Resilient Architecture Concepts
Summary
PDF p.160Cloud services can be designed to be resilient to failures at various levels, such as components, servers, networks, and datacenters. Cloud Service Providers (CSPs) use virtualization and redundancy to ensure high availability and data replication, meeting the availability criteria set out in their SLAs.
In plain words
Supplementary — not from your PDFCloud providers build resilience with redundancy and replication. High availability storage promises something like 99.99% uptime. The world is split into regions, each with independent availability zones (separate power, cooling and network). Data can be replicated locally (within one datacenter), across zones (zone-redundant), or to a distant region (geo-redundant). Hot storage is fast and costly; cold storage is slow and cheap.
Detailed explanation
PDF p.160-
High Availability (HA)
- Definition: Storage provisioned with a guarantee of 99.99% uptime or better.
- Redundancy: Multiple disk controllers and storage devices are available to a pool of storage resources.
- Replication: Data may be replicated between pools or groups, supported by separate hardware resources.
-
Replication
- Purpose: Allows businesses to copy data to where it can be utilized most effectively.
- Central Storage: The cloud can serve as a central storage area, making data available to all business units.
- Requirements: Low latency network connections, security, and data integrity.
- Storage Tiers: CSPs offer various data storage performance tiers, such as hot storage (quick retrieval, higher cost) and cold storage (slower retrieval, lower cost).
- Application Needs: Different applications have diverse replication requirements, such as low-latency, synchronous replication for databases.
-
High Availability Across Zones
- Regions and Zones: CSPs divide the world into regions, each with independent availability zones.
- Independent Datacenters: Each availability zone has its own power, cooling, and network connectivity.
- Hosting Choices: Data, services, and VM instances can be hosted in specific regions for lower latency and improved performance.
- Redundancy: Provisioning resources in multiple zones and regions increases redundancy.
-
Replication Tiers
- Local Replication: Data is replicated within a single datacenter in the region where the storage account was created, often in separate fault and upgrade domains.
- Regional Replication (Zone-Redundant Storage): Data is replicated across multiple datacenters within one or two regions, safeguarding against single datacenter failures.
- Geo-Redundant Storage (GRS): Data is replicated to a secondary region distant from the primary region, protecting against regional outages or disasters.
Important terms
taken from the text above- High Availability (HA)
- Storage provisioned with a guarantee of 99.99% uptime or better.
- Redundancy
- Multiple disk controllers and storage devices are available to a pool of storage resources.
- Replication
- Data may be replicated between pools or groups, supported by separate hardware resources.
- Central Storage
- The cloud can serve as a central storage area, making data available to all business units.
- Storage Tiers
- CSPs offer various data storage performance tiers, such as hot storage (quick retrieval, higher cost) and cold storage (slower retrieval, lower cost).
- Application Needs
- Different applications have diverse replication requirements, such as low-latency, synchronous replication for databases.
- Regions and Zones
- CSPs divide the world into regions, each with independent availability zones.
- Independent Datacenters
- Each availability zone has its own power, cooling, and network connectivity.
- Hosting Choices
- Data, services, and VM instances can be hosted in specific regions for lower latency and improved performance.
- Local Replication
- Data is replicated within a single datacenter in the region where the storage account was created, often in separate fault and upgrade domains.
- Regional Replication (Zone-Redundant Storage)
- Data is replicated across multiple datacenters within one or two regions, safeguarding against single datacenter failures.
- Geo-Redundant Storage (GRS)
- Data is replicated to a secondary region distant from the primary region, protecting against regional outages or disasters.
Examples & real-world scenarios
Supplementary — not from your PDF- Running VMs in two availability zones behind a load balancer.
- Geo-redundant storage copying backups to another region.
- Moving old logs to cold storage to save money.
Scenario
A regional power outage takes down a whole region. Only data with geo-redundant replication stays available. Local and zone-redundant copies were in the affected region.
Common mistakes
Supplementary — not from your PDF- Thinking two VMs in the same zone survive a zone outage.
- Mixing up zone-redundant (within a region) and geo-redundant (to a distant region).
Practical skills
Supplementary — not from your PDF- Choose a replication tier to match a resilience requirement.
What I should remember
Key Points PDF p.160-
High Availability (HA)
- 99.99% Uptime: Guaranteed storage availability.
- Redundancy: Multiple disk controllers and storage devices.
-
Replication
- Central Storage: Data available to all business units.
- Requirements: Low latency, security, data integrity.
- Storage Tiers: Hot vs. cold storage.
- Application Needs: Diverse replication requirements.
-
High Availability Across Zones
- Regions and Zones: Independent datacenters.
- Hosting Choices: Lower latency, improved performance.
- Redundancy: Multiple zones and regions.
-
Replication Tiers
- Local Replication: Within a single datacenter.
- Regional Replication: Across multiple datacenters.
- Geo-Redundant Storage: Secondary distant region.