Load balancing IBM Cloud Object Storage

The IBM Cloud Object Storage system is a breakthrough platform that helps solve petabyte (and beyond) storage challenges for enterprises worldwide. It uses an innovative and cost-effective approach for storing large volumes of unstructured data while still ensuring scalability, security, availability, reliability, manageability, and flexibility.

For high availability and scalability, IBM recommends that a load balancer is used to distribute client connections to the IBM Accessor node clusters. Load balancers monitor and perform health checks on a node to ensure traffic is routed correctly to healthy nodes.

Without a load balancer, an offline or failed node would still receive traffic, causing failures. A variety of load balancing methods are currently supported by IBM Cloud Object Storage Accessor nodes, dependent on customer infrastructure, including layer 4, layer 7, and geo GSLB / location affinity.

When the IBM Accessor nodes are deployed with the load balancer, clients connect to the Virtual Service (VIP) on the load balancer rather than connecting directly to one of the Accessor nodes.

The VIP can be configured using either Layer 4 or Layer 7, depending on the architecture of the network. In this case we recommend using Layer 7 as no network changes are required and SSL termination can be implemented.

This mode offers high performance and implementation flexibility, however as Layer 7 is a reverse proxy the client source IP address is not visible at the real server: instead, the IP address of the load balancer is visible at the real server. In order to retain the client source IP address, the load balancer inserts an X-Forwarded-For header into the load balanced traffic, which the IBM Accessor nodes can log for troubleshooting issues while seeing the true source IP address of connecting clients.

HTTP and HTTPS (SSL pass-through) load balancing

HTTP and HTTPs (SSL termination) load balancing

The load balancer can be deployed as a single unit, although we recommend a clustered pair for resilience and high availability. Details on configuring a clustered pair can be found on page 24 of our deployment guide, below.