- sara.next
- Technology
- Scalability and availability
Technology · Architecture
Scalability and availability
In IGA, scaling is not just about the number of accounts. The volumes of requests, review campaigns, the number of integrated applications and the reliable execution of permission changes matter just as much.
- sizing
- integration queues
- load testing
Context
Scaling processes, not just the identity register
An access management system handles uneven loads. A typical day involves individual requests and approvals, but a quarterly review can trigger thousands of decisions in a short time. Integrations, in turn, may have their own API limits, maintenance windows, database constraints and temporary unavailability. These phenomena should be taken into account separately when designing the environment.
sara.next provides a central process for managing requests, roles and attributes, reviews and activity history. For supported auto-implementation integrations, once a request is approved the change goes into a queue executed by the scheduler.
NoteThe technical documentation describes the recurring processing of this queue every minute. This is a description of a specific mechanism, not a guarantee of completion time or a statement of maximum throughput.
Design
What to take into account in the architecture
Interactive
Sign-in, searching for access, creating requests, the manager and approver panels.
Batch
Initial access imports, dictionary synchronisations, report generation and recertification campaigns.
Integration
REST requests and SQL operations executed after the authorisation process is complete.
Dependencies
The SSO provider/IdP, database, mail services, DNS, target systems, network connections and external data sources.
Reliability
The ability to identify a pending request, an implementation error and discrepancies between the target state and the IGA records.
Availability
The availability model is agreed during implementation
The recommended analysis covers single points of failure (SPOF), RTO and RPO, backups, database recovery and the procedure for resuming queue execution after a failure. In highly critical environments redundant components should be considered, but simply running several application instances does not prove support for active–active clustering. Such a topology must be confirmed for the specific version, including schedulers, job locking and shared state.
| Architecture question | What should be documented during sizing |
|---|---|
| How many identities and applications? | The number of active identities, roles, entitlements and target systems |
| What do the peaks look like? | The number of concurrent sessions, review tasks, requests per hour |
| How fast should operations be? | Target response time percentiles and the maximum time to make changes |
| What are the integration dependencies? | API limits, timeouts, availability of target systems, DB throughput |
| How do we recover? | RTO, RPO, backup/restore procedures, emergency stop of implementation |
| How do we monitor? | Latency measurements, error counts, queue backlog, review time, DB health |
Scenario
A review campaign in a large organisation
The organisation launches a quarterly recertification for many financial applications. The biggest increase in load does not have to occur at sign-in; it may be related to mass notifications, record searches and report generation. That is why the performance test should reproduce the entire campaign.
- Stage 1Campaign launchMass notifications.
- Stage 2Reviewer decisionsRecord searches.
- Stage 3Escalations
- Stage 4Corrective actions
- Stage 5Summary reportReport generation.
An implementation expert's view
The number of users is not the only measure
I would not accept the "number of users" as the only measure of IGA scalability. What matters more is whether the platform completes the control process within the required window and whether, after a failure, it leaves no permission changes in an ambiguous state.
Before production, both the interface and the real target systems should be tested, including errors and restarting jobs.
Scope
Limits and conditions
sara.next has experience in corporate environments and handles complex processes.
NoteWe do not publish general TPS figures, numbers of concurrent sessions, availability SLAs, RTO/RPO or an HA topology matrix.
Such data should come from current tests of the specific version and the agreed architecture.
We will prepare the assumptions for sizing and architecture testing
Tell us the volume of identities, the number of applications and the profile of your review campaigns.