Oninit® DBVoyeur
Oninit® DBVoyeur is a best-of-breed SQL query-level performance monitor for the world's leading relational and document databases. It captures every SQL transaction directly off the wire and stores the resulting records in a purpose-built repository designed to cope with very large volumes of performance data.
DBVoyeur is built around three principles: it is non-intrusive, adding no measurable load to production database servers; it measures performance from the end-user perspective, not the operating-system perspective; and it speaks the same operational language across every database it supports, so a single console covers Informix, Db2, SQL Server, PostgreSQL, MongoDB, SAP ASE, SAP IQ, and HCL OneDB.
Non-intrusive operation
DBVoyeur is based on a passive network capture model. It does not require configuration changes to the database, does not connect to the database as a client, and does not depend on server-side traces, event sessions, extended events, audit logs, or any other instrumented path. There is nothing inside the database engine to enable, tune, or turn off.
DBVoyeur is installed directly on the database host, and overhead is measured at about 4% of one CPU core for the capture agent at 40 statements a second on our test system (it grows with load). Capture is local to that host by design: statement text from encrypted connections, and shared-memory clients, are only observable there, so SPAN, mirror-port and network-tap deployments are not supported.
This is a stark contrast with traditional database-level performance tools, which typically consume 10–15% CPU on the database host and which often consume materially more when operators ask them for detail.
Measured from the end-user perspective
Traditional OS-level and DB-level performance tools deliver IT-centric metrics — disk I/O, buffer hit ratios, CPU utilisation — that have little obvious correlation to end-user experience. DBVoyeur measures every SQL transaction directly from the perspective of the application that issued it, removing guesswork and allowing scarce performance-tuning effort to be focused precisely where it will yield the most dramatic improvements in user-perceived response time.
This is a decisive advantage in complex, distributed environments, where critical time is otherwise wasted simply determining which area of the IT organisation should own a given issue.
At a glance
| Topic | Detail |
|---|---|
| What it does | Captures and times every SQL transaction at wire level; stores the result in a searchable repository. |
| Where it sits | Passive capture, on the database host. |
| What it sees | Every prepare, execute, fetch, commit and rollback, with full SQL text, bind values, row counts and response times. |
| Coverage | Informix, IBM Db2 LUW, Db2 z/OS DDF, HCL OneDB, SAP ASE, SAP IQ, Microsoft SQL Server, PostgreSQL, MongoDB. |
| Overhead (DB host install) | About 4% of one CPU core for the capture agent at 40 statements a second (measured on a test system; grows with load). |
| Repository | Purpose-built for very large volumes of per-transaction performance data; SQL-searchable. |
| Multi-DB licensing | One DBVoyeur instance can be licensed to monitor multiple database types and multiple instances. |
| Free edition | One database of one engine, 24 hours of retention. See Download. |
| Read-only | DBVoyeur never modifies traffic in either direction. It can only observe. |
Why use it
Operating-system tools show you a busy host. Database-level tools show you a busy database. Neither tells you which application’s users are waiting, and why. DBVoyeur shows you the SQL the user actually issued, how long the database actually took to answer it, and how that time was spent — for every transaction, for every database, in one place.
The result: faster triage, less finger-pointing between application and database teams, and a much shorter path from a user complaint to the line of SQL that caused it.