Memo: Learn GCP Global infrastructure
Global infrastructure = the global presence of data centers, network and cloud resources.
Region
region — is independent geographic areas that are collection of zones.
- at GCP has 25 regions
- when want to launch a cloud resources (eg. VM), so it need to choose the region.
Zone
zone — is physical location that made up of 1 or more datacenter
Region & zone relation: 1 Region will container 3+ zones.
Datacenter in zone will be isolated from each other but they will be close enough to provide low-latency.
- Practice to run workload in at least 3 zones to ensure service remain available (High Availability)
- Once you choose Region, you will proceed to choose your Zone when launching cloud resource.
Edge Network
edge network — is a practice to have compute and resource as closest as possible to the end user in order to deliver the lowest latency and save bandwidth.
Point of Presence (PoP)
PoP — is a location between GCP region and end user. -> The location could be a 3rd party datacenter or collection of hardware
PoP 3 types:
- Edge PoP — is a location where user can quickly enter (ingress) to GCP network for accelerated access to cloud resource.
- CDN PoP — is a location where serve (egress) cache resource (website, file or assets) for load very fast to end user.
- Cloud Media Edge — is a location specialized for delivery of media (Video content)
Resource Scoping
One zone is a deployment area for Gcloud resources with in a region. So only one zone should be considered a single failure in region. To deploy redundant resource is multiple zones (multi-zones) that help fault tolerance and high availability.
Product / service can scoped based on:
- Zonal resource — resource resides in a single zone in single region.
- Regional resource — resource resides in multiple zone in single region.
- Multiple regional resource — resource resides across multiple specific zone.
- Global service — resource resides globally and regions and zone.
- Internal service — resource used by many other service, but you don’t interact with these services directly.
Cloud Interconnect
cloud interconnect — is Gcloud service to provide direct physical connection between on-premise network and GCloud network. Its allow you to transfer large amounts of data between network which more cost-effective than purchasing bandwidth over public internet.
2 types of Cloud interconnect:
- Dedicated — a physical direct connection on-premise network and Gcloud network through a co-location facility -> can transfer 10–200 Gbps
- Partner — a physical direct connection on-premise network and Gcloud network through a trusted 3rd party -> can transfer 50 Mbps to 10 Gbps
- Co-location — aka Carrier-hotel is a datacenter are available for rental for retail customer.
Latency
Latency — is time delay between two physical systems.
Lag — is time noticeable delay between the action of input and the reaction of the server send back to the client.
- Inter-regional latency — latency between region
- Inter-zonal latency — latency betwwen zone in signle region
*Latency between region is alway more that zone
