Zabbix

Enterprise-grade infrastructure monitoring with predictive alerting

4.2(0 reviews)freemiumFounded 2001Latvia

free tieropen sourceself hostable

Integrates with

Zabbix is a free, open source enterprise monitoring solution for networks, servers, cloud, and applications (AGPLv3 since v7.0). Developed by Zabbix SIA in Riga since 2005, with paid support subscriptions and an optional managed Zabbix Cloud service.

What Zabbix actually is

Zabbix is a monitoring platform — network, server, cloud, application, service — that has been around longer than most of the infrastructure it watches. Alexei Vladishev started writing it in the late 1990s as an internal tool because the commercial options (HP OpenView, IBM Tivoli) were priced for enterprises he did not work for; the first public release appeared in April 2001, the first stable 1.0 in March 2004, and Zabbix SIA was founded in Riga in April 2005 to sell support around it. Twenty-five years of code, one owner-founder still at the helm, and the software still free: since version 7.0 (June 2024) it is licensed AGPLv3, before that GPLv2. The AGPL change matters mainly if you plan to offer Zabbix itself as a hosted service — network use triggers the source-sharing clause.

What does it actually do? Collect metrics from hosts and devices via a Go-based agent (agent2, with a plugin architecture) or agentless over SNMP, IPMI, JMX, HTTP and a dozen other protocols; store them in a relational database (PostgreSQL, MySQL, TimescaleDB and friends); evaluate trigger expressions; correlate events; and alert through escalating media — email, SMS, webhooks, chat integrations. Distributed deployments use proxies, including the load-balancing mode added in 7.0, and 6.0 LTS added server high-availability clustering, Kubernetes monitoring and ML-based anomaly detection. 7.0 added browser-based synthetic monitoring. The template system is the ecosystem's backbone: official templates for common vendors and services, plus a large community-contributed catalogue, all driven by a documented API and a Terraform provider.

The company story is steady rather than dramatic: headquartered in Riga, with offices in the United States, Japan, Latin America (Porto Alegre, operating since September 2020), Mexico, a distributor in China, and a French office created through the IZI-IT acquisition. Revenue comes from paid support subscriptions, training (the Zabbix Academy), certification, and — the recent part — Zabbix Cloud.

Zabbix Cloud is the piece that puts Zabbix in our /category/monitoring-hosting comparison at all: a fully managed SaaS instance of the same software, seven tiers sized by NVPS (new values per second), five regions, cloud proxies you can deploy near remote infrastructure, automatic upgrades, and a 5-day free trial without a credit card. If you want to run it yourself, nothing stops you — that is the whole point of the licence.

Pricing: what you really pay

Self-hosted Zabbix costs nothing for the software. Download it, install it, monitor as many hosts and metrics as your hardware sustains — no per-host, per-seat or per-metric licence, ever. The real cost is operations: a server, a database that grows, someone to maintain upgrades and tune triggers. That trade is favourable for teams who already run infrastructure and unreasonable for teams who do not.

Zabbix Cloud prices by monitoring throughput:

Nano: 50 NVPS — from $50/month
Micro: 100 NVPS — from $100/month
Small: 250 NVPS — from $250/month
Medium: 1,000 NVPS — from $750/month
Large: 2,500 NVPS — from $1,875/month
xLarge: 5,000 NVPS — from $2,500/month
2xLarge: 10,000 NVPS — from $5,000/month

Annual billing takes 10% off. New instances start with 10 GB of storage; storage is billed on top and — read this twice — the published caveats say disk can grow once per 24 hours and never shrinks, and that prices are starting points varying with configuration. Also, hitting the NVPS cap does not queue overflow values: extra values are dropped at random, which is a nasty failure mode for a monitoring system. Size up headroom. Unpaid instances are cleaned up on a schedule (stopped after roughly 5 days running past due, deleted 10 days later; trial instances without a card are deleted 5 days after the trial ends).

Compare honestly with the alternatives: Datadog bills per host plus per GB of ingest and gets expensive fast at scale; PRTG and Nagios sell per-node/per-sensor licences. A self-hosted Zabbix on a modest server monitoring a few hundred hosts is essentially free and stays free as you grow — which is why it is embedded in so many MSPs and enterprise ops teams. Zabbix's own paid support subscriptions (tiers from Bronze upward, priced by negotiation) are how the company funds development; support pricing is not public, which we count against transparency.

Performance & infrastructure

For the self-hosted form, performance is your design problem — and Zabbix's architecture is unusually well documented about it: housekeeper tuning, database partitioning, NVPS capacity planning guides, proxy fan-out. Large deployments (hundreds of thousands of endpoints, per Zabbix's own materials) are proven in the field; the release cadence is disciplined (standard releases every 6 months with 12 months of support; LTS every 18 months with 5 years — 7.0 LTS runs to June 2029, 7.4 is the current fast branch, and 8.0 LTS with CEP, OpenTelemetry support and a new UI is in beta).

For Zabbix Cloud, the infrastructure questions are shifted onto the vendor: instances are pre-tuned, updates and patching are automatic, and "24/7 uptime" is promised. What we could not find is a published numeric uptime SLA with credits for the Cloud product, nor a public enumeration of the five regions — both are things we ask any hosted service to publish, and their absence is noted here rather than papered over. On compliance, Zabbix states ISO certification and SOC 2 for Zabbix Cloud are in progress; if you need attestations signed today, verify before you buy.

We have not run Zabbix through our /benchmarks harness — our measurements cover compute and storage hosts, not monitoring stacks — so no throughput claims from us. The NVPS tier definitions are the vendor's own capacity language; use the pricing calculator with your real device counts and update intervals rather than trusting tier names.

Where it falls short

The learning curve is the reputation, and it is earned. Zabbix is configured in a web UI with trigger expression syntax, template hierarchies and discovery rules; nothing is opinionated, everything is exposed. Teams without monitoring experience will spend days getting from install to useful dashboards, where a SaaS APM would have shown something in ten minutes. The counterargument — flexibility, no lock-in, no per-host bill — is real, but it is an ops-time argument.

The UI has improved across 6.0 and 7.0 but remains functional rather than pleasant; log analytics and tracing are weaker than the Datadog/ Grafana-stack world, though 8.0's OpenTelemetry work and planned connectors (ElasticSearch, ClickHouse, QuestDB) target exactly that gap. For now, if your primary need is distributed tracing or log search, Zabbix is not the center of that stack.

On the Cloud side: the drop-overflow-at-NVPS-cap behaviour and the grow-only disk policy are the kind of fine print we dislike, and Cloud support is described as platform-level rather than the deep product support you can buy for on-prem deployments. And the free tier of the software is self-hosted only — the Cloud offer is a 5-day trial, not a permanent free plan.

Who should pick this

Our pick for ops teams and MSPs who monitor a lot of heterogeneous things — routers via SNMP, Windows boxes via agent, Java apps via JMX, a website via synthetic checks — and want one system whose cost does not scale with what it watches. If you have someone who can own the platform, self-hosted Zabbix remains one of the strongest infrastructure deals in software.

Zabbix Cloud is worth a look when you want the same tool without running the stack and your throughput is predictable enough to pick a tier — MSPs onboarding customers quickly, or teams standardizing on Zabbix without an ops hire.

We would skip it for: teams needing tracing-first observability, buyers who must have a signed SLA and compliance certs today, and small teams with no one to maintain a monitoring server who would be better served by a simple SaaS uptime checker.

Key takeaways

Twenty-five-year-old open-source monitoring platform (AGPLv3 since 7.0), built and owned by Zabbix SIA in Riga since 2005 — free to self-host, no per-host or per-seat licensing.
Broad collection: agent2 (Go), SNMP, IPMI, JMX, HTTP, synthetic browser checks, Kubernetes monitoring, ML anomaly detection; template- and API-driven with Terraform support.
Zabbix Cloud is the managed option: 7 tiers by NVPS, from $50/month (50 NVPS) to $5,000/month (10,000 NVPS), 10% off annual, 5-day cardless trial, five regions.
Cloud fine print matters: overflow values drop at random above the NVPS cap, storage grows once per 24h and never shrinks, and there is no published numeric uptime SLA.
Paid support subscriptions and training fund the company but are priced by negotiation — not on the website.
Best for ops-heavy teams watching heterogeneous infrastructure at scale; wrong fit for tracing-first needs or teams with nobody to run the platform.

Sources

Pros

Free, no per-host or per-seat licensing — cost stays flat as monitored infrastructure grows
Deep protocol coverage: agent2, SNMP, IPMI, JMX, HTTP, synthetic checks, Kubernetes
25 years of field-hardened releases with a disciplined LTS support schedule (7.0 LTS to June 2029)
Zabbix Cloud removes the ops burden without changing the tool; 5-day trial without a credit card
Open source under AGPLv3 with a documented API, Terraform provider and large template ecosystem

Cons

Steep learning curve — configuration UI, trigger syntax and template design need real investment
Zabbix Cloud drops values at random above the NVPS cap and disk only grows (once per 24h, never shrinks)
No published numeric uptime SLA or public region list for Zabbix Cloud; compliance certifications still in progress
Weaker on logs and distributed tracing than modern observability suites; self-hosting means you own the upkeep