It’s probably not news to you that a significant part of the data flows within a data centre network today are machine-to-machine. Applications inside the data centre have been decomposed into many services that involve automated communication between servers inside that data centre: what has become known as ‘east-west’ traffic.
Although every data centre is unique, it’s commonly quoted that, on average, around 70 per cent of traffic in the data centre is between servers, and this is rising. The growth in east-west traffic is driven by virtualisation inside the data centre, and the need for flexibility to run any application on any machine.
Doing this successfully requires a network fabric to deliver consistent performance that is independent of server location. In a true fabric architecture everything is connected to everything else – ideally any device is just one or two network hops away from any other device or resource. The most important advantage of this is predictable low latency for this east-west traffic.
This networking problem isn’t new. Charles Clos, who created the Clos architecture that arranges the network switches in a hierarchy, was working at Bell Labs - in 1952. These architectures are often described as a 'spine and leaf' architecture. In this architecture, each ‘spine’ device is interconnected to each ‘leaf’ device, meaning that all devices can connect to all other devices through equal length paths.
Yet, as Clos and his engineers knew, this quickly introduces complexity. Spine and leaf and IP Clos architectures deliver a network fabric, but each element in the fabric is an autonomous switch, and each switch needs to be managed. Forming part of Juniper’s Metafabric architecture, Virtual Chassis Fabric (VCF) and QFabric improve on these architectures by also allowing the fabric to be managed and controlled as a single device.
VCF is an evolution of Juniper’s Virtual Chassis feature, which enables you to interconnect multiple devices into a single logical device, to work within a fabric architecture needed within today’s virtualised data centre. A VCF supports up to twenty data centre switches, and up to four of those can be configured as spine switches.
Beyond manageability there are several other valuable benefits to using VCF or QFabric architectures for your Data Centre network. Clearly, predictable low latency is one, delivered using an algorithm that contains the built-in intelligence to forward traffic by using the optimum available path to the destination. Plus, with multiple paths across the fabric, the architecture is resilient. It is flexible, using existing configuration procedures from a Virtual Chassis. Scalability is another. The QFabric architecture uses a pay-as-you-grow model, which can scale from a few hundred ports to 6,000 10GbE ports within a single-tier network.
VCF has an auto-provisioning feature that enables you to plug and play devices into a configured fabric. As the data centre expands, you can easily expand the size of your VCF by adding new top-of-rack switches to the fabric as easily as say adding a line card into a physical chassis switch. Managing the network fabric in this way becomes radically simpler.
As you plan your network fabric, Juniper’s commitment to an open architecture is also an advantage. In the long run, vendor lock-in may restrict the flexibility that you have created, but the same devices used to build VCF can, if perhaps required later, be redeployed in open spine and leaf or Clos architectures.
As you create a more flexible, scalable data centre architecture, your network fabric will either enable it or hold you back. For an insight into how VCF or QFabric can provide the foundation for your private cloud, discover how a global financial services company, based in Europe, are deploying Juniper’s Metafabric to address their challenges to significantly enhance their application performance in their data centres, while greatly improving efficiency and reducing costs.