LavinMQ RabbitMQ Guides Last updated:

Shared vs. dedicated plans in CloudAMQP: LavinMQ and RabbitMQ

Shared vs. dedicated plans in CloudAMQP: LavinMQ and RabbitMQ
You're building an application that needs a message broker. Maybe you're processing background jobs, connecting IoT devices, or passing events between services.

CloudAMQP lets you choose between sharing a server with others or getting one of your own. This guide explains the difference between shared and dedicated instances and how each one works in CloudAMQP, for both LavinMQ and RabbitMQ.

This guide was first published in 2016 and updated in November 2026 to include LavinMQ and our current plans.

Quick answer: What’s the difference between shared and dedicated servers?

A shared instance puts you on a broker that other customers also use. You get a vhost and a set number of messages per month, plus published limits on connections, queues and queued messages.

A dedicated instance gives you your own broker, with resource isolation, predictable capacity, an SLA, advanced configuration, and private networking.

A multi-node cluster adds high availability on top of that. A dedicated single-node instance does not provide the same failover capabilities as a multi-node cluster.

Shared vs. dedicated at a glance

  Shared (multi-tenant) Dedicated
What you get A vhost on a broker shared with other customers Your own broker, or a multi-AZ cluster of 3 or 5 nodes
Isolation Logical only (vhosts); CPU, memory and disk are shared Full; no other customers on your servers
Limits A set number of messages per month, plus limits on connections, queues and queued messages No shared-plan message limits; capacity is set by the plan’s resources
SLA None 99.95% availability
Failover and replication None Automatic cluster failover and replicated queues on supported multi-node configurations
Control Use the vhost you are given Manage users, vhosts, permissions, firewall and VPC

Can’t find a plan that fits? If you’re looking for a serverless option, something new is on the way. Reach out to hear about it first.

What actually differs for your application on a shared vs. dedicated server

Isolation

On a shared plan you get a vhost on a broker that other customers also use. All vhosts run in the same broker process and share CPU, memory, disk I/O and network, so a vhost separates names, users, and permissions but not resources. If the broker runs low on memory or disk, it protects itself by blocking publishers, which affects every publishing connection on the node, not just the tenant that caused the issue. On CloudAMQP, this is regulated by the capacity of the plans.

On a dedicated plan, CloudAMQP provisions one or more VMs in the cloud provider and region you choose, and only your broker runs on them, so CPU, memory, disk, and network serve only your traffic. Read more about vhosts.

Control

Shared tenants connect to the shared server’s hostname and include their vhost in the connection URL. Because the server and its IP addresses are shared, individual tenants don’t get a firewall or VPC, and every tenant uses the same broker version, settings, and maintenance windows.

A dedicated instance gets its own hostname, can be placed in a dedicated VPC for peering and restricted with firewall rules, and you choose the broker version when you create it (LavinMQ instances can get automatic version updates). You can also change settings such as the heartbeat and the maximum number of channels per connection in the console.

Guarantees

Shared plans come with no SLA and no failover. Dedicated plans come with a 99.95% SLA and, on supported multi-node configurations, automatic cluster failover and replicated queues.

Cost model

On CloudAMQP, a shared plan has a set number of messages per month plus limits on connections, queues and queued messages. Both publishing and consuming count toward the limit, so 100,000 messages published and consumed add up to 200,000 message operations, and a spike can hit the message limit fast. When you run out of messages for the month, the instance is blocked until the count resets on the first day of the next month, or until you upgrade. A dedicated plan has no such limit: the plan’s resources set the ceiling.

Questions to ask when considering shared vs. dedicated

  1. How much downtime can you tolerate? Shared plans have no SLA and no failover. If an interruption can wait, shared is fine; if it can’t, you need a dedicated multi-node cluster.
  2. Will the instance sit idle for long periods? Free shared instances are deleted after two months without traffic; you get a warning email first. Dedicated instances are never deleted, but you pay for them while they exist.
  3. Do you expect traffic spikes or an uneven workload? Shared plans block the instance when the monthly message count runs out. Dedicated plans have no message limit, but a spike still means a larger plan or more nodes.
  4. Will you need to change plans later? Both brokers can move from shared to dedicated, but the migration differs: LavinMQ from the console, RabbitMQ through a process such as queue federation. See Migrate between plans.

Dedicated server, 1, 3 or 5 nodes

Dedicated plans come in three setups: 1 node for high throughput, 3 nodes for high throughput and availability, or 5 nodes for very high availability. In a multi-node cluster each node runs on its own VM, which is what makes automatic failover and replicated queues possible. A few RabbitMQ plans are not offered with 3 or 5 nodes, and on AWS and Google Cloud, multi-node clusters place their nodes in different availability zones.

Full feature comparison for CloudAMQP

Feature Shared LavinMQ Shared RabbitMQ Dedicated LavinMQ Dedicated RabbitMQ
24/7 email support Yes Yes Yes Yes
Management interface Yes Yes Yes Yes
Queue, connection and consumer alarms Yes Yes Yes Yes
AMQP 0.9.1, HTTPS, MQTT, WebSockets Yes Yes Yes Yes
Streams Yes Yes Yes Yes
Webhooks (push queues) Yes Yes Yes Yes
Delayed message exchange (built in) Yes No Yes No
AMQP 1.0 and STOMP No Yes No Yes
SLA (99.95%) No No Yes Yes
Critical support No No Yes Yes
Automatic cluster failover (multi-node) No No Yes Yes
Replicated queues (multi-node, supported configurations) No No Yes Yes
CloudAMQP Diagnostic Tool No No Yes Yes
Manage users, vhosts and permissions No No Yes Yes
Server metrics, node information, logs No No Yes Yes
Custom alarms (incl. CPU and memory) No No Yes Yes
Monitoring integrations No No Yes Yes
Firewall No No Yes Yes
VPC peering and PrivateLink No No Yes Yes
Plugins No No No Yes

Source: CloudAMQP plans and pricing.

Whichever you start with, you can change it later: dedicated plans can be resized, and both brokers have a path from shared to dedicated. Get started for free or talk to us about sizing and we’ll help you pick. Questions or feedback? Email contact@cloudamqp.com.