text stringlengths 0 2.67k |
|---|
• Autumn Nash recommends using blog posts as a way for companies to share their learnings and experiences with others. |
• UniSuper's account on GCP was deleted, causing all data to be lost |
• The deletion occurred due to a default setting on an internal tool used by Google Cloud |
• The tool had a one-year "deletion period" that wasn't noticed or properly configured |
• UniSuper was using VMware on top of GCP and trying to migrate away from VMware |
• Google Cloud released a post-mortem report detailing the incident and its causes |
• The report concludes that this was an isolated incident and not a systemic issue in Google Cloud |
• Deprecation of Google services and impact on customers |
• Importance of processes and observability in infrastructure management |
• Hidden time bombs in infrastructure, such as expired certificates or logs filling up |
• Impact of proprietary software on companies like VMware and Broadcom |
• Future implications for cloud-agnostic applications and lift-and-shift strategies |
• Abstraction vs. control in cloud computing, with examples from Render |
• VMware's evolution and the trade-offs between abstraction and control |
• Importance of knowledge and access in managing complex systems |
• Open-source alternatives like Kubernetes as a way to avoid vendor lock-in |
• Cloud providers becoming the new "VMware" with similar challenges |
• Balancing cost, expertise, and control in scaling IT infrastructure |
**Justin Garrison:** Hello, and welcome to Ship It, the podcast all about what happens after you git push. I'm your host, Justin Garrison, and with me as always is Autumn Nash. How's it going, Autumn? |
**Autumn Nash:** I'm tired. How about you? |
**Justin Garrison:** Tired? Man, this is a short week, too. |
**Anurag Goel:** Don't you feel like that almost makes it worse though? ...because we all know that we're going to do the same amount of work, but within less time to do it. |
**Justin Garrison:** It's a compression algorithm. |
**Autumn Nash:** Yeah. I love that use of the compression algorithm... |
**Justin Garrison:** This is deflate all over again, so... |
**Autumn Nash:** Deflate, but like workweek deflate. |
**Justin Garrison:** Yeah... I feel a bit deflated. |
**Autumn Nash:** We need more Dr. Pepper and coffee to get through it. |
**Justin Garrison:** Yeah, we'll make it there. So for this episode today we are talking to Anurag the CEO and founder of Render. Render is a platform for running applications; they have a bunch of other stuff that they do as well. So if you never check them out today, it's sort of like a platform as a service... |
**Autumn Nash:** Infrastructure as a service, sort of... |
**Justin Garrison:** Yeah, I would call it more like a modern Heroku sort of thing, where it's like "Oh, what if we took the ideas from Heroku and applied it to today's sort of infrastructure abilities?" And so it was really cool talking to him about what it took to start it, how they were scaling it, and kind of what ... |
**Autumn Nash:** I love that, for one, he was an engineer before he was CEO. I love that. And that he went out to solve a problem, and that they are smart enough to pivot and still be customer-focused... Because I feel like a lot of times right now people are starting startups -- a solution that solves an imaginary pro... |
**Justin Garrison:** Right. I mean, I guess that's the goal for a lot of cases, where it's like the ratio of application engineer to infrastructure engineer is a little more offset towards the application side, which tends to be the thing that creates companies value, or makes them money at least, whether it's valuable... |
**Autumn Nash:** It's what's available to you. A lot of times you may have a database that might be a better solution, or you may have an infrastructure that's a better solution, but where do you find the people that know how to use that, and know how to use it well enough at scale? So I think that a lot of it too is t... |
**Justin Garrison:** So let's go ahead and jump into the interview, and then we will be back afterwards with an outro. |
**Break**: \[07:39\] |
**Justin Garrison:** Thank you so much, Anurag Goel, for joining us on the show today. You are the CEO and founder of render.com. Can you explain to us a little bit how and why you started Render? |
**Anurag Goel:** Absolutely. Thank you for having me. It's very nice to meet you both, Justin and Autumn. I was previously the fifth engineer at Stripe back in the day, 2011. I join pre-launch, and then helped the company grow to hundreds of people over the next nearly five years... And I saw, just as the company was g... |
And so as I was doing that over the two years after I left Stripe - so I left in early 2016... Towards the end of 2017 I had built this other product, I was into deep learning, and I was taking these online courses... And one of the things I ran into was just setting up a Jupyter Notebook powered by a GPU on AWS. The c... |
But then I realized, as I was doing all this, that I was building the same kind of production infrastructure over and over again. And most companies do it sort of once, and it happens in parallel across multiple companies. I had been doing it serially, because I was spinning up multiple projects that were completely in... |
**Autumn Nash:** Can we talk more about what you did at Stripe? Were you like a platform engineer, so you were helping people use Kubernetes? Or was it just like the struggle that you learned from using it because you were building on it? |
**Anurag Goel:** Stripe was not using Kubernetes when I -- I mean, Kubernetes came onto the scene much later. I joined Stripe in 2011. We were doing all kinds of things. And when the company is 10 people, as an engineer you end up doing a little bit of everything. And then eventually as Stripe grew, I ended up building... |
\[12:03\] So I wasn't directly working on our AWS infrastructure, but I was still pretty close to the engineers who were... And I kind of saw all the problems, the late-night pages... As Stripe scaled in terms of volume, the number of payments we were processing across the board just continued to increase, and that led... |
**Autumn Nash:** The reason I asked is because I think that as developers we're constantly inundated with different products, and they all claim to do the same things. But when you can hear the problems that someone's trying to solve and how they got there, you can better understand the product and like the value prop,... |
**Anurag Goel:** This is something that I've been thinking of for like six years, six plus years now... |
**Autumn Nash:** This is your time to shine. |
**Anurag Goel:** So the first piece of that is you have to make it really easy for people to get started. Anytime you have to log into AWS to do anything, an angel dies somewhere. |
**Autumn Nash:** I almost spit out my coffee. |
**Anurag Goel:** I mean, I have felt that pain many times myself. And AWS'es whole spiel is "Look, we're going to give you these Legos, and we're really good at building hardware, and data centers, and keeping machines up and running at all times, and then giving you these storage, networking and compute primitives tha... |
Whenever you make something easy to do, people do more of it. And so that's the starting point. But the way we differentiate beyond that is even more important... Because when you think about the history, the evolution of platforms, Heroku was great when it started, 2005... 2007 was sort of when it publicly launched, a... |
But then, after its acquisition by Salesforce, development of the platform and adding functionality to the platform to allow applications and companies to continue to grow on Heroku - that stopped, and kind of slowed down to a crawl. And ultimately, Heroku got to a point where they just weren't giving people what they ... |
**Justin Garrison:** Is that a feature of -- because you could call it Heroku, but every other platform basically has that same problem, where if it's easy to get started, it's going to be hard to scale... Because you're gonna have all the defaults. You're tightly coupling something upfront to say "Hey, we're gonna mak... |
\[16:22\] I was at AWS when we shipped App Runner, which was literally like connect a GitHub repo and it just spins up a container and builds it for you and deploys it, and you have a web endpoint. It was like "Oh, cool." That was the thing that, like everyone said, "Oh, I want that thing." It's like "Well, do they wan... |
**Anurag Goel:** Yeah, absolutely. |
**Autumn Nash:** See, I think a lot of the problems with anything that's managed or anything that is easy to start with - just to piggyback on what Justin said - it's hard to debug after a while, after you scale, if you didn't go in there and pick those things, and you're just getting whatever is the default. It begins... |
**Anurag Goel:** Yeah, so this was the second differentiation point, which is - look, Heroku kind of stopped there. However, it is absolutely possible to build a platform that exposes the right kind of primitives, but it doesn't have to expose all of them at the same time, and all of them in the same way. So as an exam... |
So again, to give you an example, we have customers who started on Render with just the three founders paying us like 30 bucks a month. And they're now hundreds and hundreds of people. They didn't have any infra engineers, and they're now paying us well over a million dollars a year. And they have maybe one or two full... |
Another example of something like that is -- well, like you said, you have two of these services, you need to connect them to each other, and there's a VPC, and then you have to set up security groups, \[unintelligible 00:19:11.24\] all of that... Render has private networking just built in. Render has service discover... |
And we have network isolation as a feature as well, and something we're working on now is if you have multiple environments, like dev/staging/production, you will soon be able to isolate those environments from each other from a network standpoint with a single click. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.