hckrnws
> “Wait, wait, wait, but why are we doing it this way?” Barroso said on the podcast. “And it just turns out that the people who had been living in that area hadn't really thought about questioning that. And sometimes it's something that was based on a good reason three years ago, and that reason had a sell-by date, and it's time to do something else.”
This is a very humane way of thinking about the mindset of people who implemented solutions that don't quite work for your needs.
It's the other side of Chesterton's Fence: When you do understand why the fence was put there, you can remove it.
Too often people turn Chesterton's Fence into Frost's Fence: "Good fences make good neighbors" and refuse to think beyond that.
Oh, that's a shame. Luiz was one of the nicest people I met at Google and his tech contributions were excellent. He also promoted "roofshots over moonshots", which was a philosophy badly needed.
Because I was curious and others may be too, written by the man himself: https://rework.withgoogle.com/blog/the-roofshot-manifesto/
https://web.archive.org/web/https://rework.withgoogle.com/bl...
Seems the source is overloaded or down.
Thanks for sharing this article.
It feels like a hopeless endeavour to learn technical skills in later stages of life, reading the article helped me believe in maintaining my focus.
He gave "self-driving cars" as an example of "disruptive, 10X leaps in technology". I am not so sure if that what we need is even more cars on the roads.
Luiz called large meetings "all-heads" instead of "all-hands" meetings. "Why? Because we're not sailors!"
I never got to interact with him directly but he was a leader worth following. This is a sad day.
Luiz was one of the smartest, kindest people you could have ever met. He was the kind of person you'd have been proud to call your friend.
If you are ever looking for someone to aspire to as a software engineer, you should aspire to be like Luiz.
> If you are ever looking for someone to aspire to as a software engineer, you should aspire to be like Luiz.
This a million times.
I only interacted with Luiz a few times. They were eye-opening. He was so bright and, at the same time, so kind and supportive!
My team has improved some algorithms he and Jeff Dean popularized. When we presented to him, he asked us excellent questions. But, even though he was far up in our management chain, we never felt attacked nor judged, rather supported and empowered. I told my friend that it felt almost like talking to my grandfather when I was a young boy.
I also really appreciate some of his documents. They dealt with fairly general/philosophical topics (on the nature of software). They were so succinct and yet ... so deep and far-reaching! They left a long-lasting effect on me.
I admired Luiz's ability to connect with others. He showed me how not all truly smart people need to make a display of it.
I'm truly sad he died. And just after Bram...
"The Datacenter as a Computer: Designing Warehouse-Scale Machines" (3rd Edition) (2018) https://www.google.com/search?q=The+Datacenter+as+a+Computer... https://books.google.com/books?id=b951DwAAQBAJ
Also the co-author of The Tail at Scale, one of the more interesting papers on web-scale architecture that I've read: https://research.google/pubs/pub40801/
Comment was deleted :(
As a Brazilian, I'm really sad to hear this news, but I'm glad Luiz's work is getting the respect and recognition it deserves. Thanks Luiz.
"By customizing nearly every inch of Google’s data centers and the hardware within them, including power supplies and cooling kits"
This is the second time i hear about a company building its own power supply, the first being apple for the apple 1.
As i know "nothing" about electricicty, i wonder : isn't power supply a solved issue at this point ? What could have they done differently that justify having its own production line for a power supply ?
In the early days of Google, that would have been getting rid of the expensive, building- or cluster-wide UPS and adding batteries to the power supplies. Later, they switched to a rack-level UPS.
There isn't a lot to the Apple power-supply story. An Apple employee designed a switching power supply at a time when the industry was moving from linear to switching power supplies. Ken Shirriff writes more: https://www.righto.com/2012/02/apple-didnt-revolutionize-pow...
Yes, low-voltage power supplies are commodities. But they're also complex and critical to a product's operation (look up "glitching attacks" to see how squirrely power can influence behavior). If running hardware reliably at scale is part of your company's core competency, then your core competency will likely also include understanding the power being supplied to that hardware.
For datacenters, there are many design decisions for the power supplies and what voltages to use. The obvious solution is to use commodity PC hardware, so you have 120 volts AC to the computer's ATX power supply, which generates 12 volts DC, which is dropped to e.g. 1.5 volts on the motherboard to power the CPU. But this isn't the most efficient approach. You're going to get more efficiency with higher AC voltage (e.g. 208 volts). And you're going to get more efficiency with higher DC voltage (e.g. 48 volts). And also ATX power supplies aren't optimized for datacenters, since they have a lot of historical baggage and voltages that may not be needed (like standby). So there is a lot that can be done to squeeze more efficiency out of the power supplies.
Here's a Google presentation about changing to a 48V architecture: https://www.opencompute.org/files/External-2018-OCP-Summit-G...
Google has also used new circuitry for the DC-DC conversion. Here's a paper on their switched tank converter to drop 48 volts to 12 volts: https://storage.googleapis.com/pub-tools-public-publication-...
Have it hyperoptimised for your precise operating conditions?
Reliable, cheap, and available in high number power supplies are surprisingly hard to come by.
Delta makes tons of money on satisfying the "reliable, and available" points, but not the "cheap" one.
from https://www.barroso.org/ "February, 2023: I have released the album Before Bossa, alongside two fantastic jazz musicians, Zeca Assumpção and Sergio Reze. It features a selection of Brazilian and American Jazz tunes that inspired the creators of the Bossa Nova musical style."
https://www.youtube.com/watch?v=_f0BYADR7B0&list=OLAK5uy_mVO...
Luiz Barroso - Before Bossa (feat. Sergio Reze & Zeca Assumpção)
I shared an office with Luiz for 2 years - with Urs Hölze and Bart Sano - I joked with them that I got a PhD every 12 months hanging out with them. Luiz was an amazing engineer - and a great teacher, role model, philosopher and amazing friend. Rest well my friend - you made an indelible mark on me.
Sad to see another great guy go away too early :(
@dang, why is there no black bar at the top of HN?
Crafted by Rajat
Source Code