This syndicated post originally appeared at Zeus Kerravala's blog.
I believe Juniper was the first mainstream network vendor to use the term “fabric” aggressively to describe its next-generation network architecture when the company announced QFabric back in early 2011. There may have been another vendor that used it first, but Juniper made it mainstream. Since then, almost every network vendor uses the term “fabric” broadly, and Juniper itself announced its next phase in fabric-based network when it unveiled MetaFabric in late 2013.
Subsequently, I’ve heard many customers, investors, and even some other analysts ask me what the difference was between MetaFabric and QFabric. The first thing to understand is that while QFabric and MetaFabric are related, MetaFabric does not in any way replace QFabric. I’ve heard some rumors about Juniper’s QFabric being retired because of MetaFabric, but in discussions with Juniper, I can most certainly say that this is not true. QFabric is here to stay.
The primary difference between the two is that QFabric is a single data center solution, whereas MetaFabric is broader. QFabric brings together several products (nodes, directors, and interconnects) to create a single fabric capable of supporting up to 6144 low latency, 10 Gig-E server-facing ports. In addition to the original model, there’s a “mini-me” version of QFabric that has 768 server-facing ports.
MetaFabric on the other hand is an architecture that ties together switching, routing, and software into a holistic solution for data centers and provides a path to cloud networking. MetaFabric is based on the concept that single-site data centers are rapidly being replaced with a more complex environment that includes multiple, distributed data centers as well as hybrid situations that tap into cloud-based services. This drives the need for a new type of network, which brings new challenges, and MetaFabric is Juniper’s answer to this emerging problem.
One good way to think about how one might use QFabric and MetaFabric is to think of the needs of a multi-data center business. In each data center, the company would deploy a single or even multiple QFabrics (depending on the size) to create a network that can better handle the explosion of East-West traffic from the growth of virtualization, big data, SDN workloads, and other trends. The MetaFabric architecture, then, would be used to tie the multiple QFabrics together to create a single network that could be managed end-to-end.
In both cases, the same universal building blocks of simplicity, openness, agility, and intelligence are utilized so customers can be assured of a consistent experience across a single QFabric or a broadly deployed MetaFabric.
With regards to addressable market, while QFabric has obvious appeal to large organizations, it also is a viable solution to mid-sized businesses, particularly since the smaller version of QFabric has been released. While 768 server-facing ports is indeed a lot of ports, companies can start smaller than that and then grow into it as needed. QFabric can be deployed without the director in a leaf-spine mode called Virtual Chassis Fabric. Mid-sized companies can start with just a handful of QFX Nodes and get the benefits of a fabric at a modest cost, particularly when compared to legacy technology.
MetaFabric’s primary appeal is larger enterprises with multiple data centers, although there are a handful of mid-sized companies that have unique requirements, where MetaFabric would be a good solution. However, for the most part the primary target market for MetaFabric is larger, distributed enterprises looking to build a dynamic, private cloud with location-independent compute resources.
Somewhere over the past few months the messaging from Juniper regarding QFabric and MetaFabric seems to have gotten mixed up, causing some market confusion. The best way to think of the two is QFabric enables automation, agility, and simplicity within a single data center, and MetaFabric extends those same principles network-wide.
Latest posts by Zeus Kerravala (see all)
- Thinking About the Cloud 2020 - March 20, 2017
- SD-WAN: Right for the Cloud Era? - March 14, 2017
- Microsoft and NVIDIA partner to bring GPUs to the public cloud - March 8, 2017