3 ms·
I can speak to the server provisioning complaint in the Quora thread since I briefly worked on the PaaS team in their CorpEng group. Google's server and infras
by nunez 10y ago
I can speak to the server provisioning complaint in the Quora thread since I briefly worked on the PaaS team in their CorpEng group.
Google's server and infrastructure provisioning is actually not that bad. It's MILES better what I've seen at any other large company I've worked for (and I've worked for a few). Requesting new servers and instances is automated (plenty of places require you putting in a ticket), virtual instances are provisioned almost immediately and are available in minutes (much faster if your workload is on internal GCE) and many of the changes to those instances are nearly instant if they're reasonable. You could even modify login and ownership permissions with this damn tool. It was an incredible piece of work.
Obtaining hardware was a different story. Requesting it was automated just like it was for virtual compute, but getting servers racked-and-stacked took longer because the tickets generated by the request tool needed to be routed to a hardware operations team in our datacenters. (By the way, they have a REALLY strong hardware ops team. These guys and gals, but mostly guys, care about every damn resistor in their machines, especially the ones that were built by them for Borg compute.) This took a while sometimes because, as you would expect, Google had two different request systems and seeing which updates were made where wasn't always easy. As for the servers themselves, let's put it this way: their SMALLEST server was bigger than the largest servers provisioned by MOST large companies. Obtaining a beast of a server (32-cores, 256GB RAM, a few TB of disk space) was pretty easy despite being frowned upon ("can that run on Borg? No? Okay!"). They have tried adding costs to stuff a bunch of times to make teams feel bad about abusing compute (SRE/SWE-minutes/hours are a real thing, and some tools even tell you how many were spent on your request), but it didn't matter too much in the end since Google's internal IT budget was always pretty lofty and it mostly affected who got to take their entire team on an all-expenses-paid vacation at the end of the year.
The teams were siloed to high heaven, so that did slow things down (my job there: "Is it the operating system? No? Pass it onto networking/HWops/VM team!"), but those teams absolutely knew their shit.
Everything else in that post was spot-on, though, and pretty accurately described why I no longer work there. But, without question, Google is one of the most employee-friendly companies in the world. If you want a high-paying tech job in an urban city, stability and all of the fucking perks you can imagine (you can fill your entire closet on Google clothing and gear, though you'll need to buy pants and underwear; the SRE backpack they gave me is one of the best I've ever used), you can't beat working at Google.
However, if you're the slightest bit entrepreneurial, want to make as much money as a dev or whatever as possible or want a fast-tracked career of some sort, Google is, at best, a good pit stop between your current job and your next impressive gig. (Their non-compete is ridiculously limiting; you essentially can't work on any side projects while working for them that you intend on selling.)
- eecsninja 10y agoMostly correct. One nit: Mountain View is hardly an "urban city" and SF is a 90 minute commute away.