Azure Global Foundation Services (GFS)
Cloud-scale datacentres provide the core infrastructure and foundational technologies for its 200-plus online services, including Bing, MSN, Office 365, Xbox Live, Skype, OneDrive and Microsoft Azure. #Microsoft
To date Microsoft claim they have invested $15 billion in their datacenter infrastructure. Microsoft describe their datacenters as possessing the following 3 key attributes:
Microsoft Azure is currently available in 22 regions with plans for 8 more.
Is is important to note not all services are available in all datacenters and some special restrictions apply, for example:
Azure is now available in China through a unique partnership between Microsoft and 21Vianet.
The Australia regions are available to customers with a business presence in Australia or New Zealand.
- Enterprise Grade
Enterprise capabilities and features to address security, compliance, interoperability, reliability, and performance.
Services which span and integrate on premise environments with the public cloud. This is also a key differentiator between Microsoft and Amazon, the latter’s vision being far more fixed on public cloud alone.
Cloud-scale datacenters use commodity components to reduce cost. Accordingly one must plan for and expect failure. The focus shifts from Mean Time Between Failures (MTBF) to Mean Time To Recover (MTTR). This is achieved through the approach of ‘software defined everything’ and ‘automate everything’.
Services can be deployed into specific region(s) to address concerns around data sovereignty and latency.
Datacenter maintenance can be categorised as planned or unplanned. With planned maintenance Microsoft will aim to (but do not guarantee) they will:
- Provide at least 1 week advance notification
- Schedule the maintenance outside of regional business hour
It is important to be aware of this and consider using multi-instance VMs assigned to an Availability Set rather than single-instance VMs.
- Single instance VMs will result in any hosted applications or services experiencing an outage during planned or unplanned maintenance
- Single-instance VMs have no Availability SLA
- Mult-instance VMs assigned to an Availability Set will be split into separate Update and Fault Domains and are guaranteed a 99.95% Availability SLA