Skip to content

R8. Glossary

Diátaxis: Reference (a dictionary to look things up in) · Audience: operators / administrators ← Back to contents

Short definitions of the terms that come up often in this manual. For the concepts in more detail see E. Understanding the Concepts, and for how to judge a metric see E3. What the Metrics Mean.


Terms (Alphabetical)

TermMeaning
AgentThe collector attached to an instance that gathers performance data and sends it to the APM server.
Application groupThe service-level unit bundling the instances that are meaningful to look at together. → E2.1
AvailabilityThe proportion of time the service responded normally (99.9%, for example). The core SLA metric.
Context selectorThe tool at the top of the main area for narrowing the target by group, host, or instance. → R4
Datasource (connection pool)The bundle of connections the application borrows database connections from. When exhausted, requests wait. → H8
DeadlockThe state where two or more threads wait on each other's locks and are stuck for good.
Error rateThe proportion of all requests that returned a 4xx or 5xx error (%).
Garbage collection (GC)The JVM's work of clearing unused objects from the heap. A brief pause (STW) occurs while it cleans. → E3.2
HeapThe JVM memory space where Java objects accumulate.
InstanceOne process actually running (one WAS, one database broker, and so on).
MergeThe selector option deciding whether to combine several targets into one line or split them out individually. → H1
Metaspace (non-heap)The area outside the heap where class metadata accumulates. A leak from repeated redeployment causes an OOM.
OutlierA value statistically far from the usual pattern. Repeated, it is a precursor to passing a threshold.
Response timeThe time taken to handle one request (ms).
Sawtooth patternThe normal shape where heap utilization rises and falls after GC, repeatedly. A flat line suggests a leak. → H7
Thread dumpA snapshot of what every thread was doing at that moment. Used for deadlock and hang analysis.
ThresholdThe level above which an event is raised. Two levels, WARN and CRITICAL. → H15
TransactionThe handling flow of one user request (request → processing → response). → E2.2
WaterfallThe screen laying out the segments of one transaction (database, external calls, and so on) as bars in time order. → T2
WorkerThe unit in which a web server (load balancer) handles requests. When exhausted, connections wait or are refused.

Abbreviations

TermMeaning
APMApplication Performance Management/Monitoring. → E1
APDEXApplication Performance Index. The score converting response time into user satisfaction (0 to 100). → E3.1
CASThe response handling process of a CUBRID broker. When saturated, database requests wait or fail.
CogentAIThe AI feature that analyses charts and events to suggest a likely cause and an action. → H14
Full GCThe GC that cleans the whole heap. Its STW pause is large.
OOM (OutOfMemoryError)The serious error where the JVM can no longer allocate memory because the heap is short.
SLAService Level Agreement. The promised service level (usually expressed as availability). → E3.3
STW (stop-the-world)The interval where the application pauses briefly, for GC or similar.
TPSTransactions Per Second. The number of requests handled in one second (throughput). → E3.1
T-MapThe heatmap scattering transactions by response time and time of day. Problems are read from the distribution. → H5