Publications

Libera for Programmable Network Virtualization

IEEE Communications Magazine, JCR 2018 IF 11.052 Top 2.44%, vol. 58, no. 4, pp. 38-44, April 2020, doi: 10.1109/MCOM.001.1900290.

Gyeongsik Yang, Bong-yeol Yu, Heesang Jin, Chuck Yoo

Download PDF

Abstract

Current network virtualization allows tenants to have their own virtual networks. However, new demands to "program" virtual networks at a finer granularity have arisen as tenants want the ability to provision and control switches and links in their virtual networks. This study proposes a new concept called the p-NIaaS model. The p-NIaaS model enables tenants to program their own packet processing logic and monitor network status from any virtual network infrastructure, which is not possible with current network virtualization. This article presents the Libera network hypervisor that implements the p-NIaaS model. Libera overcomes the shortcomings of existing network hypervisors such as scalability, VM migration support, and flexibility. The evaluation shows that the Libera hypervisor is highly scalable and effectively supports VM migration. We also present the overheads of Libera. Libera incurs up to 11 percent overhead in comparison with a non-virtualized network, which we believe is promising in the first prototype of the p-NIaaS model.

Paper summary

Make a tenant’s network programmable

Cloud tenants can program virtual network structure and forwarding policy.

Libera implements programmable network infrastructure as a service. The design allows tenants to choose virtual topology and install network policies while the hypervisor translates their requests to the physical network.

Coordinate topology, rule reduction and migration

The hypervisor translates a tenant’s virtual infrastructure and forwarding rules into the physical network. Its scalability component can reduce rules using different matching semantics, while LITE forwarding separates location, endpoint identity and tenant information. Migration handling updates the physical forwarding path when a VM moves, without requiring the tenant controller to participate. These functions work together so that an unmodified SDN controller can manage the tenant’s programmable network.

Libera’s coordinated network-virtualization functions.
FunctionPhysical-network action
Topology translationMap tenant infrastructure and policies
Rule scalabilityMerge forwarding under selected matching semantics
Migration handlingUpdate paths when a host moves

Measure rule savings and network reconfiguration

The scalability test compares several forwarding semantics with OpenVirteX in a four-ary fat-tree. The largest core-switch rule reduction is eightfold with hop-based wildcarding; savings depend on the selected semantics and switch layer. A separate hardware-switch migration experiment measures an average of 10.5 ms to reconfigure the network path after a host moves. The paper also measures rule-processing time, memory and control messages as connection count grows.

Physical flow-rule counts by forwarding semantics and switch layer, relative to OpenVirteX.View full-size figure
Flow-rule counts relative to OpenVirteX, grouped by forwarding scheme and switch layer. Hop-based wildcarding gives the lowest core-switch count.

Key takeaways

  1. Programmable virtualization gives tenants control over network structure as well as connectivity. Libera translates that control into physical topology and forwarding policies.
  2. Rule scalability and mobility share the same translation layer. Their coordination lets physical forwarding change without requiring the tenant controller to manage the underlying infrastructure.