Facing a data-related challenge? Explore some of the most common issues and discover how we can help you tackle them, gain deeper insights from your data, and find practical solutions.
Data center infrastructure insights and analytics
Data center operations depend on reliable infrastructure, efficient resource utilization, and well-performing workloads. Q-Sensei connects infrastructure, workload, and facility data to help you investigate issues, analyze performance, assess capacity, and improve operational efficiency.Learn more
What is causing an infrastructure or workload issue?
The challenge
An application slows down, a server reports errors, or a workload becomes unstable. I need to understand where the problem originates and which workloads or services are affected. The cause may involve several connected components, such as GPU contention, memory pressure, network congestion, storage latency, or application performance.
How Q-Sensei helps
Q-Sensei connects infrastructure inventory, topology, performance metrics, logs, and traces to support investigations across related components. Examine infrastructure events alongside workload activity, compare metrics across components, and follow dependencies to narrow down where an issue may originate.
AI-assisted analysis can help you explore possible contributing factors and prioritize troubleshooting steps. You gain a more complete view of the evidence across infrastructure and workloads, making it easier to investigate incidents and determine where further action may be needed.
Follow the affected workload and align the events
Workload ARuns on server N12
Server N12Uses storage S04
Storage S04Shared dependency
Events · Shared window: 10:00–10:10
Workload AResponse warning · 10:06
Server N12Storage wait event · 10:04
Storage S04Queue alert · 10:02
10:0010:0510:10
Review dependencies and the recorded sequence in the same time window before attributing an incident. Sample data.
Data typically used in this investigation
Infrastructure inventory and topology
Inventory identifies the servers, accelerators, storage systems, and network components involved. Topology and dependency information helps you trace relationships between affected components and workloads.
Infrastructure performance metrics
CPU, GPU, memory, network, and storage metrics show how infrastructure behaved during the incident. Comparing these measurements helps identify resource constraints or performance changes that may be related to the issue.
Workload and application telemetry
Workload metrics, application performance data, logs, and traces describe how services and workloads behaved. They help connect infrastructure conditions with application-level symptoms.
Events and configuration history
Event, alert, configuration, and change records provide a timeline of operational activity. They help you investigate whether a recent event or change coincided with the issue.
Where are infrastructure bottlenecks affecting workload performance?
The challenge
A workload may have sufficient compute capacity but still perform below expectations because of memory bandwidth, GPU availability, storage throughput, or network latency. I need to identify which resources limit performance, which workloads are affected, and whether the constraint is temporary or recurring.
How Q-Sensei helps
Analyzing infrastructure performance alongside workload runtime data reveals where resource constraints coincide with performance degradation. Compare CPU, GPU, memory, network, and storage metrics with workload activity to investigate how individual resources affect execution.
Historical analysis helps you recognize recurring bottlenecks and compare conditions across workloads and infrastructure components. You get a clearer basis for investigating resource contention and assessing whether infrastructure or workload configuration changes may improve performance.
Examine a constrained resource alongside workload performance
Utilization · 0–100% of each resource’s capacity
CPU38%
Memory57%
Storage throughput92% · Investigate
Network28%
Response times · Same 0–200 ms scale; lower is better
Workload A
Workload B
10:0010:10
Utilization readings cover the same observation window.
Check storage activity during the slower workload’s requests. Sample data.
Data typically used in this investigation
Compute and accelerator metrics
CPU and GPU utilization, availability, and performance measurements show how processing resources are being used. They help identify compute constraints and accelerator contention affecting workloads.
Memory metrics
Memory usage, bandwidth, and pressure measurements describe memory demand and availability. They help reveal whether memory constraints coincide with slower workload execution.
Network telemetry
Throughput, latency, packet loss, and network utilization measurements describe connectivity and data transfer performance. They help identify network conditions that may limit workload performance.
Storage performance data
IOPS, throughput, latency, and capacity measurements show how storage systems respond to workload demand. They help identify storage constraints that may affect application response times.
Workload performance data
Runtime metrics, application latency, throughput, and workload activity describe actual execution behavior. Comparing these signals with infrastructure metrics helps connect resource conditions to workload performance.
Resource allocation and historical records
Allocation records, infrastructure topology, and historical performance data show how resources are assigned and how conditions change over time. They help investigate shared-resource contention and recurring bottlenecks.
Which changes in infrastructure behavior need attention?
The challenge
Infrastructure behavior changes as workloads shift and resource demand fluctuates. A short-lived increase in utilization may be normal, while a gradual rise in latency, error rates, temperature, or resource consumption may indicate a developing operational issue. I need to distinguish normal variation from unusual behavior and understand whether changes could affect reliability.
How Q-Sensei helps
Q-Sensei provides a view of infrastructure performance over time, supported by historical analysis and configurable threshold- and rule-based alerts. Compare behavior across infrastructure components and investigate events associated with unusual metrics.
Statistical analysis, anomaly detection, and forecasting capabilities can help you examine deviations and changing trends. You can recognize patterns that warrant attention, investigate their operational context, and prioritize issues for further analysis.
Compare a new reading with its historical context
Response times · Shared 0–100 ms scale
Historical band: 40–60 msNew reading: 85 ms
Earlier observationsLatest reading
Examine the surrounding recordsWorkload demand · Configuration changes · Related events
A reading outside the historical band prompts review of related events and workload activity. Sample data.
Data typically used in this investigation
Infrastructure telemetry
Continuous measurements describe the operating conditions of servers, accelerators, storage, and network components. They help establish normal behavior and identify changes that may need investigation.
Compute and accelerator metrics
CPU and GPU utilization and performance data show how processing demand changes over time. They help identify unusual resource consumption or changes in accelerator behavior.
Memory, storage, and network metrics
Usage, latency, throughput, and error measurements describe the behavior of shared infrastructure resources. They help reveal developing constraints or performance deviations.
Environmental measurements
Temperature, humidity, and other facility readings describe the conditions around infrastructure equipment. They help investigate whether environmental changes coincide with unusual infrastructure behavior.
Workload and event histories
Workload activity, system logs, alerts, and incident records provide context for changes in infrastructure metrics. They help distinguish workload-driven variation from other operational patterns.
Are workloads using infrastructure resources efficiently?
The challenge
Data center resources are shared across workloads with different performance and capacity requirements. Some workloads may consume substantial compute or accelerator capacity, while others leave allocated resources underused. Resource contention can also affect performance even when overall infrastructure utilization appears acceptable. I need to compare resource allocation with actual workload demand, identify underused infrastructure, and understand where competing workloads create constraints.
How Q-Sensei helps
Q-Sensei makes it possible to examine resource consumption across workloads, infrastructure components, and time periods. Compare utilization by resource type, workload, infrastructure component, and time period to understand how resources are allocated and consumed.
You can identify underused capacity, compare allocated resources with observed demand, and investigate workloads competing for shared infrastructure. This gives you a clearer basis for reviewing resource allocation and identifying opportunities to improve utilization without overlooking workload requirements.
Compare allocated resources with observed demand
CPU allocations and usage · Common 0–16 cores per workload
AllocatedObserved usage
Workload A12 cores allocated · 4 cores observed
Workload B12 cores allocated · 10 cores observed
Workload C8 cores allocated · 7 cores observed
Compare the same resource and observation window, then review workload requirements before changing allocations. Sample data.
Data typically used in this investigation
Workload inventory and runtime metrics
Workload records and runtime measurements describe which applications and jobs are running and how they behave. They help connect resource consumption with individual workloads and their activity.
Compute and accelerator allocation
CPU and GPU allocation and utilization records show how processing resources are assigned and consumed. They help identify workloads with unused allocations or high resource demand.
Memory and storage metrics
Memory consumption, storage capacity, and throughput measurements describe how workloads use these resources. They help reveal underutilization, pressure, and potential contention.
Network usage
Network throughput and utilization data show how workloads consume available connectivity. They help identify heavy network demand and potential shared-resource constraints.
Configuration and scheduling data
Workload configuration, resource reservations, and scheduler records describe how resources are assigned. They help compare planned allocations with observed workload behavior.
Historical utilization records
Historical resource measurements show how consumption changes over time. They help identify recurring patterns and distinguish sustained underuse from temporary variation.
Where is additional capacity needed—and where is capacity available?
The challenge
A data center may have available compute resources but insufficient power, cooling, network bandwidth, or storage capacity where a workload needs to run. At the same time, other infrastructure may have capacity that is not fully used. I need to compare current and historical demand with available resources and understand how workload growth could affect future requirements.
How Q-Sensei helps
By connecting infrastructure inventory, utilization, and workload demand, Q-Sensei helps you assess available capacity and changing requirements. Compare available resources with workload requirements and examine utilization across compute, accelerators, memory, storage, network, power, and cooling.
Historical patterns and forecasting capabilities can help you investigate emerging constraints and changing resource requirements. You gain a more complete view of where capacity may become limited and where existing infrastructure could accommodate additional demand.
Look for headroom across all the required resources
Capacity snapshot · 0–100% of each resource’s capacity
In useAvailable headroom
Compute45% headroom
Storage30% headroom
Network15% headroom
Power8% headroom · Review constraint
Cooling5% headroom · Review constraint
Each bar shows the share of that resource’s capacity. Check power and cooling alongside available compute. Sample data.
Data typically used in this investigation
Infrastructure inventory
Inventory records describe available servers, accelerators, storage systems, and network components. They provide the baseline for comparing installed infrastructure with current resource demand.
Compute and accelerator capacity
CPU and GPU capacity and utilization measurements show available processing resources. They help identify locations where compute capacity is constrained or underused.
Memory and storage capacity
Capacity, usage, and throughput measurements describe available memory and storage resources. They help identify constraints that may limit workload growth or placement.
Network capacity
Bandwidth and utilization data describe the available capacity of network connections. They help reveal whether connectivity could constrain additional workload demand.
Workload demand and growth
Workload requirements, activity, and historical consumption show how resource demand changes. They help assess whether current capacity is likely to meet future requirements.
Power and cooling data
Power availability, consumption, and cooling measurements describe facility-level constraints. They help identify situations where IT capacity is available but facility conditions may limit its use.
Allocation and planning records
Resource reservations, allocation records, and capacity plans describe how resources are assigned and planned. They help compare expected demand with available capacity.
How can workload placement improve performance and resource utilization?
The challenge
Placing workloads across data center infrastructure involves balancing resource requirements, performance expectations, availability, and operational constraints. A workload may have sufficient compute capacity but still perform poorly because of network distance, storage access, resource contention, or limitations in power and cooling availability. I need to assess workload requirements against available resources and determine whether alternative placements could improve performance or utilization.
How Q-Sensei helps
Q-Sensei helps evaluate workload requirements in the context of infrastructure availability, utilization, and dependencies. Compare workload requirements with infrastructure conditions and investigate potential placement constraints across different locations.
AI-assisted analysis can support the evaluation of alternative placement options in the context of resource availability, workload dependencies, and operational requirements. You gain better context for assessing placement decisions and identifying options that may improve performance or resource utilization.
Compare candidate locations and their tradeoffs
Workload ANeeds 8 CPU cores, 1 GPU and access to dataset D04
Candidates · Available compute at 10:00. Mean network latency from 100 probes per location, 10:00–10:10. Compare the same workload requirements.
Weigh the available compute at each location against latency to the workload’s dependencies. Sample data.
Data typically used in this investigation
Workload requirements and configuration
Configuration records describe workload resource requirements, dependencies, and operational constraints. They help establish the conditions that potential placement locations need to meet.
Compute and accelerator availability
CPU and GPU capacity and utilization measurements describe available processing resources. They help determine whether a location can support the workload's compute requirements.
Memory and storage capacity
Memory availability and storage capacity and performance data describe the resources accessible at each location. They help identify potential constraints on workload execution.
Network topology and performance
Network topology, latency, and throughput measurements describe connectivity between infrastructure components and locations. They help assess the effect of network conditions on workload placement.
Workload dependencies
Application relationships and workload dependency records describe connections between services and infrastructure. They help identify placement requirements that extend beyond individual resource availability.
Facility conditions
Power and cooling measurements describe the operational limits of different locations. They help assess whether facility capacity may constrain a placement option.
Scheduling and placement policies
Scheduler records, orchestration data, and placement policies describe how workloads are assigned to infrastructure. They help compare current placement decisions with operational requirements.
How do operational conditions affect power, cooling, and energy efficiency?
The challenge
Power consumption and cooling requirements change with IT load, equipment utilization, and environmental conditions. An increase in energy consumption may reflect higher workload demand, inefficient resource use, changes in cooling performance, or a combination of factors. I need to understand how facility conditions relate to the infrastructure and workloads they support.
How Q-Sensei helps
Connecting facility measurements with infrastructure and workload data provides a broader view of energy consumption and operating conditions. Compare energy and thermal trends with changes in IT load, investigate correlations between operating conditions and infrastructure behavior, and identify patterns that warrant further investigation.
Multi-dimensional analysis and relationship discovery help you explore connections across operational data sources. This makes it easier to investigate potential inefficiencies, assess facility constraints, and evaluate opportunities to improve energy and cooling performance.
Align infrastructure load, power and thermal conditions
Readings · Power plots share a 0–200 kW scale; temperature uses 20–40 °C. All share 10:00–10:10.
Shared review window: 10:04–10:06
IT power load0–200 kW
Facility power0–200 kW
Rack temperature20–40 °C
10:0010:0510:10
Compare workload and cooling events over the interval where the power and temperature readings change. Sample data.
Data typically used in this investigation
Power consumption and metering
Metering data describes energy consumption across infrastructure and facility areas. It helps compare energy demand with changes in IT load and workload activity.
PDU and UPS measurements
Power distribution unit (PDU) and uninterruptible power supply (UPS) measurements describe electrical load and operating conditions. They help investigate power availability and changes in energy consumption.
Cooling system telemetry
Cooling equipment measurements describe system operation and cooling demand. They help identify changes in cooling performance that may coincide with thermal or energy trends.
Temperature and humidity sensors
Environmental readings describe conditions in racks, rooms, and other facility areas. They help investigate whether environmental conditions relate to infrastructure performance or cooling requirements.
Infrastructure and workload utilization
Server and accelerator metrics and workload activity describe the IT demand supported by the facility. They help relate energy consumption and thermal conditions to actual infrastructure use.
Facility and location data
Rack, room, and facility management records provide the physical context for infrastructure and environmental measurements. They help compare operating conditions across different areas.
Historical energy and environmental trends
Historical measurements show how energy use and facility conditions change over time. They help identify recurring patterns and investigate potential efficiency improvements.
How could infrastructure changes affect workloads and services?
The challenge
Infrastructure changes can affect more than the component being modified. A configuration update, hardware replacement, network change, or resource reallocation may influence connected workloads and services. I need to understand which dependencies could be affected, assess the operational context, and identify potential risks before or after a change.
How Q-Sensei helps
Q-Sensei connects topology, workload dependencies, configuration records, and operational history to help assess the potential impact of infrastructure changes. Compare infrastructure and workload behavior before and after a change, and investigate events or performance shifts that occur around the same time.
You gain a clearer view of the systems and workloads associated with an infrastructure change. This helps you assess potential impact, investigate unexpected behavior, and support more informed change planning and review.
Review a recorded change and its dependent services
Network N07Configuration change recorded at 10:05
Workload AUses network N07
Order APIDepends on workload A
Order API observations · 100 requests in each 10-minute window · Mean response time uses a common 0–200 ms scale
Before the change
Observation window
09:55–10:05
Mean response time
80 ms · All 100 requests
Error responses
2 of 100 requests
After the change
Observation window
10:05–10:15
Mean response time
120 ms · All 100 requests
Error responses
6 of 100 requests
Compare the surrounding evidenceChange record · Workload demand · Related incidents
The Order API averaged 80 ms before the change and 120 ms afterward. Check workload demand and other incidents during those windows. Sample data.
Data typically used in this investigation
Infrastructure inventory and topology
Inventory and topology records describe the components involved and their relationships. They help identify connected infrastructure that may be affected by a proposed or completed change.
Configuration and change records
Configuration histories and change logs describe what was modified and when. They help establish the change timeline and compare operational conditions before and after it.
Workload and service dependencies
Dependency records describe relationships between workloads, applications, and infrastructure components. They help identify services that may be affected by an infrastructure change.
Performance metrics
Infrastructure and workload measurements describe operating behavior over time. They help reveal performance changes that coincide with infrastructure modifications.
Events and incident histories
System events, alerts, and incident records describe operational activity around a change. They help investigate unexpected effects and related incidents.
Maintenance and deployment records
Maintenance schedules and deployment histories describe planned infrastructure work. They help provide context when assessing operational changes and their potential impact.
Which infrastructure components need maintenance attention?
The challenge
Data center equipment operates under changing workloads and environmental conditions. Gradual performance degradation, recurring errors, rising temperatures, or increasing resource instability may indicate that a component needs closer inspection. I need to identify which infrastructure components show signs of deterioration and prioritize maintenance based on their condition and operational importance.
How Q-Sensei helps
By examining equipment health indicators, performance history, and environmental measurements, Q-Sensei helps identify components that warrant closer inspection. Compare current behavior with historical patterns and examine whether recurring errors, changing performance, or unusual operating conditions affect specific equipment.
Trend analysis and anomaly detection can help highlight changes that warrant maintenance investigation. You get a more informed basis for prioritizing inspections and evaluating equipment condition in the context of its workload dependencies and operational role.
Storage diagnostics · Same equipment model · Event counts per 10-minute window, 09:30–10:00 · Common 0–10 event scale
Storage S04
Recorded events: 2 / 5 / 9
Dependent workload or service
Order API · High operational importance
Service history
Last inspection: 6 months ago
Prioritize inspection
Storage S05
Recorded events: 1 / 1 / 2
Dependent workload or service
Archive jobs · Flexible completion window
Service history
Last inspection: 1 month ago
Compare with equipment history
S04 has more diagnostic events, supports the Order API and was last inspected six months ago. Give it priority for inspection. Sample data.
Data typically used in this investigation
Infrastructure inventory and lifecycle records
Asset records describe equipment types, locations, and available lifecycle information. They help identify the components under review and provide context for maintenance planning.
Hardware health and diagnostic data
Device health indicators, diagnostic events, and error records describe equipment operating conditions. They help identify recurring faults or changes that may warrant inspection.
Performance and utilization history
Historical CPU, GPU, memory, storage, and network measurements describe how components behave over time. They help reveal sustained degradation or unusual operating patterns.
Environmental telemetry
Temperature, humidity, and other environmental measurements describe the conditions around equipment. They help investigate whether environmental factors coincide with changes in component behavior.
Event and incident histories
Alerts, error logs, and incident records describe previous operational problems. They help identify recurring issues and understand the operational history of a component.
Maintenance and service records
Maintenance logs and service histories describe previous inspections and interventions. They help place current health indicators in the context of past maintenance activity.
How are infrastructure issues affecting service availability?
The challenge
An infrastructure incident does not always affect every workload or service in the same way. A component failure or performance issue may disrupt one application, degrade several dependent services, or have little visible effect on users. I need to understand the relationship between infrastructure incidents, workload dependencies, and service availability so I can assess operational impact and prioritize investigation.
How Q-Sensei helps
Q-Sensei lets you investigate infrastructure events alongside workload dependencies, application performance, and service-level measurements. Trace relationships between affected components and workloads, compare incident timelines with service behavior, and investigate how infrastructure conditions coincide with availability or performance changes.
You gain a clearer picture of which workloads and services are affected and how infrastructure issues relate to operational objectives. This supports incident assessment, impact analysis, and more informed prioritization of troubleshooting efforts.
Compare availability across dependent services
Storage S04 · I/O warning at 10:02Recorded dependency shared by both services
Service checks · 100 checks per service, 10:00–10:10 · Successful checks use the same 0–100% scale
Order API
Recorded dependency
Storage S04 via server N12
Successful availability checks
92 of 100 · 92%
Failed checks
8 of 100
Investigate the eight failed checks
Reporting API
Recorded dependency
Storage S04 via server N18
Successful availability checks
100 of 100 · 100%
Failed checks
0 of 100
No failed checks in this sample
During the storage warning, eight Order API checks failed. All 100 Reporting API checks succeeded. Sample data.
Data typically used in this investigation
Infrastructure events and alerts
Event and alert histories describe component failures, warnings, and operational changes. They help establish incident timelines and identify infrastructure conditions associated with service disruption.
Infrastructure topology and dependencies
Topology records describe relationships between infrastructure components. They help trace how an issue in one part of the environment may affect connected workloads and services.
Workload and application telemetry
Workload activity, application logs, traces, and performance measurements describe service behavior. They help connect infrastructure incidents with application-level effects.
Availability and service-level measurements
Availability records, service-level indicators, and operational targets describe service reliability. They help assess whether infrastructure incidents coincide with changes in service performance or availability.
Incident and change records
Incident histories and configuration or change records describe previous disruptions and related operational activity. They help identify recurring patterns and provide context for impact assessment.
Q-Sensei connects infrastructure, workload, and facility data to help you improve visibility, performance, reliability, and efficiency across your data center.
