Palo Alto Networks’ Strata Logging Service is the current name for the platform previously known as Cortex Data Lake. It is a cloud-delivered log storage and aggregation layer for Palo Alto Networks telemetry, including next-generation firewalls, Prisma Access, Cloud NGFW, and selected Cortex services. The operational value is not simply “sending logs to the cloud.” The value comes from building a dependable telemetry path in which devices produce the right records, the logging service receives them, retention is sized deliberately, analysts can retrieve them quickly, and forwarding to downstream systems remains observable. In the broader security operations stack, the logging layer is the evidence plane on which investigations, hunting, reporting, and many automation decisions depend.
For readers studying Palo Alto Networks SecOps-Pro, that makes logging architecture a security-control problem rather than a storage problem. A firewall can enforce excellent policy while the organization still loses investigative visibility because a forwarding profile was omitted, a device stopped sending, retention was undersized, or a downstream connector silently fell behind. Teams should therefore treat Strata Logging Service as an engineered pipeline with explicit source coverage, transport health, retention objectives, query practices, forwarding dependencies, and recovery procedures.
This perspective also keeps the service in the right place relative to a SIEM. Strata Logging Service is optimized for Palo Alto Networks telemetry and integrates tightly with management and security products. A broader cloud-native SIEM architecture may ingest Palo Alto data alongside identity, endpoint, application, SaaS, and cloud-platform events. The designs are complementary when responsibilities are explicit: Strata Logging Service provides durable, product-aware collection and retrieval, while downstream analytics may correlate across a wider enterprise dataset.
Start with source coverage, not with storage
Reliable logging begins by listing which products, tenants, devices, and log types are supposed to contribute evidence. Traffic, Threat, URL Filtering, WildFire Submission, Data Filtering, Authentication, System, Configuration, User-ID, HIP Match, and other records do not all follow the same configuration path. Some are emitted when a security rule matches and therefore depend on log-forwarding profiles attached to rules. Others are configured under device log settings. If teams treat “cloud logging enabled” as a single switch, gaps appear in exactly the places an incident later needs.
Source coverage should be expressed as an expected-state matrix. For every device or device group, record the log types required, the configuration object that enables each one, the destination, and the owner. That matrix makes change review much easier. If a new firewall is onboarded, the question is not merely whether it can reach Strata Logging Service; the question is whether its policy rules and device settings generate the same evidence classes as the rest of the estate.
This source discipline complements firewall monitoring, because investigation quality depends on what was logged before the analyst opens a viewer. The most sophisticated query cannot reconstruct fields that were never emitted.
Treat transport health as a measurable control
Once sources are configured, teams need to know whether logs are arriving continuously. Device-side status commands, management-plane indicators, inventory views, and storage dashboards all provide pieces of this answer. A healthy design defines normal forwarding delay and identifies the signals that indicate disconnection, backlog, rejected records, or misconfiguration. The important metric is not only “connected”; it is whether the last created, forwarded, and acknowledged log sequence is advancing at the expected rate.
Transport monitoring should distinguish a source outage from a query problem. If a firewall shows forwarding progress but an analyst cannot find a record, the issue may be filtering, retention, or schema expectations. If the last acknowledged sequence stops advancing, the problem is closer to source connectivity or service health. Separating those failure domains reduces the tendency to troubleshoot the entire logging stack at once.
Network teams should also protect the path itself. Certificates, service routes, DNS, egress controls, and time synchronization can all become hidden dependencies. A change to an upstream proxy or firewall policy can break logging even when the protected application traffic remains healthy. Logging connectivity therefore belongs in change-impact analysis, not just in the SOC runbook.
Retention is a risk decision, not an arbitrary number
Strata Logging Service removes the need to size local collectors for every growth scenario, but it does not remove the need to decide how much history matters. Retention should reflect incident-detection latency, regulatory obligations, forensic expectations, seasonal investigation patterns, and the cost of exporting older data elsewhere. A team that detects account abuse after 45 days will not be helped by a 30-day retention window simply because 30 was the easiest option.
Current Strata Logging Service management exposes storage quota and retention status, and 2026 updates standardized retention choices in Strata Cloud Manager. That simplifies administration, but it makes the governance decision more visible: which logs deserve the longest online retention, and which can be forwarded to lower-cost archival storage? The answer may differ by log type and business environment.
Retention changes should be treated like policy changes. Record who approved them, why they were made, what investigation scenarios depend on the old window, and whether external archives still satisfy the requirement. Purging is automatic once the retention policy applies, so the organization cannot rely on ad hoc recovery after the fact.
Query design should preserve investigative context
Centralized logs are valuable only if analysts can retrieve the right records without turning every incident into a manual search project. Query standards should define common fields, time-bounding habits, identity pivots, source and destination conventions, and techniques for excluding known noise. Strata Logging Service added ILIKE, NOT ILIKE, and NOT LIKE operators in 2026, which improves pattern handling and case-insensitive filtering for string and enum fields.
Query convenience should not become a substitute for analytical method. A strong investigation starts with a hypothesis, narrows the time window, identifies authoritative fields, and preserves the sequence of events. When analysts jump directly to broad keyword searches, they often mix unrelated events or miss the transition that matters. SIEM alert triage depends on chronology: analysts need the sequence of records that changed the incident state, not merely a collection of matching events.
Unified Logs in Strata Cloud Manager can correlate up to several log types in one table, reducing the need to pivot between screens. That helps when a single incident crosses traffic, threat, URL, or other record classes, but analysts still need to understand which product generated each field and what the absence of a field means.
Forwarding creates another production pipeline
Many organizations forward Strata Logging Service data to SIEM platforms, syslog receivers, HTTPS endpoints, email, object storage, or other destinations for broader correlation, long-term retention, or audit. That forwarding layer needs its own health model. A successful source-to-SLS path does not prove that SLS-to-SIEM delivery is healthy.
Downstream designs should document authentication, destination rate limits, payload sizes, retry behavior, buffering, schema mapping, and ownership. Palo Alto Networks added direct Microsoft Sentinel forwarding through the Azure Monitor Ingestion API in 2026 and expanded several other forwarding capabilities. Those integrations simplify connectivity, but they do not eliminate the need to monitor delays and rejected payloads.
Rate limits matter during spikes. If a security event causes a large increase in log volume, the same event can stress the forwarding mechanism analysts depend on. Capacity testing should therefore include burst scenarios, not only average daily volume.
Operational failure modes deserve explicit runbooks
Common failures include new devices that were onboarded without complete log settings, policy rules missing forwarding profiles, retention changes that shorten investigative history, downstream destinations rejecting data, filters that exclude needed records, and analysts assuming Cortex XDR alerts will appear in the same viewer as firewall logs. These are configuration-boundary failures more often than platform failures.
Runbooks should start with evidence that separates the layers: Is the device producing the log? Is it forwarding? Is Strata Logging Service acknowledging? Is the log retained? Does the query target the right fields and time? Is the external destination receiving? Each answer narrows the fault domain.
This layered troubleshooting approach also reflects the broader security-operations architecture principle that observability systems create second-order dependencies. A team cannot make detection, automation, or response reliable unless the evidence path itself is operated as a service.
A mature logging program measures evidence quality
Useful metrics go beyond gigabytes stored. Track source coverage, percentage of devices with expected log types enabled, forwarding lag, acknowledgement gaps, retention exceptions, failed downstream deliveries, query latency for common investigations, and the number of incidents affected by missing telemetry. These measures expose whether the logging program is improving investigative reliability rather than merely accumulating data.
Review these metrics with both network and security operations teams. The source device owner controls much of the evidence generation, while the SOC controls many of the search, retention, and downstream-use decisions. Shared ownership prevents the familiar situation in which one team assumes the other is monitoring the pipeline.
Strata Logging Service works best when it is treated as a deliberate evidence plane inside the wider Palo Alto Networks architecture. The target is not maximum log volume. The target is trustworthy, timely, queryable evidence that survives the operational failures most likely to occur during the incident that matters.