Netflix – Evolving Regional Evacuation(medium.com)
medium.com
Netflix – Evolving Regional Evacuation
https://medium.com/@NetflixTechBlog/evolving-regional-evacuation-69e6cc1d24c6
2 comments
Here's a glossary of terms that might help:
* Data Center: A physical location/building with racks of servers running.
* Availability Zone: A group of one or more data centers that are geographically close and connected. An availability zone belongs to a single region.
* Region: A broader geographic group of availability zones connected in a low-latency network. Data and servers within a region are usually isolated unless you explicitly set otherwise. As an example, us-east-1 in AWS is a region based out of North Virginia with multiple availability zones (groups of datacenters) in the area.
Most cloud failures or impairments tend to have their blast-radius limited to a single "availability zone" or a single "region". Traffic in one region for a service may have elevated error rates while all other ones are completely healthy. If you're architecting for high performance, you may want to run out of multiple regions and serve users that are geographically closest from each region. To also architect for high availability, you'd want to be able to have traffic shift away from regions with problems towards regions that are healthy. When shifting traffic from an unhealthy region towards a healthy one, you would of course need to scale up your server capacity in the region that will now be taking on extra load.
"Evacuation" here refers to that traffic shift. The article goes into optimization strategies for capacity scaling when dealing with these traffic shifts.
* Data Center: A physical location/building with racks of servers running.
* Availability Zone: A group of one or more data centers that are geographically close and connected. An availability zone belongs to a single region.
* Region: A broader geographic group of availability zones connected in a low-latency network. Data and servers within a region are usually isolated unless you explicitly set otherwise. As an example, us-east-1 in AWS is a region based out of North Virginia with multiple availability zones (groups of datacenters) in the area.
Most cloud failures or impairments tend to have their blast-radius limited to a single "availability zone" or a single "region". Traffic in one region for a service may have elevated error rates while all other ones are completely healthy. If you're architecting for high performance, you may want to run out of multiple regions and serve users that are geographically closest from each region. To also architect for high availability, you'd want to be able to have traffic shift away from regions with problems towards regions that are healthy. When shifting traffic from an unhealthy region towards a healthy one, you would of course need to scale up your server capacity in the region that will now be taking on extra load.
"Evacuation" here refers to that traffic shift. The article goes into optimization strategies for capacity scaling when dealing with these traffic shifts.
It means rerouting clients / webrequests to other aws regions / hosting locations when computing resources become unavailable or experience downtime in their region.
From the headline, I thought this was going to be about Netflix rolling out an "emergency broadcast system" equivalent based on the viewer's region/location. That would've been pretty neat (if a difficult technical feat)
I had originally thought it was going to be an article about emergency warnings broadcast on netflix.