Scaling OpenStack - painlessly!

OpenStack is an attractive solution for private and hybrid cloud service orchestration because it provides an open and extensible platform for delivery of virtualized compute, storage and networking services. One of the criticisms of OpenStack has been that it doesn’t scale very well. Until now the mantra has been “only big companies can afford the time and resources required to successfully make OpenStack operate ‘at scale’.”

Scaling OpenStack is becoming less of an issue, but it still has one area that is a concern with the out-of-the-box solution – the network. OpenStack Neutron is okay for small OpenStack deployments, but if you’re looking to run more than a few dozen compute nodes then you need a different approach.

For OpenStack to be a viable service delivery platform it has to be easy to use and reliable. Right up there in terms of importance is that it must be fast. Obviously there are limitations imposed by how fast virtual machines can boot. But the network should not be a bottleneck in the deployment of virtualized resources. Thus, OpenStack networking has to be proactive by design and “push” the network into place well before the virtual machines are ready to accept workloads. "Reactive mode" solutions simply don’t provide the speed and agility required to scale large OpenStack deployments.

A second aspect of scaling is multi-site connectivity to enable workload agility and mobility. Single site deployments are fine for a lot of situations, but today’s applications are becoming highly distributed. Companies are starting to take advantage of geographic economies of scale and operational efficiencies.  So the network has to effortlessly scale within and between data centers.

Connecting OpenStack instances across multiple sites must incorporate end-to-end quality of service. This requires that the multi-site connectivity support secure and reliable network protocols such as Multi-Protocol Label Switching (MPLS) to enforce bandwidth and traffic priority guarantees. But, as we all know, there are situations where the “best effort” of the Internet may be adequate for some service delivery models. In these cases Internet Protocol Security (IPSEC) tunnels may be sufficient.

Another aspect of OpenStack scaling is “cloud bursting” to public clouds such as Amazon Web Services (AWS). Sometimes it may make sense to use public clouds as an “infrastructure overflow” for short term projects, seasonal capacity requirements and unexpected flash-crowd events. Or, in many cases connection to third-party service providers that run in a public cloud may be required.

Just like connecting multiple OpenStack sites, cloud bursting should have multiple options: IPSEC for best effort, MPLS for secure, guaranteed performance, or in the case of AWS, using the AWS Direct Connect service may be desired, or required.

What does OpenStack scaling give you in terms of business and operational benefits? It can solve tough problems like legacy integration, business continuity, data sharing and replication, application scaling, etc. Workload mobility and agility are essential for a flexible DevOps-driven operations model. Applications need to be deployed fast and reliably – at the right place.

Over at CPLANE NETWORKS we’ve made single-site OpenStack networking “production ready” from day one with our DVNd product. Now we’ve taken it to the next level with our lightweight, high performance x86-based Overlay Gateway Router (OGR). OGR delivers OpenStack multi-site connectivity and cloud bursting across IPSEC, MPLS and AWS Direct Connect with as little as 3 clicks – and no scripting! When combined with our intelligent MPLS WAN automation solution, DVNi, true end-to-end quality of service can be realized.

So if you’re interested in making OpenStack scale to your demands, check us out – www.cplanenetworks.com

CPLANE NETWORKS Multi-Site OpenStack Networking and Cloud Bursting

---

Header Image courtesy of Dollar Photo Club