Extending Load Balancer as a Service (LBaaS) in OpenStack Neutron with MidoNet

Cloud load balancers manage online traffic by distributing workloads across multiple servers and resources—automatically or on demand. Cloud load balancers give users a seamless application experience even under heavy load, making it a popular valued service for cloud operators to offer to their tenants in both public and private cloud implementation.

Existing implementations supported by hardware and software appliances have limitations in deployment, scalability and fault tolerance. Also, vendor specific implementation makes configuring physical or virtual load balancers manually tedious and error prone.

LBaaS is an advanced service in OpenStack Neutron, which allows a single set of APIs to be used to leverage core Load Balancing functionality provided by a multitude of vendors. MidoNet extends the LBaaS service in OpenStack Neutron one step further.  Building on the  Layer 4 Load Balancer feature in MidoNet from last year, the LBaaS feature in latest Midokura Enterprise MidoNet (MEM) release enables administrators to provide each tenant with distributed load balancing service from OpenStack Horizon UI.

While traditional load balancer vendors often tout their plug-in integrations with OpenStack, vendor-specific plug-ins can be intricately tied to lengthy hardware release cycles which can become outdated and out of sync with the rapid release cycles of OpenStack. Some of the plug-ins leverage basic Nova networking instead of Neutron which has its own pitfall when running at scale.

Unlike the distributed Layer 4 load balancers in MidoNet, the traditional hardware and software appliances weren’t designed for distributed computing right out of the box. To achieve the same result, the operators would have to manually configure a cluster of those appliances. They would also have to ensure that the appliances are placed on different hosts but ensure that the VMs behind the appliances are clustered on the same host and are not scattered across the clusters.  If VMs are placed on the same host, the failure of a host will bring the service down. If VMs are placed on different hosts, this will generate traffic tromboning as the traffic goes via the LB on one host, leaves the host to reach the VM and in some cases comes back to the host that contains the Load Balancer.

blg-lbaas

With so many moving parts, misconfiguring any one parameter can lead to hours of troubleshooting resulting in planned or unplanned downtime.

There’s a strong desire in the enterprise to reduce manual operations. A recent ESG study shows that 39% of the 601 IT decision makers surveyed are making network management a key area of investment, second only to network security. However, there’s an equally strong resistance towards changing the network.

And IDC study called Why the Network Matters shows that 22% of the changes to the network could potentially lead to service disruptions. The two opposing sides result in a deadlock that’s being played out across the organizations we’re working with.

“Digital Darwinism is unkind to those who wait” Ray Wang, Principal Analyst & Founder, Constellation Research

With network virtualization overlay solutions like Midonet, there’s no need to make tradeoffs between agility and risk. MidoNet can provide the safety of a software abstraction layer in between the end hosts and the physical layer which means that the downside of service disruptions can be brought to a minimum while providing for high volume of change.

With the advancements in LBaaS, use cases for MidoNet are many and we can’t do it justice with a blog post. Calling out a few modern use cases for consideration.

Cloud Operators can implement LBaaS for API-driven machine to machine communication. With the advent of the Internet of Things (IOT) and proliferation of of devices to device communication, the volume of machine-generated traffic is always on and growing exponentially which means there’s no down time allowed for network maintenance.

In recent years, a new standard for building applications, called the “Microservice Architecture” is gaining traction with developers. This approach to software design breaks complex applications into small, nimble, independent components. Designing software applications as suites of independently deployable services makes monolithic code bases a relic. This approach to software design is complementary with existing agile software development principles of continuous integration and sprints. Feature enhancements and bug fixes can easily be packaged into independently deployable services.

Network changes are often required with application changes. Much like providing each tenant a load balancer service, one could provide a load balancer service for each microservice, controlling a myriad of small instances that can scale up and down according to external, environmentally-controlled factors.

A third use case is Elastic Search which is designed to take unstructured, big data from any source and search, analyze, and visualize the results in real time. The volume of incoming data transmitted through the networks could grind to a halt without proper load balancing. The application demands from Big Data workloads can be sporadic but can also peak dramatically. With LBaaS, a network virtualization overlay solution like MidoNet is well-suited for enhancing traditional networks that weren’t designed for frequent changes and high density, massively distributed workloads.

 

Susan Wu

About Susan Wu

Susan is the Director of Technical Marketing at Midokura. Susan previously led product positions for Oracle/Sun, Citrix, AMD and Docker. She is a frequent speaker for industry conferences like OSCON, OpenStack Summit, Container World, Interop ITX, All Things Open, Linuxcon/CloudOpen/Containercon, Interzone, Data360/Cloudcon and Data Storage Innovation. Honored by the Cloud Network of Women (CloudNOW) organization as one of the Top Women in Cloud for 2013, Susan is passionate about advancing womens' contribution to the cloud industry and serves on the Board of Directors for CloudNOW. Follow Susan on Twitter @susanwu88

Comments are closed.

Post Navigation