Introducing LoxiLB - A cloud-native service load balancer

Hello Vyatta Community,

We would like to introduce open-source project “LoxiLB”.


LoxiLB is an open-source external service load-balancer for cloud-native workloads written from scratch using eBPF as its core-engine and based on Go Language. ​LoxiLB turns Kubernetes service load balancing into high speed, flexible and programmable LB services. It automates the tasks of external load balancer administration: deployment, bootstrapping, configuration, provisioning, scaling, upgrading, migration, routing, monitoring, and resource management. It is primarily designed to support bare-metal on-premise and edge cluster deployments via a k8s CCM provider plugin, however, it should run equally well in any public cloud or stand-alone environment.

LoxiLB itself is fully cloud-native and stateful in nature and completely blends in with any modern microservices architecture. Its purpose-built eBPF engine gives it various advantages over its other open-source counterparts like MetalLB/Ipvs, Iptables, HaProxy etc. For example, exceptional performance, scalability and the flexibility to support services ranging from simple (tcp/udp/http) to exotic ones (5g-gtp/sctp, nat66, nat64 etc). There are also tons of visibility and configuration options built-in.

Project repo :
https://github.com/loxilb-io/loxilb
https://github.com/loxilb-io/

We would like to explore the possibility of integrating loxilb into vyatta OS or if there is any such interest in the vyatta community or some pointers on how to proceed with such an undertaking.

Thanks and Regards,
@LoxiLB Developers

1 Like

loxilb requires a special version of iproute2 tool for its operation. The customized repository can be found here. loxilb needs some patches to iproute’s tc module to properly load/unload its ebpf modules

The thing which we should avoid
But good point for testing

Thanks for the comment. It is something we are working on and either we will upstream our changes or work with the community to find a solution.

1 Like