J Metz, Cisco Systems
The most important part of any Data Center. Wait, wait, I get it – there are no unimportant parts of the Data Center. The ability to store and retrieve data, however, is the heart and soul, and the unmitigated catastrophe that befalls the company that loses data is impossible to overstate.
If you think about all the systems that are put in place to protect data – the generation, transmission, retention and retrieval (collectively known under the umbrella, “Storage”) – the bulk of the energy spent is devoted to one basic, fundamental task: give me back the bit I asked you to hold on to.
This is not new – in fact, the ability to manage Storage has been a fundamental part of every computing system ever devised. The only questions have been, 1) how good is my ability to store and retrieve the data, 2) how much can I store, and 3) how fast can I do so?
Building a Strong, Reliable, Dependable Network. For Storage.
Fibre Channel was not the first storage protocol solution – not by a long shot – but it did change the nature of how Data Centers worked. It could be argued that Fibre Channel was the first widely-adopted open standard interface that merged the performance and reliability of local storage with the connectivity and distance of networks.
Originally, and for several years afterwards, one of the main strengths of Fibre Channel was its raw power. At a time when I/O interfaces were capable of 20 MB/s (such as SCSI or ESCON), Fibre Channel rocked everyone’s world with 100 MB/s. While others worked in half-duplex mode (limiting data to being moved in one direction at a time), Fibre Channel was full-duplex. While others were limited to only a few meters (SCSI was limited to under 7m, and only 7 devices), Fibre Channel allowed thousands of devices to be connected over distances of kilometers.
The significance of this can’t be understated enough – we would not even be thinking of massively scalable storage networks had it not been for the pioneering efforts of the men and women who recognized that “status quo” is not “good enough.” Without Fibre Channel, it’s hard to imagine that there would have been a desire to push for massively scalable networks and global access to storage with any kind of reliability.
Yet here we are.
Over time, an interesting thing happened. Fibre Channel became known less for its speed advantages (though this was obviously a major draw for the first decade-and-a-half), and more for its reliability. People came to count on the technology because, well, you could count on the technology.
You knew, for instance, that the 5,000th device you added to the fabric was going to behave just like the first. You knew that it was going to be just as easy to setup, just as fast, just as reliable. You knew that you could buy Fibre Channel equipment and that it would be fully qualified, tested, and hardened. You knew that the cards you bought for your servers were going to work with the arrays that you bought, because the vendors had spent Billions of dollars (yes, with a capital “B”) to make it so. You knew that you could set this up and it would last for years.
The legacy of Fibre Channel, then, was not that it was so much faster than other technologies. (Admit it – did you remember exactly the difference between SCSI and Fibre Channel throughput in 1997?)
No, the true legacy of Fibre Channel is that in a world of disposable, commodity, unreliable components, there is still a technology that you can depend upon for storage, because it was built for storage, and that mission has never changed. Ever.
Looking to the Future
It seems that nowadays the trends of Data Center architectures are expanding and collapsing over time, much like an accordion. Depending upon who you ask, the pressure can be on increasing performance (which can limit scale), increasing scale (which can sacrifice performance), or manageability (which can sacrifice both performance and scale). Fibre Channel has solved – and continues to improve upon – each of these aspects of Data Center storage, and provides the underlying consistency that many people desire when adopting a new technology.
Throughout this Solutions Guide you’ve seen advances in speed, advances in topologies, and of course, advances in upper-layer protocols such as NVMe. Many people believe that NVMe is not just a new storage protocol – it means a fundamental, exponential shift in the capabilities and burdens inside of the Data Center. While this is exciting, and the possibilities for the future seem wide open once more, it is also scary for some – our job is still to securely and confidently store, transmit, and retrieve data without loss.
The questions become, then, how do you keep the “moving parts” to a minimum? How do you adopt new technologies like NVMe without creating upheaval in other aspects of your data center? How do you know that you’re getting the most out of your new investment, and not jury-rigging a cobbled-together network with too many untested pieces?
You know where I’m going with this. Fibre Channel has always been designed for situations like these – create a network that understands how storage is supposed to work, what the expectations are end-to-end, and know that the entire solution is qualified, from Operating System to storage media.
The Ecosystem Approach
At the end of the day, we want to make sure that when we innovate, we aren’t sacrificing what we already have. It’s okay to be excited about the potential for new technologies, but any experienced administrator knows that no technology works in isolation – simply adding NVMe to your data center is not a “one and done” situation, just like adding Virtualization wasn’t simply adding “another operating system” to your servers.
The impact and effect of these new technologies don’t change the nature of the storage problem; you still need to have store, transmit, and retrieve your data in a reliable, consistent, and guaranteed manner. Fibre Channel has two decades of an impeccable track record for these responsibilities, and will continue to remain the ultimate benchmark for complete storage solutions.