Why traditional DR approaches are insufficient in modern conditions?
For Ukrainian businesses, reliance on a single data center or region creates significant vulnerabilities, as a localized event can completely paralyze operations. Traditional DR solutions, while providing data recovery, often come with high RTO (Recovery Time Objective) and RPO (Recovery Point Objective) metrics, implying extended downtime and potential loss of significant data volumes. This is unacceptable for critical business processes that demand near-instant recovery and minimal information loss.
Businesses require an architecture that ensures active application resilience. This means not just the ability to restore operations, but the capacity to continue them without significant interruptions, even if a component or an entire region becomes unavailable.
Geographically distributed IT systems: Key concepts
Geographically distributed IT systems are architectural solutions that involve placing IT infrastructure components in multiple geographically distant locations. The primary goal of such systems is to enhance resilience against local failures, natural disasters, cyberattacks, or other regional threats. The advantages of distributed systems lie in their ability to provide high availability and business continuity through redundancy and fault tolerance.
Key components of geographically distributed systems include:
- Data Replication: A mechanism for copying data between different locations to ensure its currency and consistency. This can be synchronous or asynchronous replication, which impacts RPO. Synchronous replication ensures RPO=0 1, but increases latency and cost.
- Load Balancing: Distributing incoming requests among available servers in different locations to optimize performance and ensure continuous operation.
- Monitoring and Automatic Failover: Monitoring systems track the status of components, and in the event of a failure, automatically redirect traffic to healthy nodes in another location.
Active-Active architecture: Advantages and challenges
An Active-Active architecture implies that all geographically distributed locations actively process workloads simultaneously. This means users can connect to any available node, and the system will continue to function even in the event of a complete failure of one location. The main advantages of Active-Active are:
- Near-zero RTO and RPO: In ideal conditions, switching between active nodes happens instantly, ensuring minimal downtime and virtually no data loss. This depends on replication technologies and network infrastructure.
- High resilience and availability: The system remains fully functional even with the failure of an entire data center.
- Efficient resource utilization: All resources are constantly in use, enhancing overall performance and return on investment.
- Load balancing: Allows traffic distribution across locations, optimizing performance and scalability.
However, Active-Active architecture presents significant challenges:
- Complexity in ensuring data consistency: Maintaining absolute real-time data consistency across multiple active nodes is a very complex task, especially for transactional systems. The CAP theorem problem arises.
- High network requirements: A high-speed, low-latency network is required between locations for effective synchronous replication and load balancing.
- Complexity of management and development: Requires significantly more complex configuration, monitoring, and application development to operate in a distributed environment.
- Higher cost: Requires double or triple the amount of hardware and software resources, as well as highly skilled personnel for support.
Active-Passive architecture: A compromise between availability and cost
In an Active-Passive architecture, one location (active) handles all requests, while the other (passive) remains in standby mode. The passive location continuously receives replicated data from the active one but does not process workloads until the active location fails. In the event of an active node failure, a failover occurs to the passive node, which then becomes active.
Advantages of Active-Passive architecture:
- Simpler implementation: Less complex in design, implementation, and management compared to Active-Active.
- Lower initial costs: Although the passive node requires resources, overall costs can be lower as not all resources operate simultaneously.
- Easier to ensure data consistency: Since writes occur only on one active node, maintaining data consistency is simpler.
- Clearer failover procedures: The failover process is clearly defined.
Limitations of Active-Passive architecture:
- Higher RTO and RPO: The time required to detect a failure, perform a switchover, and initialize the passive node can lead to noticeable downtime. With asynchronous replication, there is a risk of data loss that did not replicate before the failure. Synchronous replication can minimize data loss (RPO=0) 1, but this increases latency and cost.
- Need for failover procedures: Requires clearly defined and tested failover procedures.
- Inefficient resource utilization: Resources in the passive location sit idle, awaiting a failure, which reduces investment efficiency.
Selection criteria: Active-Active vs. Active-Passive for your business
The choice between Active-Active and Active-Passive architectures depends on application criticality, budget, RTO/RPO requirements, and management complexity. To make an informed decision, it's important to evaluate each aspect:
| Criterion | Active-Active | Active-Passive |
|---|---|---|
| RTO/RPO | Near-zero (seconds/minutes) | Higher (minutes/hours) |
| Total cost of ownership (TCO) | High (requires double resources, complex network) | Medium (passive node resources, simpler network) |
| Management and support complexity | High (data consistency, load balancing, monitoring) | Medium (replication management, failover procedures) |
| Data consistency | Difficult to ensure absolute real-time consistency | Easier to ensure (one active node for writes) |
| Network infrastructure requirements | High (low latency, high bandwidth between locations) | Medium (depends on replication type) |
| Personnel qualification requirements | High (experience with distributed systems, databases, networks) | Medium (experience with DR, replication) |
Choose Active-Active if: your applications are critically important, require near-zero RTO/RPO, have high availability and scalability demands, and your budget allows investment in complex infrastructure and qualified personnel.
Choose Active-Passive if: you are prepared for short-term downtime (minutes/hours) and potential loss of a small amount of data, are looking for a simpler and more economical solution, and your applications do not require continuous operation of all nodes simultaneously.
The optimal choice of IT infrastructure for Ukrainian businesses often involves a combination of different approaches, considering specific risks and regulatory requirements (e.g., NBU requirements for the financial sector) 2.
Practical recommendations and implementation checklist
Before choosing an architecture, a thorough analysis of your application criticality is necessary. Determine which systems are vital for business continuity and what RTO/RPO are acceptable for them.
Application criticality assessment checklist:
- Define RTO and RPO: What is the maximum acceptable downtime and maximum amount of data loss for each application?
- Impact of downtime on business: What financial and reputational losses would occur if an application is down for 1 hour, 4 hours, 24 hours?
- Application dependencies: Are there dependencies between applications that might affect the recovery sequence?
- Performance requirements: Is high performance and scalability required across all locations simultaneously?
- Budget constraints: What budget is available for implementing and maintaining a distributed architecture?
- Regulatory requirements: Are there industry or legal requirements for business continuity and data retention?
Technology selection recommendations:
- For Active-Active architecture, consider synchronous data replication technologies (e.g., for databases) and global server load balancers (GSLB).
- For Active-Passive, depending on RPO, asynchronous replication or Disaster Recovery as a Service (DRaaS) solutions can be used.
- It is crucial to invest in a reliable network infrastructure that provides the necessary bandwidth and low latency between locations.
Testing and monitoring:
Regardless of the chosen architecture, regular testing of disaster recovery plans and monitoring the status of all components are critically important. Solutions such as server infrastructure and Disaster Recovery require expertise in planning and implementation.
Softline IT assists in planning and implementing server infrastructure solutions: from auditing the current state to an agreed-upon plan for changes.
Softline IT helps teams plan and implement server infrastructure, from an assessment of the current environment to an agreed change plan.
