Re:How we scaled Google Meet during Covid-19
by jeffbee on Thursday August 06, 2020 @05:46PM
(#24075822)
It's clear that the Google where you worked existed in a parallel universe from mine. On my home planet, google netcap required a forecast per-product, per region pair, years in advance, and the product group got charged back for the capacity requested, not for the usage, so nobody was requesting extra.
[ Parent
| on HN ]
Re:How we scaled Google Meet during Covid-19
by londons_explore on Friday August 07, 2020 @02:56AM
(#24079104)
Hah - except there were overestimates at every level, like where teams were confident of there 50% growth per quarter for the next 2 years, and that never materialised. Know the right people, and that unused capacity can be reallocated to you in a few hours.
Capacity planning done like that is mostly a waste of time...
Re:How we scaled Google Meet during Covid-19
by justicezyx on Friday August 07, 2020 @04:31PM
(#24086135)
Yes, planning is always there.
I was saying that this sudden increase in the traffic should not cause major issues. Most likely several SREs have emergency meetings, and get quotas across places. Then it's done.
If you read my post above, you'll notice that I did not mention SREs. I was saying app developers do not care planning.
Of course, some app developers thought they were doing planning, as in some activity to estimate demands, etc.
That, in Google infrastructure engineering, is not considered capacity planning. The typical capacity planning is done by SRE with more procedures that certainly border the most curious developers.