Ever since I first heard about Juniper's "Project Stratus" about 2 years ago, I've been poking around to get more details about this top secret initiative. Despite many glasses of red wine, my Juniper contacts remained silent but I did get a few hints at the project from some big Juniper customers. Last week Juniper finally gave some details last week while I was skiing the icy slopes of Vermont with my son.
With the announcement of QFabric (i.e. Project Stratus's new name), Juniper announced a very cool, virtual layer 2 switching architecture that has the potential to radically change data center switching. Juniper's announcement was a tad on the geeky side, claiming that QFabric is being driven by cloud computing, virtual server migration, low latency applications, and server-to-server traffic patterns. While true, most enterprises are just beginning to experience pain from these forward-looking technology issues. Nevertheless, Juniper's QFabric announcement is extremely signficant for a number of reasons that the market seems to be missing:
1. Juniper out innovated an entire industry. Remember that Juniper has only been in the Ethernet switching market for around 2 years. In that timeframe, the company has delivered some of the industries fastest switches and created a switching architecture that eliminated a network tier. So what did the company do? Moved beyond this accomplishment with virtual switching, QFabric. To do this, Juniper had to think outside the box by putting all the intelligence at the edge and lightning fast transport at the network core. In summary, Juniper went from newcomer to thought leader practically overnight.
2. The IBM partnership just got a whole lot more interesting. With QFabric, IBM can now tightly integrate a data center infrastructure that ties together processors and storage. What makes this unique? Think of QFabric as a distributed backplane for high performance computing applications, massive data analytics, or cloud computing infrastructure. As if this wasn't enough, now layer on IBM's industry and application expertise and you can imagine "turn-key" data centers for smart grids, life sciences, or quantitative financial applications. Yeah, IBM could do this before QFabric, but it just became a lot more creative and customizable.
3. Juniper enters storage networking. Juniper was doing all of the IEEE gymnastics around an alphabet soup of storage networking acronyms like DCB and Trill, and will still participate in the storage-over-Ethernet game as it evolves. In the meantime, QFabric edge switches can be configured with native Fibre Channel ports that plug into the global transport tier. How's that for network convergence? The storage guys might not like it, but the balance of power is shifting toward the network -- Juniper QFabric just gave it another push.
4. QFabric advances Juniper's reputation for network performance. By flattening the network, every data center network node can be one hop from another. This combined with the transport-centric core vastly increases performance and reduces latency. Other vendors tend to leap frog each other, Juniper is always focused on raising the bar on the industry.
5. QFabric is not a "fork lift" upgrade. You can add edge switches and the transport core in a piecemeal fashion. This should generate a lot of pilot programs.
6. Economics, economics, economics. QFabric has the potential to do for Ethernet switching what server virtualization has done to physical servers -- substantially lower capital costs, operations overhead, and provisioning time for the network.
These advances lead to one additional conclusion that may trump all others: With QFabric, every organization considering a large data center project would be crazy not to bring Juniper in for a pow-wow. Wall Street doesn't seem to get this fact but I'm certain that it will.