See your entire estate. Know what needs attention.
Ekalix Radar is unified monitoring for SQL Server, Windows, Applications and MongoDB - one command centre instead of four separate tools, reading what's already there rather than asking you to install anything new.
Four estates, one screen.
A database that stops answering, a host running low on disk, an application pool stuck restarting, a replica set falling behind - Ekalix Radar watches all four the same way, in the same place.
SQL Server
Backups, jobs, blocking, waits, Availability Groups, capacity and two years of history.
Windows Server
CPU, memory, disk, services, scheduled tasks, patching, certificates and the event log.
Applications & IIS
Sites, app pools, worker processes, error rates and certificate expiry.
MongoDB
Replica set health, replication lag, the oplog window and current operations.
What each estate actually gives you.
Chosen from what Ekalix Radar genuinely does today - not a wish list.
- Availability, every ~60 seconds. A background check that backs off to every 5 minutes for a server that stays down, so a real outage doesn't get lost in noise.
- Backups. Last full/diff/log and their age, aware of Availability Groups - a database backed up on another replica isn't wrongly flagged.
- SQL Agent jobs. Failures, and jobs still running long past their own history - watched in real time, not just at the next collection.
- Blocking, to the head of the chain. Who's actually blocking, how many sessions are waiting, resolved to the SQL Agent job when that's the cause.
- Wait statistics. The waits that matter, with the idle/background ones already filtered out.
- Availability Groups. Replica role, synchronization health, and log send/redo queues, watched in real time.
- Capacity. A statistical forecast (not a straight-line guess) of when a drive will fill, with a confidence figure and a note when the trend genuinely changes.
- Two years of history. Detail for two days, 15-minute resolution for three months, hourly for two years - so a slow Tuesday can be compared with last quarter.
- Deadlocks and execution plans, captured and drawn as pictures - a deadlock's sessions and locks, a plan's operators - not raw XML.
- 54 read-only checks on demand, from backup chains to missing indexes to tempdb pressure.
A blocking chain resolved to its SQL Agent job, in the real command centre. Video coming soon.
Two things it doesn't do, so this stays accurate: a single slow ad-hoc query has no automatic alert of its own (current activity shows it live; jobs are what get watched for overrun), and there's no separate "failover happened" log - a failover shows up as the replica and database alerts above changing state.
- Availability, CPU and memory read from the host itself - a separate signal from SQL Server's own CPU figure, so you can tell "the box is busy" from "SQL Server is busy."
- Disk capacity, with the same statistical forecast used for SQL Server's data and log drives.
- Services - a stopped automatic service is flagged, and marked critical on the services a database host depends on.
- Scheduled tasks, with the result code decoded into plain English.
- Patching - days since the last successful install, and whether a reboot is still pending. (This is age-since-patch, not a count of missing security updates.)
- Certificates, cross-checked against what SQL Server is actually using, including whether the certificate's name really covers this host.
- The event log, grouped and de-duplicated rather than left as a raw feed.
A real cross-estate link: open a SQL Server and there's a "Host" tab - the Windows snapshot for that same machine, inline. Video coming soon.
- Sites and application pools - state, recycle counts, and a stopped auto-start site or a crash-looping pool raised as a finding.
- Worker process (w3wp) CPU and memory, measured per process, not guessed from the total.
- Request rate, per site and per worker process.
- HTTP error rate, two ways: tailing the site's own IIS log for the last 15 minutes, and a counter-based 500-rate that doesn't depend on logging being on.
- Certificate expiry for every HTTPS binding, including whether the certificate actually covers the site's host name.
- Also watched: the IIS/WAS services, a synthetic local health check, the .NET runtime versions installed, and disk space on the host.
Called plainly: host, process and log-level monitoring - not code-level or distributed tracing. We say "Application Monitoring," not APM. Video coming soon.
- Replica set health - each member's role (primary, secondary, arbiter), and the set's overall state.
- Replication lag and the oplog window - how much replay time you actually have before a secondary could fall irrecoverably behind.
- Current operations, storage and database statistics, and connection counts.
- Slow operations, by shape, not by value. A slow query is recorded as its field names and structure - never the literal values inside it.
- SCRAM authentication and TLS, read the way the driver would, with no separate agent on the replica set.
Built, not bolted on - the same command centre, alert format and history as SQL Server, Windows and IIS. Video coming soon.
Built to be trusted with production.
Not a rating - an architecture.
We won't name other products here. This is the shape of the decision, however you compare it.
Explanations you can trust, without handing over your estate.
Every alert already gets a plain-English explanation with no AI involved. An AI model is an optional upgrade to that explanation - never a requirement.
What it does
- By default: a built-in rule engine. Every alert already carries a what/why/next-step explanation, written by rules, with no network call and no AI at all. This is what most installations use.
- Optional: a fuller explanation from an AI model. Turn on OpenAI, Google Gemini, Anthropic Claude, or a self-hosted Ollama model, and "Explain" asks that model to write the same three-part answer with more nuance.
- Off until an administrator turns it on. New installs ship with no provider selected. If a call fails or times out, Ekalix falls back to the rule engine automatically - you're never left without an answer.
- Ollama runs on your own infrastructure. Nothing leaves your network with that option, and no API key is needed.
What actually gets sent, if you turn it on
The slice is built to be small: for most explanations, short excerpts - an alert's own text, part of a failed job's error message, a line from the error log. Never a whole database, and never connection strings, passwords or the AI provider's own key. One honest exception: asking Ekalix to "explain this query" sends up to about 200 characters of the cached SQL text itself, which could include a literal value if your application doesn't use parameterized queries - worth knowing before you use that specific button with a cloud provider. Every AI explanation is recorded in the audit log, so this is never silent. What each provider does with what it receives after that is governed by your agreement with that provider, not by Ekalix.
What we're building next.
Plans change as customers teach us. This is the honest state of it, not a promise of dates.
Four estates, one product
- SQL Server, Windows, IIS and MongoDB
- Command centre, live views and two years of history
- Alerts to Teams, Slack and email
- Optional AI explanations - OpenAI, Gemini, Claude or self-hosted Ollama
A fifth estate, deeper AI
- MySQL, alongside SQL Server
- More of what an AI model can be asked, and more of the estate it's allowed to reason over
- Online trials and payment
Running in an afternoon.
Copy the folder
Put Ekalix Radar on any Windows Server in your network. Nothing is installed on the servers it watches.
Give it a database
It keeps settings and history in your own SQL Server - 2016 SP1 or later; Express is fine.
Add your servers
SQL Server, Windows hosts, IIS applications and MongoDB replica sets, added with read-only credentials.
See what needs attention
The command centre groups every alert by estate and environment, worst first.
Frequently asked.
Is this APM?
No. The Applications estate is host, process and IIS-log monitoring - real and detailed, but not code-level or distributed tracing.
Do I need an AI key to use this?
No. Every alert has a plain-English explanation from the built-in rule engine, with no network access, before you ever look at AI.
Where does my data live?
In one SQL Server database you name and control, inside your own network.
How is it licensed?
By the number of servers you monitor, with a free trial first. See pricing.
See what needs attention before your users tell you.
A 20-minute demo on a live estate, or a trial on your own servers.