Failure-domain isolation
Multi-zone services, durable queues, health-based failover and independently recoverable tenant storage reduce the blast radius of component failure.
Choose the tenant, network, region, keys, model-processing path and operator-access policy that match your risk model—without losing one normalized view across a global estate.
Every deployment uses tenant-scoped identity, encryption and audit controls. Regulated environments can increase infrastructure, network and key isolation without changing the operating workflow.
| Control | Enterprise SaaS | Dedicated managed tenant | Customer cloud / air-gapped |
|---|---|---|---|
| Compute | Regional, horizontally isolated service with tenant-scoped workloads | Dedicated application and ingestion resources | Runs in the customer-controlled account/VPC or approved offline environment |
| Database | Logical tenant isolation enforced in application and row-level policy | Dedicated database and backup boundary | Customer-owned database and storage boundary |
| Encryption keys | Service-managed keys with scheduled rotation | Dedicated tenant key; customer-managed key option | Customer KMS/HSM and key policy |
| Network | Authenticated TLS over approved FQDNs | Private ingress and egress controls available | PrivateLink/private endpoints, proxy-only or no-internet operation |
| Reasoning model | Managed regional model or deterministic-only | Dedicated approved model path | Customer Bedrock, Azure OpenAI, Vertex or local inference |
| Operations | Araghatta operated under audited access policy | Named change windows and customer-approved support access | Customer-operated or jointly managed under documented runbook |
| Data residency | Selected supported service region | Contracted tenant region and backup region | Customer chooses region, sovereign boundary and replication policy |
The management plane coordinates tenant policy and configuration. Regional data planes ingest and store telemetry inside the selected boundary. Export and response paths are explicit and independently controlled.
ESTATEEndpoints, servers & cloudSigned sensors, workload collectors and read-only cloud connectors produce normalized, redacted evidence.INGESTRegional data planeAuthenticated ingestion, schema validation, deduplication, bounded queues and encrypted durable storage.INTELLIGENCERisk & threat engineDeterministic policy, CVE/threat correlation and optional tenant-approved reasoning.GOVERNTenant control planeInventory, policy, evidence, approvals, exceptions, RBAC and immutable administrative audit.OPERATESOC & responseSIEM, SOAR, ITSM, API and a separately authorised response channel with rollback.Multi-zone services, durable queues, health-based failover and independently recoverable tenant storage reduce the blast radius of component failure.
Encrypted backups, documented RTO/RPO by deployment tier, scheduled restore exercises and customer-visible recovery evidence.
Bounded local spooling, exponential backoff, deduplication and reconciliation prevent a temporary control-plane issue becoming silent data loss.
Customer identity remains authoritative. Human and non-human access is least-privilege, time-bound where required and represented in the same immutable audit trail.
Okta, Entra ID, Google Workspace, Auth0 and standards-compatible OIDC providers with domain enforcement. SAML is on the roadmap.
Provision, update, group-map and deactivate users automatically. Offboarding can trigger AI-access review and revocation workflows.
Viewer, Analyst, Admin, Owner and scoped custom roles with separation of duties for policy, evidence and response.
Disabled by default, reason-bound, time-limited, least-privilege and recorded. Emergency access is separately governed and reviewed.
The discovery sensor cannot execute arbitrary commands. Response uses a narrow, authenticated policy channel with explicit target, action, approval, expiry and rollback semantics.
| Stage | Required control | Evidence | Failure behavior |
|---|---|---|---|
| Detect | Read-only or metadata-focused collection under tenant policy | Asset, signal, identity, source and timestamp | Spool locally and retry without weakening device security |
| Simulate | Dry-run evaluates policy against affected assets before activation | Projected actions, exceptions and blast-radius summary | No endpoint action |
| Approve | Role, ticket/reason, target scope, duration and separation-of-duties rule | Approver, change reference and before-state | Expired or incomplete approvals fail closed |
| Act | Pause or server-side revoke today; warn, redact, block and quarantine via the customer-authorised policy path (roadmap) | Target acknowledgement, result and after-state | Bounded retry; no arbitrary shell or remote desktop channel |
| Rollback | Versioned policy and emergency disable at tenant and fleet level | Rollback actor, reason, affected assets and completion state | Revert to the last known approved policy |
Exact contractual objectives are stated in the enterprise order form and architecture pack, including regional availability, recovery, retention and support commitments.
Enterprise evaluations receive a controlled package aligned to the selected deployment model and data boundary.
Threat model, data flow, network paths, IAM, encryption, key lifecycle, support access and response design.
DPA, SCCs, subprocessors, field dictionary, retention/deletion policy, DPIA and works-council support.
SLA, support policy, BCP/DR, RTO/RPO, maintenance, lifecycle, deprecation and incident communications.
Deployment guides, exact permissions, API/event schemas, test plan, rollback plan, SBOM and sample evidence.
We will map your identity, MDM/UEM, SIEM, cloud hierarchy, network boundary, data residency and change-control requirements into one evaluation plan.