A database that is "up" can still be the reason your app is slow. CPU looks fine, disk looks fine, and p95 latency has tripled because a query plan flipped overnight. General server monitoring never sees that. Database monitoring tools exist to read the engine itself: which queries run, how long they wait, what they lock, and whether the replica is falling behind.
This post is about picking one. It covers what separates the tools from each other, four worth evaluating with pricing read from each vendor's site on 2026-10-17, where each one is weak, and how to choose by situation rather than by feature count.
What Separates One Tool From Another#
Most tools in this category show the same top-level charts: connections, queries per second, cache hit ratio, replication lag. If that were the whole job, the choice would not matter. The differences are underneath.
How deep into the query they go. Some tools stop at "this query is slow". Others capture the execution plan, show which index it skipped, and tell you what index would fix it. That second kind turns monitoring into a tuning tool, which is where most of the value is for a team with nobody whose whole job is the database.
Which engines they cover, and how evenly. A tool built for PostgreSQL will read pg_stat_statements and auto_explain natively. The same vendor's MySQL support may be a thin wrapper. If you run two engines, check that both get the deep treatment, not one.
Where the data lives. Hosted SaaS means the vendor stores your query text and parameters, which some compliance teams refuse outright. Self-hosted tools avoid that at the cost of running one more service.
What the price is tied to. Per database server, per host, per monitor, or per node. The unit decides whether adding a read replica doubles your bill or barely moves it.
Four Tools Worth Evaluating#
| Tool | Engines | Deployment | Pricing (read 2026-10-17) |
|---|---|---|---|
| pganalyze | PostgreSQL only | SaaS, or on-premise on the Enterprise plan | Production $149/mo for 1 server; Scale $399/mo for 4 servers, then $100/mo per extra server; Enterprise on request |
| Percona Monitoring and Management | MySQL, PostgreSQL, MongoDB | Self-hosted | Open source, no license fee |
| Site24x7 | Database monitoring is one plugin inside a general infrastructure platform | SaaS | All-in-one Lite $9/mo paid annually (2 servers, 5 websites); Professional $42/mo paid annually (5 servers, 20 websites, 4 GB logs); Enterprise from $625/mo |
| SolarWinds Database Performance Analyzer | SQL Server, Oracle, MySQL, MariaDB, PostgreSQL, Db2, SAP ASE, Azure SQL | On-premise or cloud | Quote only; the pricing page is a quote-request form |
pganalyze#
pganalyze does one engine and goes deep on it. It collects statistics and logs from PostgreSQL, visualizes EXPLAIN plans, and runs an Index Advisor that recommends specific indexes from real query workloads. The Scale plan adds a Query Advisor, a VACUUM Advisor, and log insights with PII filtering.
Pricing is per database server. The Production plan is $149 per month for a single server with 14 days of history. Scale is $399 per month for 4 billable servers with 35 days of history, and each additional server is $100 per month. A replica counts as half a billable server, so a primary with 2 replicas is 2 billable servers (pganalyze.com/pricing, read 2026-10-17).
Where it falls short: if you are not on PostgreSQL, it does nothing for you. And at $149 a month for one server it is one of the pricier entry points here, so the index advice has to be worth real engineering time to pay off.
Percona Monitoring and Management#
PMM is open source software you run yourself. Percona describes it as a single pane of glass across MySQL, MongoDB, and PostgreSQL (percona.com, read 2026-10-17), and its query analytics view ranks queries by how much load they put on the engine, so the thing to fix first is at the top of the list.
There is no license fee. The cost is operational: a PMM server to host, a client to install beside each database, and upgrades to track. Percona sells support and managed services around it, but the software itself is free.
Where it falls short: you own the uptime of your monitoring. A PMM server that runs out of disk stops collecting, and nobody at a vendor is going to page you about it. The dashboards are built on Grafana and there are a lot of them, which also means they take time to learn.
Site24x7#
Site24x7 is a general monitoring platform where database monitoring is one plugin among servers, websites, APIs, and logs. The pricing page lists database monitoring as available on both the Lite and Professional infrastructure plans. The All-in-one Lite plan is $9 per month paid annually and covers 2 servers and 5 websites; Professional is $42 per month paid annually for 5 servers, 20 websites, and 4 GB of log ingestion, and Enterprise starts at $625 per month (site24x7.com/pricing, read 2026-10-17).
Where it falls short: the database view is a tier shallower than a dedicated tool. You get engine metrics and alerting, not plan analysis or index recommendations. It fits a team that wants one bill for host, database, and website checks and is happy to do query tuning in the database's own tools.
SolarWinds Database Performance Analyzer#
DPA is the broad coverage option. Its product page lists SQL Server, Oracle, MySQL, MariaDB, PostgreSQL, Db2, SAP ASE, and Azure SQL (solarwinds.com, read 2026-10-17), all from one product, with wait time analysis as its core method. Instead of asking "which query is slow", it asks "what are sessions waiting on", which points at locks, I/O, and contention a latency chart would hide.
Pricing is not published. The product's pricing page is a quote-request form (solarwinds.com, read 2026-10-17), so budget for a sales conversation before you can compare it on cost.
Where it falls short: you cannot put it on a cost comparison without talking to sales first, and it is built for shops with a DBA function. A team of 5 with one Postgres instance will find most of it unused.
Choosing by Situation#
One PostgreSQL database and no DBA. pganalyze, because the index and query advisors do work a DBA would otherwise do. Check the per-server cost against how many servers and replicas you actually have.
Several engines and someone who can run infrastructure. Percona PMM. It is the only option here that covers MySQL, PostgreSQL, and MongoDB at full depth without a license fee. Treat the PMM server as a production service with its own disk alerts.
Small team that wants database, server, and uptime checks in one place. Site24x7. Accept that query tuning stays in EXPLAIN and the slow query log.
Mixed enterprise estate with Oracle or SQL Server in it. DPA, and start the quote early.
Whichever you pick, decide first which database metrics you will alert on. A tool with 200 charts and no alert on connection pool saturation will still miss the outage that takes everything down at once.
What None of Them Tell You#
Every tool above reads the engine from the inside. That is the right place to diagnose a slow query. It is the wrong place to find out that users cannot reach the application at all.
Replication can be healthy, the pool half empty, and every query fast, while a bad deploy has the API returning 500s before it ever touches the database. Internal metrics will show green. The only check that catches it is a scheduled request against a real endpoint from outside your network, asserting on status and response time. Database monitoring covers this split in more detail.
Conclusion#
Database monitoring tools separate on query depth, engine coverage, and pricing unit, not on whether they can draw a connections chart. pganalyze is the deepest single-engine option, PMM is the free multi-engine one, Site24x7 is the cheapest way to get database alerts next to everything else, and DPA is the enterprise pick with a sales call attached.
ObserveOne does the outside-in half: scheduled API and uptime checks from multiple regions, so a healthy database and a broken application never read as the same green dashboard.