Hey speaking of which, I'm curious and would love to hear... what is your favorite *useful* definition of "platform"? (must be falsifiable, not just a marketing or fundraising term du jour, or snarky, etc)
I'll tell you mine if you tell me yours. 🤔
I'll tell you mine if you tell me yours. 🤔
Thanks for your interesting answers! As an infrastructure engineer, my working definition of a platform goes something like this (i'm now seeing how infra-specific it is):
"You have a platform when you invite your users' chaos to sit on your systems, not theirs."
"You have a platform when you invite your users' chaos to sit on your systems, not theirs."
by "users' chaos" i mean not the chaos your users render unto your business, but unto the users' own businesses on your servers. the chaos dealt to your users by users-of-users.
to me, platforms are characterized by:
1) you didn't write the code paths or make the content that your users' users are pounding away on. you can't track down the developer and shake their lapels or reverse their commits; these things were created far from your watchful eye.
1) you didn't write the code paths or make the content that your users' users are pounding away on. you can't track down the developer and shake their lapels or reverse their commits; these things were created far from your watchful eye.
2) every one of the users of your platform has a unique performance footprint or quality of service. in addition to caring about your own site's perf and quality, you have to care about *every single one of your users'* perf and quality as experienced by *their* users.
So you simultaneously have much less control over/visibility into what is created (or uploaded or whatever), and much, *much* higher demands in terms of quality and observability.
You also need to distinguish between self-inflicted problems vs edge cases that are your fault.
You also need to distinguish between self-inflicted problems vs edge cases that are your fault.
For example:
You have a customer who's doing 4 req/sec (your system does "millions"). All of their requests are timing out at 60 sec -- the app is down. But their errors or latency will never, ever show up in your dashboards or KPIs or alerts. And you have millions of users.
You have a customer who's doing 4 req/sec (your system does "millions"). All of their requests are timing out at 60 sec -- the app is down. But their errors or latency will never, ever show up in your dashboards or KPIs or alerts. And you have millions of users.
Unfortunately, this customer is a big deal. Maybe it's the NFL testing their app before some sportsball event. Why is it timing out? Is it your fault, or theirs? Can you explore perf/quality questions for any given app, the way you (hopefully) can for the site as a whole?
This is why platforms are so much fun. :D once you start building platforms and solving category problems, you're pretty much spoiled for going back to build something for just one solitary customer, ever again.
so anyway this is what my talk is about tomorrow at #DOES18 🌈📈
so anyway this is what my talk is about tomorrow at #DOES18 🌈📈
