Guide to Evaluating Software Architecture Characteristics

Performance and Scalability

Performance metrics are critical indicators of a system’s effectiveness. Response time and throughput must be carefully monitored and optimized to ensure user satisfaction. Systems should be designed to handle increased load through horizontal or vertical scaling, while maintaining optimal resource utilization across computing resources.

In our architectural recommendations, we consider both the historic and estimated usage levels for a software product.

• For backend services, solutions can involve load balancers and scalable pools of servers that can expand and contract to accommodate usage. Caching services are also important tools for reducing server loads.
• For front-end code and static assets, content delivery networks can be involved to ensure quick responses for targeted geographic areas. Many other services can also help improve performance for specific use cases.

Reliability, Availability, and Recoverability

Modern systems must maintain high uptime to meet user expectations and business requirements. This involves implementing robust fault handling mechanisms and meeting or exceeding defined Service Level Agreements (SLAs). Monitoring tools and redundancy measures play crucial roles in maintaining system reliability.

Load balancing technologies are also frequently recommended as part of the Reliability and Availability solution for a software product. They can be configured to evaluate the health of the servers that deliver the site and avoid ones with issues. Cloud technologies allow solutions that can increase the reliability and availability by distributing traffic across different data centers or regions in the event of large outages.

Recoverability addresses the meeting of Recovery Time Objectives (RTO) and Recovery Point Objective (RPO) of a system. RTO is the maximum acceptable amount of time that it can take to bring the system back online after major outages. RPO is the maximum amount of data that can be lost during an outage. An example of these goals for a low-traffic financial transaction system may include a longer RTO (due to low traffic) but a near-zero RPO because it is critical to lose none of the transactions.

Security and Compliance

Security cannot be an afterthought in modern software architecture. Comprehensive data protection measures, including encryption and secure data handling practices, must be implemented. Strong authentication and authorization mechanisms need to be in place, and all relevant regulatory requirements must be met and continuously monitored.

In the solution design, we first identify the different roles that need access to the system and then determine the authentication and authorization mechanisms to be used. We look for opportunities to work with organizations’ existing mechanisms and processes when it is a good fit. We also work to ensure the use of industry best practices, such as two-factor authentication. Establishing role-based authorization policies can help promote easy maintainability and grant the appropriate access for different parties, using the principle of least access. In addition to users’ access, solutions should consider external threat mitigation technologies, employing tools that scan for vulnerabilities in different tiers of the application.

Compliance can be specific to a department, organization, industry, or jurisdiction. Additionally, there are general best practices to follow. For instance, one area frequently covered in compliance requirements is Data Retention. In our solutions, we take these requirements into consideration, recommending databases and storage products that allow for soft deletes, versioning, and configurable retention policies. Cloud providers offer robust tools for policy and compliance management for many aspects of a solution and organization.

Time, Budget, and Cost Attribution

While not considered a traditional core “software architecture characteristic,” timeline and budget underlie every solution design. It is important to make – and document – the appropriate trade-offs. The architectural characteristics for a small proof of concept or minimum viable product for a start-up would be very different from those for a large enterprise or medical application with life-or-death criticality.

Additionally, it’s always important to monitor the costs of cloud solutions to make sure that the solution remains cost-effective over time. In a large organization, it can be difficult to track costs, so consider a tagging strategy to help organize resources for cost attribution. Cloud providers offer services for monitoring costs in real-time, allowing you to configure alerts when services reach budget thresholds.