A managed cloud service
Cisco-hosted Crosswork Assurance gives customers the benefits of a continuously operated assurance platform without requiring them to deploy, patch, monitor, or recover the hosted infrastructure themselves. Cisco Site Reliability Engineers administer the cloud offering, monitor the service 24×7, and respond when monitoring identifies a service issue. Cisco also delivers software enhancements and fixes through a continuous-delivery release that is updated every few weeks.
The hosted service uses public-cloud infrastructure, container orchestration, automated health monitoring, and resilient storage to reduce the impact of individual application, compute, and storage failures. Cisco operates these layers as part of the service so that customer teams can focus on assurance workflows and network outcomes rather than another big data platform infrastructure.
For current delivery information, see Cloud Continuous Delivery Release Notes. For an overview of the managed analytics service, see Crosswork Assurance Analytics.
How data moves through the service
Crosswork Assurance accepts performance and telemetry data from several customer-side sources. Hardware and embedded Assurance Sensors report through the Legacy Orchestrator before their data reaches the Sensor Collector. Supported software Assurance Sensors and telemetry sources can use direct Sensor Collector paths. The Sensor Collector provides the secure telemetry path into the hosted service. An on-premises Gateway connects the customer provisioning system to both the Legacy Orchestrator and Cisco-hosted cloud services. The Gateway can also support alerting and streaming integrations.
Customer connections reach the hosted service through an external interface and load balancer. From there, performance and telemetry data is routed to the ingestion pipeline, while provisioning and configuration traffic is routed separately to control and integration services. Alerts and supported streams can flow back through the same external interface. It's also important to call out the separate handling of late data - the default threshold between current-data and latent-data processing is two hours.
The diagram is a logical service view. Exact customer-side components vary by deployment and by the sources being monitored. For additional component descriptions, see Sensor Collector and Data Collection.
Layered resilience across the data path
Crosswork Assurance separates data collection, ingestion, processing, storage, and presentation into distinct service stages. Durable queues between processing stages help isolate temporary downstream interruptions. When a stage recovers, retained data can resume processing, subject to the retention capacity of the source and collection path. Current observations remain the priority while older backlog data catches up.
For a detailed platform view, see Architecture Overview.
Resilience across the hosted service
Resilience is provided at several layers rather than by a single component.
Layer | How resilience is provided | What the customer can expect |
|---|---|---|
Service workloads | Health monitoring and container orchestration restart or reschedule affected workloads where supported. | A component failure is normally handled by Cisco without customer infrastructure work. A brief interruption or data delay may still occur. |
Cloud compute | Failed compute instances can be replaced and service workloads restored on healthy capacity - most of which is automated | Service processing resumes after the affected workload is restored. |
Storage | Hosted block storage uses multi-zone persistent disks that replicate data across availability zones within one cloud region. Object storage is protected separately by the managed object-storage service. | An individual physical disk or single-zone storage failure does not require customer action and is not expected to result in customer data loss. A service interruption may still occur while affected workloads recover. |
Data processing | Current and older data can be handled through separate processing paths so that recovery traffic does not overwhelm current processing. | Contemporary data remain the priority while retained latent data is processed into the historical analytics view at a slower pace. |
Service operations | Monitoring alerts Cisco cloud operations and Site Reliability Engineers to potential and active service problems. | Cisco investigates and restores the hosted service, advertising impact through in-app and email messaging; customers can also engage Cisco Support if they observe an issue. |
Data resiliency and continuity
Crosswork Assurance emphasizes data resiliency as well as service availability. The objective is to preserve source data through temporary downstream interruptions and resume delivery when connectivity or processing is restored.
Supported customer-side collection paths retain data close to the source. When communication is restored, retained data resumes flowing toward the hosted service. Legacy Sensor Management workflows can store performance data in binary files with a default retention period of one week, subject to available disk capacity and configuration. Some Assurance Sensors also buffer measurements locally; their retention time depends on the sensor model, firmware, configuration, and collection interval.
To protect current visibility during recovery, real-time data is prioritized over older backlog on supported legacy collection paths. Older data is sent in a controlled manner and may arrive out of chronological order. In the hosted service, current data and latent data are processed separately so that a large backlog does not prevent contemporary observations from reaching analytics. As retained data is ingested, it can fill temporary gaps in the historical view without displacing current processing.
Streaming export services follow the current-data path and do not replay latent or backfilled records. For example, a gNMI subscription does not synchronize data received before the subscription began.
Recovery processing handles duplicate and out-of-order records so that retained backlog can be reconciled with current data. This approach is designed to reduce data gaps during temporary interruptions, subject to the retention capacity of the source, collector, or gateway and the duration of the interruption. It does not provide a zero-data-loss guarantee. Customers should size their customer-side retention for the outage window they need to cover.
For documented collection and recovery behavior, see Metrics Collection and Export and Guidelines for Collecting Metrics from Assurance Sensors.
Data freshness during normal and recovery operations
Data freshness is the time between generating a measurement and making it available for query in the analytics UI or APIs. Crosswork Assurance aims for approximately five minutes of data-processing latency, although actual availability depends on the source collection interval, batch size, and processing volume.
During a reconnection or backlog recovery, current data remains the priority. Data beyond the default two-hour threshold is processed through the latent-data path and is added to historical analytics over time. This prevents a large recovery backlog from overwhelming the current-data path and helps maintain a useful view of present network conditions while history catches up.
For more information, see Data Freshness.
What happens during common interruptions
Event | Cisco response | Customer experience and action |
|---|---|---|
Hosted application workload stops | Monitoring detects the unhealthy workload, and the platform restarts or reschedules it where supported. | A short interruption or increased data delay may occur. No customer infrastructure action is required. |
Hosted compute instance becomes unavailable | Cisco restores affected workloads on healthy cloud capacity. | Processing resumes after restoration; buffered or retained data may continue to catch up afterward. |
Hosted physical disk or single-zone storage fails | Cisco restores affected workloads on healthy cloud capacity. | Processing resumes after restoration; buffered or retained data may continue to catch up afterward. |
Customer-to-cloud connectivity is interrupted | The hosted service remains under Cisco operation while the customer restores its connectivity. | Supported customer-side components retain data within their configured capacity. After reconnection, current data is prioritized and retained backlog resumes flowing. |
Customer-side source or collector fails | Cisco Support can help isolate whether the hosted service is healthy. | The customer restores the affected on-premises component unless it is covered by a separate Cisco CX or partner-managed service. |
Planned cloud update | Cisco applies enhancements and fixes through the continuous-delivery process. | Updates are managed by Cisco, and notifications are published in-app in advance. A reconnect or temporary delay may occur when required by a particular change. |
Responsibility boundary
Everything deployed in the customer environment is customer-operated unless a separate agreement assigns it to Cisco CX or a managed service partner.
Area | Cisco-hosted service | Customer environment |
|---|---|---|
Hosted platform | Operate, monitor, update, and recover the cloud application and supporting infrastructure. | No hosted-infrastructure administration required |
Hosted data protection | Operate hosted storage, backups, and supported recovery procedures. | Confirm customer-specific retention and recovery requirements through the applicable service agreement. |
Sources and collectors | Provide documented interfaces and supported ingestion behavior. | Operate network devices, Assurance Sensors, Legacy Orchestrator, Telemetry Collector, Sensor Collector, and the on-premises Gateway unless separately managed. |
Provisioning integration | Provide the supported hosted integration endpoint and documentation. | Operate the provisioning system and its integration with the on-premises Gateway if available. |
Connectivity and trust | Operate the hosted endpoint and Cisco-side certificates. | Operate routing, firewall, proxy, DNS, bandwidth, and customer-side certificate trust. |
Incident support | Investigate hosted-service issues and coordinate recovery through Cisco Support and service operations. | Use in-application support or open a Cisco support case and provide the affected service, time, symptoms, and relevant customer-side details. |
Service objectives
The service-level objective and customer-specific service commitments for Cisco-hosted Crosswork Assurance are defined in the applicable contract. This article explains the resilience design and operating model; it does not replace or modify the contracted SLO.
Frequently asked questions
Do we manage the hosted infrastructure?
No. Cisco operates, monitors, updates, and recovers the hosted Crosswork Assurance platform in cloud. Customers remain responsible for components and connectivity in their own environment unless those components are covered by a separate managed-service agreement.
What happens if a hosted virtual machine or service fails?
Cisco monitors the hosted service and restores affected workloads on healthy infrastructure. Customers may observe a brief interruption or data delay while recovery completes, but they do not need to rebuild the hosted component.
Is our data lost if a hosted disk fails?
No. Hosted block storage uses multi-zone persistent disks that maintain redundant copies across availability zones within the same cloud region. An individual physical disk or single-zone storage failure is therefore not expected to result in customer data loss. This protection does not provide cross-region replication. Cisco restores any affected storage or workload as part of operating the service, and customers may experience a temporary service interruption while recovery completes.
Will data catch up after an interruption?
Supported customer-side sources and collection components retain data within their configured capacity. When connectivity returns, current data is prioritized and retained backlog resumes flowing. Older data is processed through the latent-data path and can fill temporary gaps in historical analytics over time. Recovery processing handles duplicate and out-of-order records during catch-up. Export services do not replay these latent or backfilled records, and the recovery behavior is not a zero-data-loss guarantee. Retention capacity and recovery behavior vary by source and configuration.
How do we know whether the service is healthy?
Current data in the analytics UI is an important indicator of end-to-end health. If data appears stale or the application is unavailable, customers can use the in-application support experience to contact Cisco Support or open a Cisco support case. Cisco Support can engage service operations and Site Reliability Engineers when investigation of the hosted service is required.
For support options and instructions, see Contact Technical Support.
How often is the cloud service updated?
The cloud continuous-delivery release is updated every few weeks with software enhancements and fixes. Cisco Site Reliability Engineers handle installation and administration of the cloud offering. See Cloud Continuous Delivery Release Notes for the current release history.