Home › SysPeep Monitor › Why SysPeep › Coverage
04 · Coverage

From the switch closet to the plant floor to the cloud account


"One place to see everything" is the wish behind most monitoring purchases, and it is usually answered with three products, two of which are add-on packs for the first.


Key points

  • 38 collection methods and 280 catalog sensors from one Windows service, under one alert model and one on-call process.
  • A compiled MIB set names the counters of the common network vendors; a WMI explorer turns any Windows property into a sensor.
  • A Flows page in the main navigation: exporter status, top talkers, conversations, interfaces, QoS classes and traffic by hour of the week.
  • Discovery runs continuously once you have run the first sweep; the map comes from LLDP/CDP on the wire.
  • Network, Windows and Linux servers, hypervisors and containers, applications and databases, cloud accounts, plant-floor protocols, hygiene checks and AI endpoints.
  • Flow analysis and directory sign-in in every edition; scheduled reports and PDF export in the paid editions, not as separate packs.

On this page: Discovery The map Network Flow analysis Servers and hypervisors Applications Cloud and plant floor Security hygiene Composite sensors How collection runs


SysPeep covers what most teams currently spread across three tools from a single Windows service. This page is the inventory: each section is a shipped collection method or sensor family, with a note on what it means to the person on call. Five of these areas have a page of their own: SNMP monitoring, the NetFlow analyzer, bandwidth monitoring, discovery and mapping and Windows server monitoring.


Discovery that runs whether or not you are logged in


On a new install, discovery waits for you to run the first sweep from the dashboard, with live progress, so a fresh installation never fills with devices before anyone has chosen what to scan. From then on, active discovery scans and passive LLDP/CDP capture run continuously as a service. New devices appear in a "Discovered" filter with a nudge on the dashboard, so onboarding is a click rather than an audit.


Smart groups with membership rules — subnet ranges, device types, discovering probe, manual pins, subgroups with first-match precedence — mean a new switch in the branch subnet joins the branch group on its own. Retired hardware leaves the working views but keeps its history for reports; test boxes and mistakes are deleted cleanly if they were never monitored.


Two details that matter on a real estate: every discovered device gets its own random key, never the model's SNMP object ID or the address, so twelve identical switches are discovered as twelve devices rather than as the first of each model; and a discovered name follows the host when it is renamed, while the names your team chose are kept, with "Discovered as" shown beside them.

What discovery does for you

  • Finds devices and interfaces on its own
  • Files them into groups by subnet, type or probe
  • Lists flow exporters that are sending but not yet monitored, with an Enable button
  • Offers every sensor on every device; the ones discovery did not detect are marked, with why
  • Keeps history when hardware is retired
If you have ever inherited an undocumented network, this is the part that matters first: within an afternoon you know what is actually out there, as opposed to what the spreadsheet says is out there.

A map that reflects the cabling, not someone's memory of it


The device workspace shows the tree and the topology map in one place. Groups visibly contain their devices; dependency parents and children are stepped and connected; a Show filter, search and per-device tabs get you to the device you need in seconds, and the page remembers your view, filters and open tabs so you land back where you left off.


Because the dependencies come from LLDP/CDP on the wire, the map is also what the alerting engine uses to roll forty downstream hosts under the one switch that actually failed.


Network


Method and what it watchesWhat it means on call
SNMP v2c and v3, with instance walks and interface scans
Interface utilisation and error rates, routing, device health, printers, firewalls, VPN concentrators; v1 and v2c traps and v2 informs received
The whole switching and routing estate in one view, with per-interface trend history, no vendor-specific tooling required
Vendor and IETF MIBs named out of the box, plus your own
A compiled MIB set beside the service names walked OIDs, types and enumerations for the standard IETF and IEEE modules and for Cisco, Juniper, Arista, Fortinet, Palo Alto, HPE, Dell, APC, MikroTik and Synology, among others — about 2.2 million objects in 14,400 modules. Upload a device's own MIB files to name what the set does not know; uploads take precedence and are compiled again at every start
A vendor's counters are picked by name instead of by a dotted number, usually without finding and uploading a MIB first
ICMP, TCP port, DNS query, NTP clock, SIP OPTIONS
Reachability, service ports, resolver health, clock drift, VoIP signalling
The "is it up, and is it answering correctly" checks that catch the cheap failures before users report them
NetFlow v5/v9, IPFIX and sFlow v5 flow analysis
Listeners on the server and collection at remote probes; top talkers, conversations, per-interface traffic with interface names and speeds, QoS (DSCP) classes; sampling rates honoured. See Flow analysis below
"What is eating the bandwidth" answered with evidence — included in every edition, not sold per interface
QoS packet train between probes
Loss, jitter and latency measured across a real path
Proves whether the WAN or the VoIP complaint is the network's fault, with numbers
Passive LLDP/CDP topology capture
Neighbour discovery from the wire, no polling (needs Npcap on the server)
Dependencies appear on their own; the map matches the cabling

Flow analysis, on a page of its own


The Flows page sits in the main navigation and brings the flow records from your switches and routers into the same console as the sensors. It shows where to point an exporter, the live status of each one — receiving, no records, not recognised, silent since — and the top talkers, conversations, interfaces and QoS classes, filtered by site, time range and exporter. Every total says how much of the window the collector observed, and every time on the page is the server's clock, whatever the browser's time zone.


Heat strips show interface use in and out as a percent of the link speed, traffic by exporter and each QoS class's share, one cell per hour over a day or a week and coarser over 30 and 90 days, with an hour the server did not observe hatched rather than drawn as quiet. A seven-by-24 pattern of the last four weeks shows the Monday-morning surge and the nightly backup window as what they are: a pattern, not a one-off.


Interfaces are named as the device's SNMP sensors name them, scanned automatically when a monitored device starts exporting, so "ifIndex 14" becomes the uplink's real name and utilisation is shown against the speed the device reports. Two predefined sensors turn interface use in and out into a threshold in percent, seeded with a warning at 80%, so a saturating uplink alerts like anything else without a bandwidth figure typed in per interface.

On the Flows page

  • Exporter status: receiving, no records, not recognised, silent since
  • Top talkers, conversations, interfaces, QoS classes
  • Heat strips by hour, and traffic by hour of the week
  • Traffic sent by country, from a bundled DB-IP Lite database with no online lookup
  • Interface utilisation sensors in percent of link speed
  • Flow silence events on the device within minutes
  • Unmonitored exporters listed with an Enable button
  • Ports, retention and storage ceiling set in the console
A router whose export stopped becomes an incident on that device within minutes, not a quiet gap in a chart found days later.

When every exporting address of a monitored device has been quiet for longer than the silence rule, the server raises a flow-silence warning event on the device, naming the silent addresses, and closes it when flow arrives again; an event rule makes it an incident and a notification. Each monitored device that sends flow carries a flow chip in the tree and on the map, with "Sending flow" and "Flow stopped" filters, and its device page links each exporting address to its traffic. Long ranges read hourly totals kept in their own tables with their own retention, so a 90-day traffic view costs the server hourly rows rather than 90 days of per-minute records, and no total is ever drawn from data that was already pruned.


Servers, virtualisation and the operating system


Method and what it watchesWhat it means on call
WMI for Windows
Services, Hyper-V memory pressure, DNS server counters, terminal sessions, battery charge, and more; fans and most temperatures through LibreHardwareMonitor on the host
Deep Windows visibility with no agent to install or patch on every server
WMI explorer
Browse a Windows host's WMI classes property by property, with sample values and units where the host reports them; turn any numeric property into a sensor
The counter nobody pre-packaged is one click from a graphed, thresholded sensor, with no script to write
Windows event log subscription
Events become alerts through rules with severity mapping
The event that matters reaches the phone; the thousand that do not stay in the log
SSH commands, scripts and programs, file and folder checks
Linux hosts, custom checks, backup drop-folders, growing logs
Anything a shell can measure becomes a graphed, thresholded sensor — the escape hatch for the odd system every site has
IPMI and Redfish out-of-band hardware health
BMC temperatures, fans, power, via the management port
Hardware faults are seen before the OS notices, and while the OS is down
VMware vCenter, Proxmox, Kubernetes, Prometheus / OpenMetrics, node exporter
Hypervisor and container estates, plus anything already exporting metrics
Fits whichever hypervisor you landed on, and reuses metrics endpoints you already run instead of duplicating them

Applications, databases and services


Method and what it watchesWhat it means on call
HTTP sensors
Status, latency, JSON and text matching, certificate days remaining, minimum TLS version, weak ciphers, security headers
A web service checked the way an auditor would check it, with the certificate expiry shown as a date
Multi-step web journeys
Log in, navigate, verify — scripted end to end
"The site is up but nobody can log in" is caught as an outage, because it is one
Database queries; Redis; MongoDB; RabbitMQ
Any SQL result as a sensor, plus native health fields
Queue depth, replication lag and order backlog monitored beside the infrastructure they run on
Mail flow round trips (SMTP, IMAP, POP3)
Send, receive, time the loop
Email problems detected minutes before the "did you get my message?" calls start
LDAP bind and RADIUS authentication checks
Directory and network-access authentication, tested continuously
The failures that lock the whole company out at 8:55 a.m., caught by the monitor instead of the helpdesk queue
LLM / AI endpoint monitoring
OpenAI-compatible (including Grok), Claude and Gemini endpoints: availability, latency, throughput and token counts
As AI services become dependencies, they are monitored like any other dependency — something few monitoring tools offer yet

Cloud, plant floor and things that cannot be polled


Method and what it watchesWhat it means on call
AWS CloudWatch, Google Cloud Monitoring, Azure Monitor
Cloud metrics with the same thresholds and alert rules as on-prem
One on-call process for a hybrid estate
Modbus TCP registers, MQTT topics, OPC UA endpoints
Plant and building systems, spoken to natively
The warehouse, the clinic's building controls and the production line watched by the same team and tool as the office network
Passive push sensors
Devices and scripts POST readings in
Batch jobs, embedded devices and remote scripts still get graphs, thresholds and alerts

Security posture, as a by-product


Method and what it watchesWhat it means on call
Certificate Transparency issuance watch
Alerts when a certificate is issued for your domain that you did not expect
Mis-issued or unauthorised certificates — a phishing precursor — caught from the public logs before anyone abuses them
Domain hygiene: SPF, DMARC, DNSSEC
Continuous checks, alert on regression
The protections that quietly break during migrations stay verified — useful evidence for the cyber-insurance questionnaire
TLS certificate, protocol and cipher sensors; backup checks
Expiring certificates, legacy TLS, silently failing backups
Tickets with lead time instead of incidents

Numbers management actually asks about


Calculated and composite sensors

"Checkout service health" as a single sensor rolled up from the web journey, the database query, the queue depth, the mail loop and the payment API, with its own threshold and its own alert. It is the number the business asks about, and it lives in the same console as the interface counters that explain it.

The monitoring server is a device too

A built-in device for the SysPeep server itself, in the Self-Monitor group, watches collection failures and timeouts, poll-cycle timing, event rate, disk on the data volumes, host CPU and memory, database latency, and the date the disk runs out. Every remote probe reports its own health the same way. The usual blind spot is closed by default.


How collection runs


Whether a graph can be trusted depends on how the poller schedules and bounds its work. Every method above runs on the same collection engine, and it was built so that one slow device never costs the rest of the estate a sample.


A clock grid, not a queue

Polling runs on a fixed 30-second cycle; intervals are whole cycles from 30 seconds to a day, and a late cycle serves its missed slot rather than drifting the schedule or bursting catch-up polls. Sixteen collection workers by default (adjustable from 1 to 64) keep one session per device and method, with all of a device's WMI queries in flight at once, so one slow server no longer delays everyone else's readings.

Bounded, and honest when it overruns

The cycle waits at most 25 seconds; a slow device keeps collecting in the background and a hung one delays only its own sensors. When a poll takes longer than its interval, the chart, the device row and the sensors table say so in plain words — "Collection took 75 s, longer than the 60 s interval" — with the advice to lengthen it. A failed WMI or script check raises an alert that reads "error:" and the reason.


  • Discovery and housekeeping run off the poll thread, so a long sweep or a slow mail server never costs a sample.
  • One unit model everywhere: 9.7 GB, 5 d, 5 h 07 m read the same on pages, tests, maps, dashboards and notifications, so nobody converts kilobytes at 3 a.m.
  • Per-instance sensors: one sensor for a set of volumes, services or interfaces, whose status is the worst instance's and whose value is the worst instance, the total, or "2 of 9 failing" as the type demands, with each instance its own chart line. One full drive among nine is reported as the problem, not averaged away.
  • Test before Add: catalog sensors sit in five categories and 40 sub-categories with a protocol chip per row; search by what it measures, test it against the device, and only then add it. A card names which credential the test will use.

Questions administrators ask about coverage


Does it do cloud-based network monitoring: AWS, Azure and Google Cloud infrastructure monitoring, hybrid estates?

It is on-premises software that monitors cloud accounts: CloudWatch, Azure Monitor and Google Cloud Monitoring metrics come in with the same thresholds and alert rules as the switches, so a hybrid infrastructure has one console and one on-call process. It is not itself a cloud-based network monitoring service.


Does it monitor databases: Oracle, PostgreSQL, MySQL, SQL Server?

A database query sensor turns any SQL result into a reading with a graph and a threshold, which is how replication lag, queue depth and order backlog are monitored beside the server they run on; Redis, MongoDB and RabbitMQ expose native health fields. Oracle database monitoring and PostgreSQL monitoring are the same sensor with a different query.


Does it monitor Wi-Fi activity and the devices connected to Wi-Fi?

Access points and wireless controllers that answer SNMP expose radios, client counts and per-radio traffic, so Wi-Fi activity is a set of sensors like any other device. Who is on the wireless network and what they are sending comes from flow records at the switch the access points uplink to.


Does it monitor internet connection stability?

ICMP, DNS and HTTP sensors against the gateway, the ISP's next hop and a public address, on a 30-second clock, give a continuous internet connection monitor with latency and loss graphed and every outage on the record; between probes, a QoS packet train measures the circuit itself.


Does it monitor DNS, DHCP and IPAM?

DNS query sensors prove the resolvers answer correctly; DNS and DHCP server counters come over WMI on Windows servers; DHCP scope exhaustion and lease churn are the practical checks. IP address management as a category is covered in the IPAM and subnet monitoring series.


Does it monitor Apache, nginx and web server performance?

HTTP sensors measure status, latency, content, certificate days remaining and TLS settings; multi-step journeys verify the application behind the server; the host's processes, CPU and logs come over WMI or SSH. Apache web server performance monitoring is those three sensors on one device.


Does it monitor Dell, HP, Netgear and other vendors' switches?

Any switch that answers SNMP, with the compiled MIB set naming the counters of HPE, Dell, Cisco, Juniper, Arista and others out of the box and a MIB upload for the rest. Switch port monitoring is the same sensor on every vendor: utilization, errors, discards and state per port.


Can SysPeep monitor SD WAN (SD-WAN) edges and HPE iLO5 over SNMP or Redfish?

An SD-WAN edge appliance is monitored like any other router: SNMP for interfaces, CPU and tunnels where the vendor exposes them, NetFlow or IPFIX if the appliance exports flows, and latency and loss probes across each underlay link. There is no integration with a vendor's SD-WAN controller API. HPE iLO5 is covered twice over, by its SNMP agent with the HPE MIBs compiled in and by Redfish, which gives power, fans, temperatures and component health without an agent on the server.


Point it at the estate and see what discovery finds

The free edition monitors up to 5 devices at full depth with every collection method above and flow analysis included. The 30-day trial unlocks probes, scheduled reports and PDF export: every feature except clustering.

Download Free TrialNext: One executable on your server