Skip to main content

Q1 2026 Key Features and Improvements

Info

Applicable versions: Service 2.25.X ~ 2.27.X

WhaTap expanded its service in Q1 2026 across LLM, database, and infrastructure. Beyond simple feature additions, analysis capabilities have been strengthened to examine systems based on end-to-end flow rather than individual metrics.

LLM Observability Launch​

LLM Observability collects request flows from LLM-based applications as transactions and provides integrated analysis of performance, cost, token usage, and errors. It enables analysis of LLM request flows and response quality issues that were difficult to identify with traditional APM.

  • Visualize the entire call flow from user request to response generation as a transaction
  • Analyze response time, call count, cost, and token usage
  • Trace problematic requests based on prompt logs and response data
  • Analyze cost structure and usage patterns based on per-model token consumption
  • Correlate GPU metrics with LLM performance degradation and infrastructure impact

Even when system metrics show no issues but response quality degrades, you can identify root causes by analyzing LLM request flows alongside GPU usage patterns. Additionally, even when the response code is normal (200 OK), you can analyze cases where response quality is inadequate from the user's perspective — providing differentiated analysis beyond traditional APM.

Currently supports Python and Java, with plans to expand to Node and other languages, including OpenTelemetry environments.

Note

Reference

Application AI Analysis Enhancements​

AI analysis capabilities in the Application (APM) area have been expanded with Active Stack Analysis (Beta) added after thread analysis. Previously, performance issues required manual comparison of logs, transactions, and metrics — now AI automatically correlates spike causes with the relevant transactions.

  • AI automatically analyzes transactions at the time of performance anomalies
  • Identifies bottlenecks based on stack dumps and classifies severity (High/Medium/Low)
  • Automatically detects major performance issues such as deadlocks, lock contention, and I/O waits
  • Suggests improvement directions based on priority

With AI-based analysis from high-level execution flow down to detailed execution units, users without deep stack analysis experience can quickly identify root causes and determine next steps.

You can review potentially problematic prompts or abnormal token requests through prompt logs, and navigate directly from those prompts to the transaction detail view to continue analysis.

Note

Reference

Database Monitoring Improvements​

Multi-Instance and Visualization Enhancements​

Previously limited to single-instance monitoring, you can now view data in a variety of ways across multi-instance environments.

  • Integrated monitoring of multiple instances on a single screen with various chart types
  • Flexible dashboard composition using diverse widgets
  • Intuitive visualization support including CPU and memory gauge charts

SQL Analysis Consolidated Around Top SQL​

Previously centered on individual SQL statistics, analysis is now consolidated around Top SQL, enabling faster identification of queries affecting performance.

  • Integrated view of execution time, trends, and wait information
  • Compare change history with execution plans and plan history
  • Reproduce problem scenarios with bind value-based SQL reproduction

AI Tuning Guide Improvements​

AI suggests performance improvement directions based on SQL analysis results. Previously, the same request could yield different results — now analysis results are saved to maintain consistent output, with on-demand re-analysis available for the latest results.

Note

Reference

Infrastructure: Centralized Operations with OPM​

OPM (Operation Manager) has been added, enabling centralized execution of server tasks and collection of results. OPM is provided as an on-premises product.

  • Deploy scripts and collect execution results via server agents
  • Manage server-level tasks centrally without individual server access

Summary​

WhaTap has expanded capabilities across LLM, application, database, and infrastructure in this update.

  • Extended analysis coverage to AI service request flows and response quality via LLM Observability
  • Expanded AI-based analysis in APM from threads to active stacks
  • Reorganized database monitoring around multi-instance support and Top SQL, with improved AI analysis consistency
  • Extended infrastructure operations with OPM for centralized server task execution and management

WhaTap continues to evolve beyond metric-centric monitoring toward analyzing systems based on end-to-end flow.

Last updated on