Home › SysPeep Monitor › Why SysPeep › Deployment
05 · Deployment

One executable on a server you own


A monitoring tool is the one piece of software that has to keep working while everything around it is on fire, and the one whose data you would least like to find in somebody else's hands.


Key points

  • On-premises software: one MSI, one Windows service, on a server you already own.
  • Console, collectors, event receivers and the time-series store in a single service.
  • Configuration and history in your choice of SQLite, MySQL or SQL Server, with retention you set and device credentials sealed under a per-install key.
  • Remote probes over mutual TLS that report what they run and what they reach; a sensor checked from up to four places; Windows Failover Cluster in the High Availability edition.
  • A console that works in any browser, with or without scripting, reviewed against WCAG 2.2 AA.

On this page: One executable Installing Upgrades and upkeep Probes and clustering Any browser Administration you can prove


SysPeep's deployment model is built around both of those: it runs as one Windows service on a server you already own, with the console, the collectors, the event receivers and the storage engine in a single service, and your monitoring data in a database you administer. This page covers how it installs, how it scales, how the console behaves in whatever browser you have, and what it does for the people who have to sign the security questionnaire.


One service


The entire console — templates, scripts, stylesheets, fonts — is compiled into the service along with the collectors, the event receivers and the embedded time-series store. There is no separate web server to configure, no runtime to keep patched beside it, and no console page, script or font sitting outside the binary to be edited or tampered with.


Monitoring data and configuration stay on your server, in your database, under your retention policy. There is no hosted platform between you and your own status page.


Install in minutes on a server you already have


  1. Run the MSI. It installs one Windows service: console, collectors, event receivers and the embedded time-series store in one process. There is nothing else to license, patch or babysit.
  2. Choose a database, or do not. Start on SQLite with zero database administration; move to MySQL or SQL Server when policy requires it. The schema is kept in verified parity across all three, with Unicode text in every one, so device names and log lines in any language are stored exactly as written.
  3. Let Quick Start open the firewall and start the service. Its last step lists every port the service will listen on, reads Windows Defender Firewall for rules that name SysPeep, adds the missing inbound rules and starts the service, so a fresh install is reachable from the network on the first try. A port changed later offers the rule again.
  4. Run the first discovery from the dashboard. Automatic sweeps wait for that first run, with live progress, so the install never fills with devices before you have chosen what to scan. From then on discovery runs as a service, whether or not anyone is logged in.

Retention is set rather than guessed: metric history and events share one horizon, 90 days by default; flow history has its own setting (7 days by default), the hourly flow totals theirs (90 days), and a storage ceiling (2048 MB by default) removes the oldest whole days of flow first, rolling each day into hourly totals before it goes. Disk growth is bounded by settings you chose, and a missing or mistyped value means 90 days, never "keep forever".


Upgrades and upkeep that do not need babysitting


  • Upgrades do not hang. The installer stops the old service itself and, if an older release is stuck in Stopping, ends that release's own process; the new service bounds its own stop at 25 seconds and names the step that hung. If the service was running before the upgrade, the installer starts it again.
  • The control panel says why the service stopped — missing configuration, an unconfirmed or unreachable database, an unusable HTTPS certificate or a stop that timed out — each with a pointer to the log. A slow first start is shown as starting, not as a failure.
  • A database restart does no damage. After a lost connection a repeat-safe statement is retried once and never an insert; multi-step changes are all-or-nothing; and if the inputs that judge sensors cannot be read for three cycles, one critical incident says monitoring is holding instead of quietly keeping old statuses.
  • Help is next to the setting. F1 in the control panel, and Help in the Licensing, HTTPS Certificate and Directory sign-in dialogs, open that topic of the manual.

Remote probes for branches, DMZs and OT segments


A probe polls a branch or a segmented network from the inside and reports back to the central service over mutual TLS, so nothing inbound is opened on the branch firewall. One console download carries the probe installer with its configuration; the probe retries until its one-day enrollment secret expires, picks up new configuration without a restart, and a refused enrollment or a failed handshake is named on both machines rather than failing silently. Each probe is itself a monitored device with silence, status, CPU, memory and spool sensors: a dead probe is an incident, not a mystery of missing data.


A probe is also a place to see devices from, and the console says what each place saw, reached and polls. The Probes card lists the collection methods each probe runs — 21 of the 38, from SNMP and WMI to Modbus, OPC UA, LDAP, RADIUS and MQTT — and sensors a probe cannot run are marked on its devices, with how to collect them from the server instead. A "seen by" chip in the tree and a Reach block on the device page show which place reached a device and what answered, so choosing the probe that polls it is a decision made on a real try rather than a guess, and a whole branch can be assigned to a new probe in one step.


A sensor can also be polled from up to four places — this server or any probe — each with its own history, thresholds and status, and the device still counts once toward the license. The same check from more than one place shows whether a failure is the device or one site's path, and with each place set as an availability source, the device goes down only when every place fails. Every alert names the place that raised it.


For the monitor's own survival, the High Availability edition runs as a native resource on Windows Failover Cluster — infrastructure the organization already owns and already knows how to operate. It is a competitive capability rather than a unique one, and we mention it here rather than on the home page for exactly that reason.

Probes

  • Report back over mutual TLS; no inbound firewall rule
  • Collect flows at the branch
  • Self-update, and pick up configuration without a restart
  • One console download, with the configuration inside
  • 21 collection methods at the branch
  • Say which devices they saw and reached
  • A sensor checked from up to four places; the device counts once
  • Monitored devices in their own right
  • In the paid editions

Works in the browser you have


The console is built the way web pages used to be: rendered on the server, readable and operable without JavaScript. Status, filters, search and alert acknowledgement all work from a locked-down jump host, a kiosk browser or the old crash-cart laptop that lives in the server room. Scripting is layered on top for the modern experience, with feature-detected polyfills so older engines are not left behind, and WOFF fonts for the ones that cannot render anything newer.


When something is on fire, the console does not care which browser you reached for.

Its keyboard operation, screen-reader announcements, named controls, focus management and contrast were reviewed against WCAG 2.2 AA in an internal review. That supports accessibility procurement reviews and, more importantly, works for operators who rely on assistive technology. A text-size control in the page header offers normal, large and extra-large, remembered per browser; the layout reflows to a 320-pixel window with no sideways scroll, and Windows high-contrast and reduced-motion settings are honoured.


Administration you can prove


Every configuration change is written to an audit log with the administrator's identity, so compliance evidence is generated as a by-product of doing the job rather than as a screenshot marathon the week before the audit.


The questionnaire asksSysPeep's answer
Where is the monitoring data stored?On your server: configuration and history in your choice of SQLite, MySQL or SQL Server, time-series data in SysPeep's embedded store, with retention you set and device credentials encrypted under a key that stays in the data folder.
Is sign-in integrated with the directory?Yes: local or Active Directory / LDAP. The directory answers the password only; accounts exist locally and the local password is tried first, so a domain-controller outage never locks the administrator out.
Is access role-based?Three administrator tiers — junior, mid, senior — enforced at every endpoint that changes anything, mapped per account or by LDAP group. The helpdesk can acknowledge and investigate without being able to reconfigure, delete or clear.
Are configuration changes logged?Every change, with who made it, in the audit log.
How are integrations authenticated?API tokens with read or write scope, shown exactly once at creation and revocable individually.
Is the monitoring path encrypted?TLS 1.2 or later on the console, with the certificate the organisation already issued — PEM, DER, PKCS#7, PKCS#12/PFX, PVK, or one from the Windows certificate store by thumbprint — and a Check certificate button that reports expiry, key strength and missing intermediates before anything is saved. Mutual TLS between the service and its probes; session limits.
Where are the device passwords?Device credentials and the passwords in the configuration file are encrypted with AES-256-GCM under a key kept in the installation's own data folder, restricted to SYSTEM, Administrators and the service account, never compiled in and never in the configuration file. A copied configuration file or a database backup without the key file does not hand over the network's passwords.
What happens when the database fails mid-action?It fails closed: a failed permission read refuses the sign-in, a failed device count refuses monitoring beyond the license, and a failed read refuses a credential save instead of erasing the stored password.
What is left after a crash?A minidump and descriptor per crash in a locked-down folder, as evidence for support. Reports go to SysPeep's crash-report service.

Secrets are shown once. API tokens and probe credentials are displayed at creation and never again. If you did not write it down, you generate a new one. This is mildly inconvenient on purpose.


One binary you can hash, one service you can restart, one database you administer, one audit log you can hand over, credentials that are useless without the key on your server, and monitoring data that stays on hardware you control. It is not the most exciting list of features on the site, and for the people who need it, it is the one that decides the purchase.


Install it on a server you already have

Download the MSI, run it on a Windows server you already own, and run the full 30-day trial on the segment that matters. Up to 5 devices stay free at full depth afterwards.

Download Free TrialNext: Agentless and extensible