Connect through one endpoint and distribute connections across available backends. Build an L4 load-balancing architecture on Cloud IaaS, with Ruk-Com expertise for clusters and autoscaling.
Illustration: new connections route to ready backends
INTERACTIVE / TRAFFIC LAB
Follow every connection. Understand every change.
Select a scenario. Watch round-robin skip unavailable backends and admit a returning node only after its readiness gate.
TCP :443 / IPv4SIMULATION · not live metrics
Client connections pass through an L4 load balancer to three ready virtual servers using round-robin.
Client connections
↓
L4 Load Balancer · Round-robin
↓
VM 01 · VM 02 · VM 03New connections route to ready backends
Client → BackendResponse to clientState / ready gate
Logical flow: validate the actual return path and health-check design for your deployment. Timing and readiness gates are illustrative. Existing connections may need to reconnect or retry after failure; they are not migrated automatically.
TWO OPERATING MODELS
Choose how your backends grow.
Use existing virtual servers or provision from a template using scaling rules. Align the model with your release process and traffic profile.
01
Load Balancer Cluster
Choose and manage existing VMs. Useful when each server needs explicit configuration review before joining the pool.
Select each backend virtual server and IP
Add or remove nodes through your change process
Plan capacity and distribution together
02
Auto-Scaling Cluster
Provision backends from an image template using scale-out and scale-in rules. Best suited to repeatable boot and deployment workflows.
Define minimum, maximum and resources per node
Choose a compatible template and nodes network group
Test both capacity expansion and node removal
TRANSPORT / CONNECTION DESIGN
Define the listener. Then shape the traffic.
Start with the protocol, port and backend pool. Design the application to operate across multiple nodes.
L4 · TCP / UDP
Distribute transport-layer traffic with round-robin selection. Match service ports to the processes listening on your VMs.
Connection affinity
The TCP example maps one connection to one backend. A new connection is a different event from another HTTP request on an existing connection.
TLS is part of the application design
For HTTPS through L4, define backend TLS termination and certificates. Include port mapping and certificate renewal in the deployment plan.
Example listenerTCP :443
→
Backend pool10.20.10.11–14
→
Service scopeIPv4 / Virtual Servers
DEPLOYMENT / BLUEPRINT
Good topology starts with complete inputs.
Work with Ruk-Com to align network, compute and backend design, with inputs your operations team can review and use.
Bring your diagram, port inventory, peak traffic profile and application constraints to the Cloud IaaS design discussion.
Match the compute zone/resource to the VM or template. The referenced modes exclude VMware virtual servers, Smart Servers and bare metal.
02
Network & addressing
Backends need an IP in the selected network zone and must share the same IP range. Also select the nodes network group for autoscaling.
03
Listener & service
Record the hostname, protocol and port set, plus the services required on each backend.
04
Capacity envelope
Set balancer port speed and CPU priority separately from template-node memory, CPUs and rate limit. Align min/max node counts with available resources.
AUTOSCALING / CONTROL LOOP
Scale out with readiness. Scale in with a plan.
Scaling rules are the starting point. Validate templates, application startup and available capacity together.
MIN 2 — MAX 4Example bounds in the Traffic Lab
Minimum preserves the configured node floor. Maximum sets the planned ceiling. Provisioning still depends on available resources and a working template.
01
Observe & decide
Evaluate traffic and scaling-rule conditions.
02
Provision
Create a VM from the template within the bounds.
03
Validate readiness
Validate the service before admitting new connections.
04
Join the pool
Include ready backends in the routing rotation.
Scale-in design: Plan to stop new connections, complete or handle in-flight work, then validate removal. The illustrated drain step is a design requirement to confirm for your deployment.
ENGINEERING / PRODUCTION READINESS
Design beyond distribution. Prepare for production.
Consider the network and the application together, so growth does not introduce new bottlenecks.
Health ≠ Application ready
An open port does not necessarily mean the application is usable. Plan dependency checks and test unavailable-backend behaviour.
Session & shared state
Review sessions, caches and runtime file writes. State kept only on one VM may not be available when a user reaches another node.
Database is a separate layer
Adding a backend does not replicate databases or files. Design a data layer and backup plan that support every node.
Retry & timeout budget
Set client timeouts, retry behaviour and idempotency around the workload, so interrupted connections do not duplicate transactions.
Capacity across the path
Test CPU, network, connection counts and downstream bottlenecks. More VMs do not guarantee linear throughput for every workload.
Availability & change plan
Multiple backends spread workload risk. Design load-balancer availability and maintenance windows as explicit parts of the architecture.
RUK-COM / ENGINEER TO ENGINEER
From network design to an operable system.
Work with a team that considers compute, networking and real workloads. Define the scope and handover evidence from the beginning.
01 / Assess & design
Review the traffic profile, listeners and backend dependencies before choosing the cluster model and Cloud IaaS resources.
02 / Validate & rehearse
Exercise distribution, backend failure, recovery and scaling against the plan. Record the results and limitations.
03 / Operate & improve
Hand over the diagram, configuration and runbook, including monitoring points and escalation steps within the agreed scope.
ENGINEERING / FAQ
Questions worth settling before you deploy.
Align the architecture and readiness expectations across your engineering team.
Cluster or autoscaling?+
Choose a cluster for explicit control of individual VMs. Autoscaling suits repeatable template-based backends with tested startup and shutdown behaviour.
Does round-robin distribute every HTTP request?+
This page covers L4. Its TCP example selects a backend for a new connection. HTTP keep-alive or HTTP/2 may carry multiple requests within that connection.
What happens to an existing connection after failure?+
Do not assume existing connections can migrate. Clients may need to reconnect or retry while new work routes to available backends. Test the actual application and timeout behaviour.
Does adding a node synchronize application data?+
The load balancer distributes traffic; it does not replicate application data. Plan sessions, databases, shared storage and backups for multi-node operation.
What should we bring to Ruk-Com?+
Start with the current diagram, protocols and ports, VM count, traffic profile and application constraints. These inputs guide the Cloud IaaS planning discussion.