DigitalOcean Load Balancer

Managed load balancer from $12/month

3.7(0 reviews)paid

managed optionhourly billingeu datacenterus datacenterapac datacenter

Integrates with

AWS Cloudflare Kubernetes Let's Encrypt IPv6

DigitalOcean's managed Load Balancer distributes traffic across Droplets automatically. Health checks, SSL termination, sticky sessions, and no configuration of Nginx required. From $12/month.

What DigitalOcean Load Balancer actually is

DigitalOcean's load balancing line is three products wearing one name, and the difference between them decides both your bill and your uptime guarantees. The default is the Regional HTTP Load Balancer: a managed Layer 7 balancer that lives in one region, terminates SSL, runs health checks against your Droplets, and costs $12 per month per node. There is also a Regional Network Load Balancer at Layer 4 for TCP/UDP workloads ($15 per node per month), and a Global Load Balancer ($15 per month flat) that routes across regions with CDN caching attached. We cover the whole line in our load balancers category.

The selling point is the same as the rest of DigitalOcean: you do not configure anything. You point the balancer at Droplets — individually or, better, at a tag, so any Droplet wearing that tag joins the pool automatically — it runs health checks, pulls unhealthy backends out of rotation, and returns them when they recover. Let's Encrypt certificates are provisioned and renewed automatically at no charge. There is no Nginx to patch, no HA pair to babysit, no cert-expiry alert at 3 a.m.

Pricing: what you really pay

The regional HTTP balancer is $12 per month per node, and each node adds up to 10,000 requests per second, 10,000 simultaneous connections, and 250 new SSL connections per second (50 with 4096-bit RSA keys). You can stack up to 200 nodes on one balancer if your account limits allow it, resize once per minute, and billing is prorated by the hour at each size — so scaling up for a product launch and back down afterward does not cost a month of unused capacity. For a small web app this means the honest minimum is $12 per month, and it is worth comparing against rolling your own: a $6 Droplet running Nginx does the same job for half the price, minus the redundancy, the automation, and your time.

The regional Network Load Balancer (Layer 4) is $15 per node per month, each node adding 50 Mbps of ingress throughput, with no defined ceiling on connections or requests per second and native client-IP preservation. The Global Load Balancer is $15 per month flat and includes 25 million HTTP/HTTPS requests, 1 TB of data transfer, and up to 5 connected domains; overage is $0.70 per additional million requests, $0.02 per additional GB, and $0.14 per month per extra domain. Those are the published rates from DigitalOcean's own pricing documentation, and we find them legible — no minimum-commitment games.

Two billing details worth knowing. First, the balancer itself does not generate bandwidth charges, but public outbound traffic passing through it counts toward your account-level transfer allowance and is reported under the load balancer rather than the backend Droplets — expect some invoice archaeology if you reconcile egress line by line. Second, the network load balancer's ingress bandwidth counts separately (50 Mbps per node is a real cap for L4 traffic; there is no equivalent RPS cap).

Performance & infrastructure

Feature depth is the pleasant surprise at this price. The regional HTTP balancer speaks HTTP, HTTPS, HTTP/2, and HTTP/3 (with an alt-svc header so browsers negotiate the upgrade), forwards TCP and UDP alongside them on the same balancer via configurable forwarding rules, supports WebSocket connections with a one-hour inactivity timeout instead of the usual 60 seconds, does sticky sessions, redirects port 80 to 443, and passes the real client IP through via PROXY protocol v1. External regional balancers can be created IPv4-only or dual-stack IPv6. Backend traffic stays on the private VPC network by default.

The health-check behavior is the part we care about most in production: checks auto-evict a failing backend and re-add it on recovery, and because backends attach by tag, autoscaling groups work without editing the balancer config. Combined with per-minute resizing, that is a credible story for handling a traffic spike without human intervention.

The uptime guarantees differ by product, and this is where reading the fine print pays. Regional Load Balancers carry a 99.9% monthly uptime SLA with tiered service credits (10% below 99.9%, 25% below 99%, 100% below 95%), and — per our reading of the SLA coverage details — a single-node regional balancer is not covered; the commitment applies to balancers of size 2 or larger. Global Load Balancers carry a stronger 99.99% SLA with the same tiered structure. Unavailability on the regional product is defined as five or more consecutive minutes of lost connectivity, so a pattern of short blips never enters the claim math. Independent outage tracking shows no major regional-LB outage in the last 365 days, which is consistent with our own experience of the service being boring in the way infrastructure should be.

Where it falls short

Start with the SLA structure, which we consider the weakest part of the product. A $12/month single-node balancer — the configuration the pricing page leads with — has no SLA at all. To get the 99.9% commitment you must run two nodes ($24/month), and 99.9% is a weaker commitment than AWS and Google publish for their multi-zone balancers (both 99.99%), and weaker than DigitalOcean's own Global Load Balancer. If load balancer downtime is existential for you, the regional product at size 2+ is the minimum honest configuration, and the five-minute unavailability floor means brief failure windows are yours to eat.

Second, the backends are DigitalOcean's. The balancer targets Droplets (and DOKS nodes) inside the DigitalOcean ecosystem; there is no healthy-backend routing to external origins the way Cloudflare or a DNS-based global balancer offers. That is fine if you are all-in on DigitalOcean and a real constraint if you run hybrid infrastructure. The Global Load Balancer narrows the gap — it spans regions and fails over across them — but it is still routing to DigitalOcean backends, and its 25 million request / 1 TB inclusion makes the $15 base price a starting point rather than a ceiling for busy sites.

Third, feature gaps versus dedicated traffic platforms: no WAF, no DDoS mitigation specific to the balancer beyond DigitalOcean's platform-level DDoS protection, no fine-grained traffic steering policies (weighted, latency-based, or rule-based routing beyond region proximity), and limited observability compared to, say, an Envoy-based mesh. Power users will still put Cloudflare in front for anything internet-facing and hostile.

Fourth, cost scales linearly with nodes, so a 10,000 RPS ceiling per $12 node is good economics at small scale and merely fine at larger scale; at 100,000+ RPS the comparison against a self-managed Envoy fleet or a hyperscaler L7 balancer should be run with real numbers rather than intuition.

Who should pick this

We'd pick the regional HTTP balancer when we are already on DigitalOcean and want HA web traffic handling without operating Nginx/HAProxy ourselves. Two nodes at $24/month with tag-based backends, auto-renewing certificates, and hourly proration is a clean, defensible setup for a small production API or storefront, and the DOKS integration (a Service of type LoadBalancer provisions one automatically) makes it the path of least resistance for Kubernetes users on the platform.

The Network Load Balancer is our pick for databases, game servers, and UDP/TCP streams where L7 features are dead weight and client-IP preservation matters. The Global Load Balancer is the pick for apps with users on multiple continents — cross-region failover plus CDN caching for $15/month is competitive with anything in the market at that scale.

We would not pick any of them for multi-cloud backends, for workloads that need 99.99% on a single region (the number simply is not offered there), or as a substitute for a CDN/WAF in front of a public site exposed to attacks. And if every dollar counts, a pair of Droplets running Nginx with keepalived undercuts the managed product — the question is whether your time is cheaper than $24 a month, and for most teams we advise, it is not.

For how the underlying nodes perform per euro, see /benchmarks.

Key takeaways

Three products: Regional HTTP LB at $12/node/month (L7), Regional Network LB at $15/node/month (L4), Global LB at $15/month with 25M requests and 1 TB included.
Managed end to end: health checks with auto-eviction, Let's Encrypt provisioning and renewal, HTTP/2 and HTTP/3, sticky sessions, WebSocket, PROXY protocol, tag-based backends, hourly-prorated resizing.
Read the SLA carefully: regional balancers are 99.9% and only at size 2+ (a single $12 node has no SLA); Global LB is 99.99%; unavailability requires five consecutive minutes.
Backends are DigitalOcean-only (Droplets/DOKS) — a real constraint for hybrid setups; no WAF or advanced traffic steering included.
Bandwidth passing through counts toward your account transfer allowance and is reported under the LB, not the backends.
Our view: the right default for DigitalOcean-native apps, the wrong choice when you need multi-cloud backends or hyperscaler-grade SLA numbers on a single region.

Sources

Pros

Fully managed: health checks with auto-eviction, tag-based backends, automatic Let's Encrypt renewal
Transparent per-node pricing ($12 HTTP / $15 network) with hourly proration and once-per-minute resizing
Modern protocol support: HTTP/2, HTTP/3, WebSocket with 1-hour timeout, PROXY protocol, dual-stack IPv6
Global Load Balancer at $15/month with cross-region failover and CDN caching is strong value
No SLA on a single-node balancer is stated openly in the SLA structure, and size-2+ gets 99.9% with tiered credits
Tight DOKS integration — Kubernetes LoadBalancer services provision automatically

Cons

A single $12/month node has no SLA coverage at all; the 99.9% commitment starts at size 2
99.9% regional SLA is weaker than AWS/Google multi-zone balancer commitments (99.99%)
Backends limited to DigitalOcean Droplets/DOKS — no external origins, no multi-cloud
No WAF, no balancer-specific DDoS mitigation, no weighted or rule-based traffic steering
Five-consecutive-minute unavailability floor deletes short blips from SLA math
Egress through the balancer draws down the account transfer allowance and is billed to the LB line item

Best For

DigitalOcean customers who want managed HTTP/TCP load balancing with automatic certificates and health checks without operating Nginx or HAProxy, and multi-continent apps that can use the $15/month Global Load Balancer.