Azure Compute¶

PaaS¶

App Service Plan is equivalent to a server with software stack and provisioned resources
whereas App Service is an application that uses resources

VM can scale via Availability Set, VM Scale Sets, Azure Batch
VM Scale Set allows a group of identical, load-balanced VMs to be managed, configured as a unit
- Number of VMs in a VM Scale Set can change dynamically depending on the load
Azure Batch enables scheduling a batch job with a pool of VMs
- Azure installs application and stage data on VMs
- runs identified tasks, re-queues work on failure
- scales down the pool as work completes
Desired State Configuration DSC prevents drifts from the des*ired configuration
- maintains server, roles, registry, files/directories, packages, env var
- it's a powershell file attached to VM as a zip file
Accelerated Networking allows VMs to be on a faster, dedicated network. Not supported for all instance types
Dedicated Hosts and Dedicated Host Group allow dedicated physical servers.
- It is possible to have both Availability Zone and Fault Domain
- Can create multiple VMs, but all have to be same series as the physical server

SLA for VMs using premium disks: 99.9% single instance, 99.95% Availability Set, 99.99% for Availability Zone
Availability Set
fault-domains share power and cabling, essentially a rack, provide protection from unplanned outages -- usually 2 or 3 per AZ
- different fault-domains don't share hardware such as network, power, cooling etc
update-domains provide protection when performing planned maintenance -- can be up to 20 per AS
VHD is not TODO
Resiliency refers to ability of system to recover from failures
Availability refers to the time that system is functional and working

A Json file, either modify an existing one or save VM creation as template by selecting "Download Template for Automation" in final ste;
Template has: template.json, parameters.json and deploy*.*sh* files to deploy it
template.json
- $schema : must not be changed
- contentVersion: User modifiable for versioning
- parameters: a dict of name and another dict for parameter properties (such as type)
- variables:
- resources: Use parameter values
- outputs:
Needs a parameters.json file, which contains value for the parameters defined in template.json
instantiation required template.json + parameters.json

run-time debug logging: add calls to trace writer in code
App Insight delivers a lot more telemetry data
- kudu allows looking at raw performance data

is a messaging service
can emulate Kafka protocol
has a namespace that is similar to Kafka brokers, except it is a single endpoint
supports very large amounts of data
capacity measured by Throughput Units ~ 1MB/Sec and 1000 events/Sec ingress, egress double of ingress