Many vendors will claim to have invented the term “network fabric.” I’m not 100% sure who actually used it first but in my mind, Juniper was the first to go mainstream with it. It’s a little like MPLS was back in the day. Ipsilon actually invented it but Cisco evangelized it and made it a real market. So I’ll give Juniper the nod here and acknowledge that launching the project “Stratus” is when the term “network fabric” actually went mainstream. I believe Brocade was the first vendor to ship a fabric, allowing them to wear that badge of honor.
The launch of Stratus was several years ago and, since then, much has transpired. Stratus became QFabric, David Yen jumped to arch rival Cisco and QFabric has fallen under tremendous media scrutiny as a product that has failed to come close to the initial hype around it. So, the question I pose is whether QFabric is indeed the game-changing technology Juniper says it is, or it’s just a cool idea that never really catches on.
On the positive side, I will say that the concept behind QFabric is interesting. It’s very simple and elegant – a big distributed switch. I remember having a discussion with Juniper’s Andy Ingram about Occam’s Razor. That is the theory that states when there are multiple solutions to a problem, the simplest one is the best one. What’s simpler than a single-tier network? Nothing. So, according to that principle, QFabric should be a smashing success.
I also believe that Juniper truly believes the product is going to change the world (well, the networking industry) and, if deployed correctly, it can significantly reduce costs, improve performance and power in the era of cloud computing. In theory, I agree with this, if it all works as advertised.
So, why all the naysayers around QFabric? Well, there are a number of reasons. First, Juniper, where’s the beef from a customer perspective? Upon release of the “mini” QFabric and larger solution, the company claimed 150 customers. However, to use some NFL terminology, “upon further review,” when digging under the covers we find that many of those customers use QFabric components but aren’t deploying the QFabric architecture. If Ed Hochuli were reading the press releases he would wind up over turning them (another NFL reference, sorry for those who don’t watch football!).
To explain a bit more clearly - there are three components to QFabric: a node (top of rack switch), an interconnect (core switch) and director (controller) that make a complete solution. I’ve talked to a number of customers that are running parts of QFabric or running it in a lab but I can’t find anyone that has the total solution running in a production environment. Shamus McGillicuddy actually wrote a very nice article outlining this as well, which I will include for your viewing pleasure. So, anyone at Juniper who reads this PLEASE post in the comments section the name of a customer that is running the total solution in production. This doesn’t mean a customer using the 3500 as a top of rack switch. Don’t get me wrong, I believe the 3500 is a great TOR switch, but it isn’t QFabric.
Also, despite Juniper’s insistence that the product isn’t late, it is and Cisco has been letting everyone know this, beginning with the satirical pizza delivery ad where a customer had to wait 18 months for the Juniper pizza delivery guy to come. Whether it was late or not, the perception is that it’s late and, unfortunately, perception is reality. Since Juniper announced Stratus we’ve had a number of other vendors, including Alcatel-Lucent, Arista, Avaya, Brocade, Cisco, Extreme, Force10 and others, launch their data center fabrics. I might have missed someone in that list but you get the point. If Juniper’s intention really was to announce Stratus and then have almost everyone else beat them to market with a shipping product, you have to wonder what the company was thinking. I’d rather think that the complexity and magnitude of what they were building made it late before I believed this was the plan all along.
I’ve asked many customers about QFabric and the perception is that it’s the ultimate “all in” decision. It’s big, it’s proprietary and it’s expensive. I think it’s OK to be proprietary when you’re first to market or you have Cisco’s install base. But you certainly can’t be last and proprietary. That’s a tough combination, particularly when Brocade, Avaya and ALU can get you in with a handful of top-of-rack switches.
Again, I really like the vision behind what Juniper is trying to accomplish with QFabric. However, vision alone doesn’t cut it. Heck, vision and product often don’t. Look at the Apple Lisa. That flopped because it was way too early. Perhaps what Juniper needs is to have Scott Kriens come back and re-create that cult-like aura that Juniper had when he was CEO.
Anyway, to answer the question that titled the blog – Given the entire scenario, QFabric is tipping the scales on the side of flop. A handful of meaningful customer wins could tip the scales back, but clock’s a ticking…