" If Google can get away without using it then why do you need to have it?" I th...

beagle3 · on July 27, 2017

Do you have the capital that Google has?

Do you have the expertise that Google has?

Do you have the scale that Google has?

Do you have the problems that Google have?

I had an employee once who, when tasked with designing an API for use by a partner company, copied MS conventions a-la "CreateWindow", when I asked about adding more functionality he said "I'll just do CreateWindowEx" and "IWebBrowser2". His claim was "If MS is doing that and MS is successful, this must be a good way to do it".

After I pointed out that, despite several attempts, there is no MS-Compatible API (this was in 2001, I think reactos and wine had both been started, there had been other attempts at WinAPI emulation, but none was worth anything), and that we had 2-people working on it, not 400, he agreed that perhaps the Unix API is easier to maintain, document and reimplement -- as is evident by many (re)implementations available.

For Microsoft, a labour-intensive-to-maintain-and-labour-intensive-to-document API is a benefit, because they can afford it and they are already holding the dominant market position (making it hard to be compatible with them). For a lean team, maintaing this kind of API is a penalty.

Similarly in this case, what's good for Google and what's good for you are not necessarily equivalent.

Google solves speed, redundancy and reliability by making and managing copies. ZFS makes one single machine the most reliable file system available today. But Google cares not about any single machine.

Do you have 100 copies of every important datum? If you don't, then you shouldn't copy Google.

supremesaboteur · on July 27, 2017

I assume the thinking goes like this :

ZFS was designed for enterprises. Enterprise software has requirements of high stability, scale etc. Google is probably on the extreme end of these problems. So if Google considered and rejected ZFS it probably means it was not very good.

What probably really happened was that Google was already too invested in Google File System and therefore did not give other file systems a fair chance

mjevans · on July 27, 2017

Google's big enough that it, most likely, has something that the public ceph project is a pale imitation of.

Actually, they're big enough that something between what ceph can do and what backblaze does is probably their /archive/ backend.

They probably have all of the fast stuff on 'disposable' temporary copies in RAM or SSDs.

petters · on July 27, 2017

GFS does not really solve the same problem as ZFS, I think.