As I'm getting set for EMC World (5th-8th of May) I was going through the things that I think will be important for customers to understand about the GEN5 technology. And there is an important feature called ClearLink that I think is becoming more more critical to customers especially in the data center. The concept is actually critical to the data center as a whole, but at the moment I will focus on the GEN5 infrastructure.
It is true that in the data center people consistently underestimate the importance of cable infrastructure to the environment. And frequently that includes a desire to reuse existing cable infrastructure in the next speed iteration. You saw this on the wired Ethernet side in particular in the transition from 100Mbit to 1Gbit or from CAT5 to CAT6. There were people who were convinced that they could "make do" with the existing cabling and the dangerous part of this is that depending on the distance and the quality the cabling you can. But the real danger is that the failure mode is not absolute, it sneaks up on you. And you see it in packet loss or retransmit rates.
On the Fibre Channel side of the house we saw this in the early days where people had 62.5µ multimode fiber (some of which was leftover from the days of FDDI) and wanted to reuse it for gigabit mode Fibre Channel. Similarly when we moved from the SC connector at 1Gbit to the LC connector and 2Gbit sppeed and 50µ multimode fiber became more consistently deployed. With the move from GBICs to SFPs and the standardization of the LC connector people became more cavalier about doing their own cabling, or about the people they hired to do it. And frequently the companies providing the cabling service or the people internal to the customer who get assigned the job have more experience doing copper cabling than optical cabling. This problem is made even worse by the fact that data centers are no longer "clean rooms" the way some of the early data centers were. And so dust and static electricity are a fact of life in the modern data center.
When you watch the way some people handle optical cabling it's almost as if they forgotten that light, nevermind coherent light, travels in a straight line and that we use the reflective properties of the cladding on the optical fiber to cheat and make light go where we want it. But you can only cheat so much and that's why there are definitions on distance and radius bend. Almost none of my data center customers wake-up on a Saturday morning and think to themselves "oh what a glorious day, tonight I get to go do cabling!" It can be a difficult job and all too often you see people treat the optical cable as if it was made of copper. When you see them running it through the cable trays overhead or underneath the flooring and it gets stuck you'll see them yank on it. Or you see them accidentally step on it when they laid it in a coil on the floor and then look down and act as if it's okay. We see them unplug the cable from an SFP and setting it on top of the next box in the rack "just for a second", ignoring the fact that static electricity will pull the dust to the connector. They plug the cable back into the transceiver and pack the dirt into the connection. The challenge here is that in most cases this won't lead to a hard failure. It is easiest when something dies, you find the dead component replace it and your running again. But when something is a little broken or it's an intermittent failure that's when you end up with problems that lead to the worst kind of headaches. And troubleshooting those in the middle of production is not how you want to start your day with the application owners.
Enter the Brocade ClearLink technology. Now many of you may be familiar with the D_Port function that we support in GEN5. This allowed us to test/verify optical infrastructure between switches and then eventually to Brocade branded HBAs. A useful function I'm certain you'll agree, but not all encompassing. With ClearLink you will begin to see HBA products very shortly from both Emulex and QLogic that will support this diagnostic capability to the servers. And since those same HBAs are also gaining design wins to be used in "target mode" in the storage arrays you will begin to see (my guess is within the calendar year 2014) array products from the storage OEMs that will ALSO support ClearLink. This means that in the GEN5 ecosystem (and ONLY in the GEN5 ecosystem) customers will be able to KNOW that their optical infrastructure is sound BEFORE they use it for production. And lets be honest, who wants to have a conversation with an application owner about a service window to check the integrity of the optical cabling.
Because Brocade built this functional set into the GEN5 ASIC, among some of the tests we have the ability to run are:
Run an electrical loopback from the switch ASIC to the back end of the SFP.
Run an optical loopback from the switch to the server HBA (and eventually the array).
Understand the latency on the link (and for a long distance link even the length of the cable).
And several other functions as well. As one customer discovered after an infrastructure company had completed the cabling job this can save you a huge amount of time and cost. When the infrastructure company left the premise the customer found that almost 15% of the ports in the environment had issues. Not hard failures mind, but thresholding issues that over time would turn into serious problems. They discovered that improper radius bends, improperly seated SFPs, dirty connections and glass fractures were some of the causes. With every speed step in networking the error margins shrink. Additional signal loss due to preventable issues in the optical infrastructure should no longer be tolerated by customers with critical performance environments.
So if you are at a point where you are looking at new or additional optical infrastructure, or you're going to be at one of the OEM shows sometime soon; stop by the booth and see us or give us a call and find out how ClearLink can save you some gray hair!