Skip to main content

Network Management

The Network Management module in eEKAS provides a centralized,centralized and intuitive interface for configuring and monitoring all network interfaces and connections across the cluster. In a Ceph-based, high-availability environment, reliable and well-structured networking is a fundamentalcritical requirementfoundation for both storage performanceperformance, cluster stability, and uninterrupted service continuity.availability.

This module enablesallows administrators to define, adjust,modify, and troubleshoot network configurations without disrupting runningactive services.workloads, ensuring consistent operation during both normal conditions and failure scenarios.

Network Architecture and Requirements

To ensureguarantee stable cluster operation, eEKAS requires a minimum of two logically separated networks., Theseeach networksserving servea distinct purposes and areessential criticalrole towithin maintainingthe availability, consistency, and predictable performance under both normal operation and failure conditions.cluster.

Cluster / Internal Communication Network: This network is used for node-to-node communication, cluster coordination, and heartbeat signaling. It allowsenables therapid clusterfailure to continuously monitor node health, detect failures,detection and trigger automated failover actions. A dedicated anddedicated, low-latency internal network is essentialmandatory to ensure fast failure detection and stablereliable high-availability behavior.

Ceph Storage Network: The Ceph network ishandles used exclusively forall storage-related traffictraffic, such asincluding data replication, recovery, and rebalancing between nodes. Isolating this traffic prevents storage operations from interfering withimpacting client access and ensures consistentpredictable performance, especially during rebuild or recovery scenarios.events.

In addition to these mandatory networks, one or more client-facingclient access networks are typically configured to provideserve access to storage servicesprotocols such as SMB, NFS, iSCSI, NVMe-oF, and S3.

Separating

IP client traffic from internalGroups and storageClient replication traffic significantly improves scalability, security, and overall system stability.

Access

For block and file services, eEKAS networking is built arounduses IP Groups—logical collections of one or more IP addresses that can be assigned to specific services. IP Groups can be moved transparently between cluster nodes during plannedmaintenance maintenanceoperations or automatically during failover events, ensuring uninterrupted client connectivity.

Key Capabilities

  • Centralized Interface Management – Configure physical interfaces, bonded interfaces for redundancy and performance, and VLAN-tagged interfaces across all nodes from a single interface.
  • VLAN Support – Logically separateClient-facing traffic typesis suchstrictly asseparated management, cluster communication, storage replication, and client access without requiring additional physical NICs.
  • Multi-Network Architecture – Enforce strict separation betweenfrom internal cluster traffic,communication Cephand storage replication traffic, improving scalability, security, and client-facingoperational servicesstability.

    to

    Network improveBandwidth Requirements

    Network Bandwidth Guidelines: To ensure predictable performance and resilience.efficient Ceph replication, high-bandwidth and low-latency networking is required for all internal and storage-related networks.

    • Minimum: 10 GbE per network – suitable for small clusters, test environments, or moderate workloads.
    • HighRecommended: Availability25 IntegrationGbE or 40 GbE – Automaticideal failoverfor mechanismsproduction ensureworkloads serviceswith remainconsistent available evenperformance during noderebuilds.
    • Best Practice: 100 GbE or hardware failures.
    • Real-Time Monitoringhigher – Monitor link status, bandwidth utilization, and traffic patternsoptimal for proactivelarge-scale diagnosticsor andperformance-critical capacity planning.deployments.

    ByNetwork combiningBandwidth aComparison

    multi-networkarchitecturewithcentralizedmanagement logic,eEKASensuresstable, fortheservicesthat
    Tier BandwidthTypical Use CaseOperational Impact
    Minimum10 GbESmall clusters, testing, low I/OLonger rebuild and automatedrecovery failovertimes
    Recommended 25 / 40 GbEProduction, virtualization, mixed workloadsStable performance during rebuilds
    Best Practice100 GbE+Large clusters, high-performancedensity connectivitysystems Minimal bothimpact theduring storagefailures backendor andmaintenance
    depend on it.

    Example VLAN Scenario

    Layout

    • VLAN 10 – Management & Cluster Communication: Node management,Administration, monitoring, and heartbeat traffic.
    • VLAN 20 – Ceph Storage Network: Replication, recovery, and internal Ceph communication.rebalancing
    • VLAN 30 – Client Access: SMB, NFS, iSCSI, NVMe-oF, and optional S3 access.

    This separation ensuresminimizes thatcontention, criticalimproves clustersecurity, and storageensures operationspredictable remainperformance isolatedunder fromall clientoperating workloads, reducing contention while improving performance, predictability, and security.conditions.

    The model can be extended with additional networks or VLANs for backup traffic, dedicated S3 access, or other specialized workloads as required.

Back to top