

Accept: Finally, I decided to stop trying to force this paper into a conference. Even though the amount of work we put into it was a lot, our findings were clearly not enough for a conference publication. In my team, we really like the EuroMLSys workshop (Workshop on Machine Learning and Systems) that is collocated with the EuroSys conference. Therefore, it was a natural choice for us, and the paper got accepted with a presentation slot at EuroMLSys 2024.

Reaching the Edge of the Edge: Image Analysis in Space

Robert Bayer, Julian Priest, Pınar Tözün

DEEM 2024 – https://dl.acm.org/doi/10.1145/3650203.3663330

This paper characterizes the performance of several resource-constrained hardware devices to determine their suitability for an image-filtering task on a small (hence, extra-constrained) satellite.

The roots of this paper also go back to 2021, though, the actual work on our end didn’t start till Spring 2022. In 2021, Julian Priest joined our lab. He is the main representative of the DISCO (Danish Student CubeSat Program) at our university. DISCO is an educational project that involves several Danish universities. It gives the students the opportunity to design and operate a small satellite. The target use case is Earth Observation; more specifically taking images of Earth from the satellite and analyzing them. The challenge with this use case is that the communication link between the Earth and the satellite isn’t your typical on-Earth internet connection; it is weak and temporary. Hence, sending all the images captured on the satellite is not an option. There is a need for image filtering on the satellite to send to Earth only the images that are of substantial interest. This need for filtering images leads to a follow-up challenge: the computation power that can be deployed on a small satellite is also small due to both space and power restrictions of the satellite. Hence, there was a need to identify the hardware device(s) to deploy on such a satellite that can satisfy the required size, power, and image filtering latency.

Since I joined ITU, I have also been interested in analyzing the performance of a variety of small hardware devices. In general, I always look for good excuses for benchmarking hardware. :) Hence, DISCO was a fantastic excuse. We also had the perfect student to lead the work, Robert Bayer, who was a student assistant with me then and is now one of my PhD students.

1^st reject: The hardware benchmarking for DISCO started in Spring 2022. I thought it could be interesting to write up about the results and submit something to CIDR (Conference on Innovative Data Systems Research) 2023. CIDR values papers on interesting and challenging data systems, and in my opinion the image processing pipeline of DISCO fits into this category. The reviewers, however, didn’t agree with me on the data systems connection, so the paper got rejected. Two out of three reviewers had a positive tone, otherwise.

2^nd reject: After CIDR rejection, I thought the SIGMOD 2024 Data-Intensive Applications track could be a fit for this topic. This was also suggested by one of the positive CIDR reviewers. We added one more hardware device to our study, re-measured power consumption on all devices with a more precise external device, included details on the satellite components, and submitted the paper. Around the paper submission time, April 2023, the first DISCO satellite, built based on the results presented in the submitted paper, was launched in space. I thought this submission was the best paper I had ever co-authored in my entire career (no offense to the co-authors of my other papers), but no one else agreed. The paper got rejected once again mainly due to being a misfit for SIGMOD’s data management focus.

3^rd reject: After two trials with data management venues, I thought it is better to target a systems venue as also suggested by some of the reviewers who rejected the paper. Thus, we made minor adjustments to the paper based on the feedback from previous reviews and submitted it to ASPLOS 2024’s summer 2023 round. We got more detailed feedback, since no one thought the paper was a misfit to ASPLOS. However, overall, the reviewers found the results not novel and surprising enough for ASPLOS and the focus on a single application too narrow, even though they all appreciated the motivation of the work. Hence, the paper was rejected once again.

10 days after receiving this rejection, Robert won the best Computer Science MSc thesis award in Denmark for the same work.

4^th reject: When we received the 3^rd reject, the submission deadlines for MLSys 2024 and EuroSys 2024 were already over. They would have been other relevant systems venues for this work. The other option, which was also recommended by one of the CIDR reviewers, was MobiSys, but this was a whole different world for me, and I wasn’t sure if I wanted to jump into a third community while already doing a bad job juggling data management and systems communities. Therefore, I recommended Robert to target VLDB’s (International Conference on Very Large Databases) Scalable Data Science track. Based on the call for papers, both Robert and I thought the paper’s topic fits there. We were wrong once again. The paper got rejected mainly due to being unfit for VLDB.

Accept: This paper was tied to a real-world application deployment; the first DISCO satellite. Hence, there wasn’t much room to improve the work to please conference reviewers. We could do more benchmarking, but the satellite was already in space based on our existing results. The paper as is had closure and real-world impact. Thus, to avoid delaying the publication further, I once again gave up on conferences and started to think about relevant workshops. Robert also needed to move on. I thought the DEEM (Data Management for End-to-End Machine Learning) workshop, which I like very much, collocated with SIGMOD conference would be a nice venue for this work. I emailed the workshop chairs to double-check the suitability of the topic for the workshop to avoid another “this is unfit” rejection. They kindly confirmed that the topic is in scope for them. So, we submitted the paper to DEEM 2024, and it got accepted.

I personally enjoy and value some conference workshops more than the main conference. Workshops gather the subset of people in a research community with similar research interests. They can be way more effective for exposing your work to the right audience than the conference itself. Similarly, the talks at a workshop in your research area are usually more relevant for you content-wise. So, I am happy that my students had a chance to present their hard work at these workshops that I very highly regard.

However, a workshop publication unfortunately doesn’t count as much as a conference publication on one’s CV when people evaluate you for academic positions or grant submissions. A couple of years ago, a postdoc candidate I wished to hire mentioned that I didn’t seem to have so many publications recently. This wasn’t the main reason he declined my offer in the end, but it was something he noted down, and I am sure others do the same. This is how our profession works.

It has been more than 6 years since I joined ITU and almost 3 years since I had my first PhD student. I still don’t have a conference paper with my own PhD students. If ITU had a more traditional tenure-track scheme, I wouldn’t have gotten the tenure. Earlier this year, I went down the rabbit hole trying to figure out what I was doing wrong and what I can do better in the future. The list was too long, but none of the answers were soothing. Deeper into the hole I questioned whether I was a shit advisor or a complete failure at my job. Tori Amos’ Crucify played over and over in my head, especially the lines “Nothing I do is good enough for you, so I crucify myself every day.” and “got enough guilt to start my own religion.”

Luckily, Crucify ends with “Never going back again to crucify myself every day.”

I know I made mistakes and misjudgments and will likely keep making them. I know the struggle is partly due to changing my research field and trying to build up my own research group from scratch without any starting funding. I know systems work takes time to get published; 2 years or more is the common case. I know the 3-year PhD duration in Denmark freaks me out as a result and makes me more impatient than I should be for publications. I know everyone’s papers get rejected; even the works of the people I admire. I know one of my favorite conferences, CIDR, was founded by people whose work was underappreciated and rejected by VLDB and SIGMOD. I know I still get invited for talks, and when I present my team’s work to others, I get positive feedback overall, unless people are lying to my face. I know many colleagues at ITU appreciate me. Most importantly, I know, at my job, regardless of the rejections, I learn a lot and get the most fulfillment from the work I do with my students. 

Note: Parts 0 and 1 of these posts are here and here, respectively. The only connection among the three is they are all inspired by the Danish language classes I took during my first three years in Denmark. This one was the hardest to write, just like the Danish homework that inspired it. I made the first attempt in October 2022, before writing Part 1. This was my third attempt.

We start a new book chapter titled “Dansk Frisind” in Danish class.

The literal translation of “frisind” would be “free mind”, a more nuanced translation could be “tolerance”. “Dansk Frisind” is usually translated as “Danish liberal way of thinking”. The first page of the chapter gives the following as supporting examples of Dansk Frisind: Danes don’t mind hanging out naked at the beach and allow teenagers to drink alcohol. Based on these facts, we are asked to discuss with our classmates how our home countries compare to Denmark in terms of frisind.

Giving over-generalized statements about my country, where almost 85 million people live, is something I dislike very much. When I moved abroad back in 2009, I didn’t know that I would have to be a representative for that many million people at various language classes and work events. To me, this is a burden that I didn’t sign up for. On the one hand, if I go ahead with stereotypes for the sake of keeping things uncomplicated, I risk erasing my own identity and the identities of many others that are dear to me. On the other hand, especially in a language class, I neither have the time nor the vocabulary to explain things in a non-black-and-white manner.

I know this challenge isn’t unique to me. Based on the stereotypes of our countries and the initial examples given for “Dansk Frisind” in the book, none of us in the classroom come from places that have “Dansk Frisind”.

To give credit to Danes, they have more meaningful (at least to me) examples to support their frisind. In Denmark, women earned the right to vote in 1915, abortion became legal in 1973, same-sex couples were given the right to be recognized as domestic partners in 1989 ... While not the first except for the last example, Danes were relatively ahead of many other countries in the world. The book mentions these facts later in the same chapter, but first we have to discuss the nude people and drunk teenagers.

During the discussions, to my non-surprise, a couple of my classmates express surprise after learning that I come from a family where the majority are Muslims.

Their reactions remind me of the questions I sometimes have to answer as the unintentional representative of a religion in the Western world.

“Do you (have to) cover your head when you go back home?”

No, unless I visit a mosque. When it comes to my hair, the only religion I follow is the Hair musical.

Among the women of my close family, my grandmother (mom’s mom) was the only one who wore a headscarf based on her religious beliefs and personal choice. In my extended family, there are others who wear headscarves as well.

“Are you allowed to drink in Turkey?”

Yes. The legal drinking age is 18, but my parents and bars allowed me to drink even younger. Till I was 16, I didn’t like the taste, though. In my early teens, I sincerely believed I wouldn’t drink as an adult. That didn’t turn out to be true.

“Are there any churches in Turkey?”

Yes, a quick internet search says over 300. Many Muslim-identified members of my family like to visit churches as much as mosques when they wish to pray inside a religious house. My mom and her sister are among them. They took me to both mosques and churches when I was a kid. As far as I know, the synagogues in Turkey aren’t open to public; you need a permission to enter for security reasons. They would have gone to the synagogues as well, otherwise.

“Are there any atheists in Turkey?”

Yes, I called one of them uncle. He died when I was 10. I have some of his belongings with me at home in Copenhagen today. They moved with me from place to place. One of those belongings is his copy of The Last Temptation of Christ. His library in general was the best source I had in Turkey when it came to learning about religion(s).

“Why don’t you fast?”

While I am legally registered as a Muslim in Turkey, I had a secular upbringing. I neither identify with nor practice any organized religion beyond some basic traditions. For example, I used to call my grandmother when the Ramadan holiday started to acknowledge the day. This is similar to wishing people “Merry Christmas” or “Happy Hanukkah”. On the other hand, I called my grandmother every Sunday regardless, when she was alive.

…

My experiences are my own. They are not representative of the entire Turkish population or Muslim families, but neither are the stereotypes of Turkish and Muslims families. Such stereotypes trigger questions and surprise reactions like the ones above, which in turn trigger me.

I know I am being unreasonable when I get frustrated by having to answer these questions or facing people’s surprise reactions. As Sister in Movies reminded me while I was venting one such frustration, it is better that people are curious and ask questions when they don’t know something rather than make assumptions. As someone who chose to become a professor, I want people to feel safe when they ask me questions. Also, I don’t know everything about every other country or culture either, so I would like to be able to ask questions to others without frustrating them.

On the other hand, these questions are like bullets from a loaded gun, even if they aren’t intended that way. Religion is one of the strongest and most violent dividers of people. My Muslim-majority family is the main reason why I am “the other” in Europe. My secular-majority family is the main reason why I am “the other” in Turkey or among other Muslims. I have listened and been in the middle of countless heated political debates on alcohol, fasting, women’s hair … both in the context of Europe’s attitude towards Islam and secular vs non-secular camps in Turkey, both with strangers and dear friends and family members. I am not blind to the, at times brutal, consequences of these conflicts throughout history.

The cliché Western narrative for my case would be “stuck between two cultures”. I am not stuck; this is my culture. I have been realizing that part of getting older is to learn how to hold contradictions together without breaking apart. I would like to own my culture and welcome questions on it even if the answers are not straightforward and cause unease in me.

Let’s go back to Danish class.

One of the homework assignments in the “Dansk Frisind” chapter is writing an essay on the alcohol consumption in our home country in comparison to Denmark. In addition, the essay must discuss high alcohol consumption among teens and whether harder alcohol rules can prevent it.

Among all the homework I have done for the Danish classes, this one gives me the hardest time.

I know, instead of torturing myself, I can write something like “Danes drink a lot.” and “In Turkey, we like drinking rakı with meze dishes.”  and be done. The point of these Danish essay assignments is to prepare you for the essay section of the Danish language exam, which you have to pass if you want Danish permanent residency or citizenship. People who grade these exams don’t care about your arguments in the essay, they only care about whether you can create coherent and grammatically correct Danish text in concise space (max 4-5 paragraphs).

The problem is I care. The topic of alcohol is a loaded and political one, not just in Turkey but also in Denmark.

Why does this chapter showcase alcohol habits in Denmark as an example of frisind and then ask us to discuss prevention methods for high alcohol consumption of teens? Why does alcohol consumption pop up often in Danish news and in political debates? Why is Denmark the country that created a film like Druk (which I must admit I don’t like as much as it is beloved in the popular culture)?

How do I represent the attitude of 85 million people, who are Muslim-majority and range from extreme secular to extreme conservative and all the shades in between, toward alcohol? Rakı and meze dishes don’t cut it, and instead of trying to find a way to represent the 85 million, my mind keeps going back to one person, my grandfather, my mom’s dad.

My grandfather was a Muslim. He did the prayer 5 times a day going to the mosque for it almost every day. He fasted. He read passages from the Quran every night before going to bed. He was a small business owner. Once he owned a restaurant, among its frequent customers were the workers of a nearby brothel. For a long time, his primary source of income was the liquor shop he owned in one of the most religiously conservative districts of Istanbul called Eyüp. He got shot in front of that shop for not supporting the district’s right-wing groups. He survived the shooting but turned the shop into a regular market with no alcohol sales after that. After my mom finished elementary school, he decided that she should stop her education, but backed down after my grandmother’s fierce resistance. He didn’t force his two daughters to cover their hair. He took them to the beach wearing a bikini. He let them drink alcohol together with him when they were teens. He stopped drinking alcohol after one of his grandchildren died at the age of three. He didn’t interfere with anyone consuming alcohol at his house even after that.

My grandfather is one person, and hence doesn’t represent the 85 million. But the complexities in his story is to me more representative of Turkey’s relation to alcohol and frisind than any overly-simplistic generalization I can come up with for the essay I am supposed to write.

In the end, I don’t include my grandfather in the essay. My Danish isn’t good enough to do justice to his story and I don’t have the space to elaborate. I will likely get a low grade if I try. I know this grade doesn’t matter, but I need constructive feedback to prepare for the big exam. So, I write something simpler about not generalizing either Turkey or Denmark. I end it with “For at opsummere er generalisering generelt forkert.”, which is me quoting Butler Lampson’s “Generalizations are generally wrong.” advice from “Hints for Computer System Design”, which I also quote several times to my students in the courses I teach.

After completing this essay, I promise myself to become emotionally detached from any Danish homework topic.

In Notes on a Foreign Country: An American Abroad in a Post-American World, Suzy Hansen mentions that James Baldwin expressed feeling free in Istanbul as a gay black man. The city offered him the refuge he needed in the 1960s.

Last year, a German man I interacted with told me that the older conservative Germans like to say, “We should be harsher like Denmark in our immigration laws.”

A place not really known for its frisind may still be capable of giving you the kind of acceptance you need, and it doesn’t mean that a place that is considered a stellar example of frisind will welcome you with open arms. 

Maybe frisind is more about being open to the contradictions of people and places rather than creating strictly-shaped boxes for them.