Splitting Light: Season 2 - Episode 37


Splitting light

Season 2 Episode 37

Hardware made redundant

If you are no longer interested in the newsletter, please unsubscribe

Around May 2019

In May 2019, the hardware lab’s C3 was in production. It was the third generation compute hardware. The last one I had worked on before switching to storage. There was an issue with the compute node. How it had been fitted with RAM and storage created a voltage drop in certain conditions and that would shut the node down. A capacitor had to be soldered by the hardware team on-site to fix the issue. But how did that issue come in the first place?

The nodes had been procured with a configuration that had not been tested in the lab. Even though since the C2 in 2015 we, the lab, had requested that every deployed configuration be tested in the lab before procurement and deployment. This step, for some reason, had been skipped this time.

Around May 2019, I had initiated the discussion to start the process to create custom storage hardware. Théo (a) and Greg (b) were defining the requirements. It was now appropriate since we had gained a lot of experience and could focus on what would really make a difference. What would bring leverage.

It was a surprise when one day a message was posted on slack. It shocked everyone. For me, it shocked me to the core. It nearly broke me. It was suddenly announced that Scalway would no longer be doing custom hardware. The hardware team and the lab was declared redundant. We would immediately stop the R&D and no longer have custom hardware.

I correlated the shutdown of the lab to the capacitor issue. But who was at fault? Who had created the bug? Was it the hardware team that had not installed that capacitor? Or the person who made the procurement order on components that had not been tested?

Since we had done custom hardware, the instruction had always been given to test before procurement. The lab had to approve the configuration. Yet for some reason, that step had been skipped.

After confusion came anger, I could not find any logical reason why such a decision had been taken. But then, none of us were in the meeting where that decision was taken. Making the hardware team redundant would create a very long lasting impact. We, in the storage team, would feel it deeply. Some things were only possible at the hardware level.

When Scaleway would start having to do hardware again they would have to start from scratch again. That time would come, I was 100% certain of that. Almost every single hyperscaler does custom hardware. The vendor relations would have to be rebuilt. The experience re-acquired. The processes streamlined again. Everything would have to be redone. It was such a waste. Scaleway did start doing hardware 3 years later and expanded later on.

For me, personally the impact was very hard. All that metaphorical blood and sweat. All those years spent, all that effort invested, it was for nothing. There was no legacy. I thought the lab was this single differentiator. The one thing that made us different from other small cloud providers. Our pride. Just like Free, the telecom operator, builds its own hardware, we did it ourselves and it brought us business value. Or so I thought. I would later understand that it did have a legacy, just not the one I thought.

After the hardware team was suddenly disbanded, they were told to find another product to work on. After a few weeks they found a product to work on, they built the IoT product line. Gregoire continued to work on heavy R&D within Scaleway and built many other products.

For me, the anger eventually led to sadness. I can now see in me the legacy of the hardware team as well as the legacy of this distressing event. From that event, I learned to always prove the value of what I built with business numbers and focused a lot more on potential issues before they happened. The legacy of my work, I find it everywhere with me. The knowledge, the work process, the resourcefulness, the resilience… This is the legacy. The legacy was not bundles of metal, silicon and plastic. The legacy is what was etched in my mental circuits.

In parallel, as I was signing my mortgage. At work we had to assess the tech debt.

  1. https://x.com/Scaleway_fr/status/530421327686299648
  2. https://x.com/Scaleway/status/707617919434412034
  3. https://lafibre.info/opcore/dc5/msg613411/#msg613411
  4. https://lafibre.info/opcore/dc5/msg613678/#msg613678
  5. https://x.com/Scaleway_fr/status/974559724938170368
  6. Photo by me

(a) Théotime Rivière, Storage Product Manager then, now Founder of Freedom From Scratch

(b) Gregoire De Turckheim, Manager of the hardware lab in season 1 & 2, now doing R&D at Scaleway

If you have missed it, you can read the previous episode here

To pair with :

  • Hard to say goodbye - Washed out, Lone remix
  • Bug Jack Barron by Norman Spinrad

Vincent Auclair

Connect with me on your favorite network!

Payson, Payson, AZ 85541
Unsubscribe · Preferences

Symbol Sled

Business, tech, and life by a nerd. New every Tuesday: Splitting Light: The Prism of Growth and Discovery.

Read more from Symbol Sled

Splitting light Season 3 Episode 13 Bills reconciliation If you are no longer interested in the newsletter, please unsubscribe January to March 2019 When you have a complex system there are always going to be discrepancies between them. In our case the discrepancies were in the billing systems. We had five different data sources. None agreed on the numbers. Let me introduce our five systems. Observability of the disk usage as well as the incoming and outgoing bandwidth. The client...

Splitting light Season 3 Episode 12 Warsaw If you are no longer interested in the newsletter, please unsubscribe End of January 2020 to beginning of February 2020. In January 2020, we had a new challenge. A new opponent to ring in and control. Let me introduce him. It was the Warsaw region! We had to deploy object storage in Warsaw. Back in September 2019 we had selected the hardware setup and had done a brief discovery of it. Then the whole region’s hardware had been shipped by truck. Once...

Splitting light Season 3 Episode 11 OpenIO reaching out If you are no longer interested in the newsletter, please unsubscribe End of 2019 While we were dealing with the 500k object problem in object storage, Jean-Francois Smigielski (a), OpenIO’s CTO reached out to me on their community Slack. They were looking to hire more engineers and several people that they trusted had told them I was what they needed. Most likely I had also done a good impression when we had worked together to bootstrap...