Skip to content

Comparison

pygslb is not the only way to answer a DNS name with a healthy backend. This page puts it next to the families of product an infrastructure team actually evaluates, states what each one is good at, and gives the public price of the ones that publish one.

How to read it. Every capability in the first table is one of: implemented in this repository (the file is named), or documented by the vendor. The prices in the second table were read on 14 November 2026 from the vendor's public page, in USD, taxes excluded; they change often and are given to set an order of magnitude, not to produce a quote. Where a vendor does not publish a price, the row says quote instead of inventing a number.

What this project is

An authoritative DNS server, one YAML file, a web console: a client in a given network gets the healthy backend of its own site, and a pool that loses all its backends falls back to a default pool instead of answering nothing. Three containers, no SaaS, no licence fee, one Python process on the resolution path.

Capabilities

pygslb F5 BIG-IP DNS Citrix ADC (NetScaler) Infoblox NIOS IBM NS1 Connect Route 53 Azure Traffic Manager Cloudflare Load Balancing PowerDNS + dnsdist gdnsd Knot DNS BIND 9
Form 3 containers appliance, virtual edition, cloud appliance, virtual edition DDI grid, physical or virtual hosted hosted hosted hosted daemons daemon daemon daemon
Runs on your own hardware yes yes yes yes no no no no yes yes yes yes
Routing key client CIDR, longest prefix, per zone wide IP, topology, geo, IP site/region, static proximity network, geo, site chains, geo, ASN, real-user metrics weighted, latency, geo, geoproximity, IP-based, failover, multivalue priority, weighted, performance, geographic, multivalue, subnet geo-steering, latency ("dynamic"), failover Lua rules in front (dnsdist), no CIDR-pool model geo map, weighted — views, RPZ
Health model http / tcp / PromQL probes; up, down, unknown; maintenance; opt-in per-pool breaker monitors (LTM) monitors monitors health-check monitors, feeds health checks, calculated and CloudWatch-based HTTP/HTTPS/TCP probes monitors backend checks in dnsdist plugin (metafo) — —
Declarative config file, reviewable in a pull request one YAML no (tmsh, GUI, iControl) no (GUI, API) no (Grid Manager, API) no (portal, API) no (console, API) no (portal, API) no (dashboard, API) no (SQL, files, Lua) yes (config file) yes (config file) yes (named.conf)
Console with per-pool state, routing counters, config history and diff yes yes yes yes yes yes (metrics in CloudWatch) yes (Traffic View add-on) yes no no no no
Licence cost AGPL, free subscription subscription subscription from $108/mo per zone + per query per query + per endpoint from $5/mo + plan GPL-2.0 GPL-3.0 GPL-3.0 MPL-2.0

Sources and prices read on 2026-11-14, re-checked against this release on 2026-11-16 (no cited capability changed at 0.7.0): F5 BIG-IP DNS, Citrix/NetScaler, Infoblox Universal DDI, IBM NS1 Connect, Route 53 features, Azure Traffic Manager, Cloudflare Load Balancing, PowerDNS/dnsdist, gdnsd, Knot DNS 3.6.0, BIND 9.

The pygslb column is traceable to the code: routing in gslb/mapper.py, health in gslb/health.py and gslb/probe.py, the breaker in gslb/breaker.py, the console in web/ and api/, the licence in LICENSING.md.

Cost, at public list price

Option What you pay for Price on 2026-11-14
pygslb Nothing to licence it. Two VMs and your time. 0 licence; support optional
Route 53 Hosted zone, queries, health checks Zone: $0.50/mo (first 25), $0.10 after. Queries: $0.40 per million (standard), $0.60 (latency), $0.70 (geo), $0.80 (IP-based) under 1 billion/month. Health check: $0.50/mo, plus $1.00 per optional feature (HTTPS, string matching, fast interval, latency measurement). aws.amazon.com/route53/pricing
Azure Traffic Manager Queries, monitored endpoints Queries: $0.54 per million (first billion), $0.375 above. Health check: $0.36/endpoint/month for an Azure endpoint, $0.54 for an external one. HTTPS or fast-interval add-on: $1 (Azure) / $2 (external) per endpoint/month. Azure Retail Prices API, "Global", USD
Cloudflare Load Balancing Add-on on top of a plan "Starting at $5/mo"; plans: Pro $20/mo billed annually ($25 monthly), Business $200/mo annually ($250), Enterprise custom. cloudflare.com/plans
IBM NS1 Connect Plan, queries included, monitors Trial free 30 days. Essentials from $108/mo (promotional $43.20, offer ends 2026-12-14) with 30–80M queries and 2 monitors. Standard from $380/mo (promotional $152). Premium: quote. ibm.com/products/ns1-connect/pricing
F5 BIG-IP DNS Subscription, sized by throughput Quote; F5 does not publish a list price.
Citrix ADC (NetScaler) Subscription or perpetual, per edition Quote.
Infoblox NIOS Grid licences plus GSLB add-on Quote.
PowerDNS + dnsdist, gdnsd, Knot DNS, BIND 9 Nothing to licence 0; support contracts and your own operations team are the real cost.

A worked example

Two sites, one internal name, six monitored backends, two million queries a month — a deployment that comfortably fits one pygslb node per site:

Option Monthly, list price, 2026-11-14
pygslb $0 licence (two VMs)
Route 53 $0.50 zone + 2 × $0.40 queries + 6 × $0.50 checks + 6 × $1.00 HTTPS = $10.30
Azure Traffic Manager 2 × $0.54 queries + 6 × $0.54 external checks + 6 × $2.00 HTTPS = $16.32
Cloudflare $5 load balancing + $25 Pro monthly = $30.00
IBM NS1 Connect $108.00 minimum plan (2 monitors; six checks need Standard at $380)

Numbers are arithmetic on the list prices above, not quotes: they exclude support, egress, and the staff time each option needs. They also show the real trade-off — a managed service costs a few dollars a month at this size, and the reason to self-host is usually where the name must be answered, not the bill.

Where the alternatives are the better answer

Being fair to the field is the only way this page stays useful.

  • Internet-facing names at scale. Route 53, Azure, Cloudflare, NS1 and F5 run anycast networks with DNSSEC, DDoS absorption and a support contract behind them. pygslb has none of that and does not pretend to.
  • Routing by measured latency or by geography. Route 53 (latency, geoproximity), Azure (performance), Cloudflare (dynamic steering), NS1 (real-user metrics via Pulsar) and F5 (topology) measure or model the client's position. pygslb routes on the source address, full stop.
  • A full DDI suite. Infoblox brings DHCP, IPAM, discovery and compliance to the same grid. pygslb is a DNS answer path and a small console, nothing else.
  • Regulated procurement. A commercial licence with an indemnity clause, a 24/7 number and a vendor to escalate to is a legitimate requirement; F5, Citrix, Infoblox and IBM sell exactly that.
  • Very high query rates. Knot DNS and BIND are built for orders of magnitude more queries per second than a Python process. pygslb answers about 10 000 q/s per core, 29 000 q/s on a node of eight workers (capacity, measured 2026-11-15); a resolver that must absorb hundreds of thousands of queries per second is not what this is.

Where pygslb wins

  • The name lives inside your network. No egress, no hosted dependency, no data leaving: the deployment is three containers and a YAML file on hardware you already run, and the DNS keeps answering when the management console is down.
  • The decision is reviewable. Weights, CIDR rules, probe definitions and maintenance flags are one YAML file in Git, diffed and versioned by the console itself. No GUI state to reconstruct.
  • Health that says "I do not know". A probe that cannot conclude leaves the backend in its previous state and logs a warning, so a monitoring outage never becomes a DNS outage. Vendors mostly model up/down.
  • The bill does not follow the query volume. At list price Route 53 charges $0.40 per million queries, so 2 billion queries a month is roughly $800 of DNS alone; a self-hosted node answers as much as the hardware allows for the same licence cost, which suits internal traffic that grows without a budget line.

What this page does not compare

  • DNSSEC, ECS, DoT/DoH, anycast, multi-tenant. pygslb has none of them; most authoritative servers on the market have at least some, with very different maturity. We did not verify each vendor one by one, so no column claims it.
  • Performance. Only our own measurement is quoted (bench/README.md), because vendor benchmark numbers are not comparable between products.
  • Features sold as add-ons. NS1 Pulsar, Azure Traffic View, Route 53 Application Recovery Controller and similar options exist and are priced separately; none of them has an equivalent here.

The use cases page shows what the shipped code does, and the security page states what is hardened and what is not.