IT RULE #4: If you can’t measure, you can’t manage. (And if you can’t see, you’re not going to get anywhere)
It might be a bit strange to start a blog quoting a rule. But over the last 37+ years, I’ve learned a number of rules from my IT customers and this one is foundational. How is any IT team going to be successful at keeping the business operational and meeting SLAs without the ability to measure? It’s like driving with a blindfold on. Should you accelerate? Brake? Turn left? Turn right? You have no idea what to do to improve your performance or to avoid potential problems.
How does this tie into a discussion of Brocade’s Gen 6 Fibre Channel environments? Frequently when I’m introducing the latest Brocade technology, I’ll start with an acknowledgement of the speed increase and move on. But in the Gen 6 discussion, we need to walk through this a little more deliberately.
PERFORMANCE: The ecosystem for Gen 6 has accelerated faster than the previous generations of Fibre Channel.
For the first time since the beginning of Fibre Channel, HBA adapter vendors (Emulex and QLogic) delivered their new generation of technology ahead of everyone in the ecosystem. This translated into market availability of the first Gen 6 32 Gbps capable storage array last September from NetApp, a mere 5 months after the launch of the first switch platform: the Brocade G620. This gives application owners access to a serious level of performance. But it is also a critical point of discussion due to the rate at which Hybrid Flash and All Flash Arrays have appeared in the market. In fact, I had two customers in recent months divulge that they no longer plan on buying anything but Flash. The impact on existing infrastructure is when new Flash systems are mixed into older networks, the customer doesn’t see the performance gain that was expected with the purchase. Testing has shown that an 8 Gbps All Flash Array will see significant performance improvement (as much as 4x increase in IOPS and 4x decrease in latency) by leveraging a Gen 6 infrastructure. Customers want to be able to serve the maximum number of servers from their Flash storage and Gen 6 meets that challenge. A single 8 port Gen 6 switch, like the Brocade G610, can be a starter platform for mission critical business applications that need the performance at an excellent entry price. And as your needs grow? You simply enable additional ports up to 24 ports via software all without impacting your running applications.
MITIGATION: But what do you do when challenges occur when there are mixed technologies in the environment?
Imagine the 80-year-old driver in their 1980 Volvo who is in the high speed lane on the highway. They aren’t broken or illegal. They are still critical (they are somebody’s grandparent) and you want them to get where they are going, but they are creating traffic issues. In a Brocade Gen 6 platform, latency is measured on every port in the network every 2.5 microseconds. When a slow device is identified, we carefully move it to a lower priority lane where it can perform at its level without impacting the high performance traffic. All without dropping a single frame.
INCLUSION: As new technologies such as NVMe make their way into the market, how will customers manage the application needs? Which applications or VMs will be able to take advantage of an NVMe performance? Should you build NVMe resources into every server? Should you stack all the NVMe-type applications in a few platforms? The former seems terribly expensive and creates a lot of stranded expensive resource. The latter would put too many of the heavy workload/critical business applications into too few servers. But what if you could seamlessly include this new technology as a shared option in your existing Brocade SAN? What if the same server could concurrently send traffic over the same Fibre Channel HBA to either a traditional SCSI target or a new NVMe target? And do so all without changing the SAN fabric? That’s the reality of Brocade Gen 5 and Gen 6 today.
MEASUREMENT: Now we come to the crux of the problem. IT staff are asked to commit to an SLA to the business line or application owner. How will you monitor it? How will you know that you are meeting the SLA? How will you know how much the application is consuming? Brocade Gen 6 platforms actually measure every frame on every port. Not sampling and forwarding to an analyzer or to the switch CPU, but actually measure every frame. No other product has this capability. I was in a discussion with a customer recently who stated that in their VM environment they would see latency spikes in sub 20 ms intervals. This was causing disruption to the applications, but the problem interval was too small for the network monitoring to pick up. To make it worse, it’s common not to be able to poll the Ethernet switch for performance data more than once every 5 minutes without causing a DDoS effect on the management port. In other words, it’s the functional equivalent to looking at only one car out of 100 on the freeway and asking the traffic cop about it once every 5 minutes. Is it any wonder that most network teams respond to the storage admin’s plaintive query of, “Is there a problem with the network?” with the answer, “We don’t see a problem.” But in a Brocade Gen 6 environment using Fabric Vision you can.
With the VM Insight capability built into Gen 6 switches with Gen 6 HBAs and storage, there is now the ability to track IO performance all the way to the VM! In my recent discussions with more than 1,000 customers, I asked them all this question:
“Do you have, anywhere in your environment, an application owner you believe that you can look in the eye and ask the IO profile of the application and get an answer you would trust?”
They are generally laughing before I can finish the question. I’ve had only ONE positive answer from a customer and he told me that he thought that one application owner of the over 600 in his environment might be able to do that. The next question I ask is, “Would you like to be able to measure it?” And the answer is an absolute yes. The problem isn’t that people haven’t wanted to be able to see this, the problem is that the granularity hasn’t existed. Seeing past the hypervisor to the VM and being able to track that traffic across the network has previously never been possible. Now with VM Insight in Brocade Gen 6 platforms, it is. Ask yourself this question, when the freight trucks are running at Ferrari speeds on your network, will your already inadequate network monitoring be “good enough?” I don’t think so.
So to come full circle: “How do you manage what you can’t measure?” How can you know which direction to take?
The answer is: you can’t. Measurement and its analytics are critical to determining your performance against SLA demands. Measurement is critical to diagnostics and troubleshooting. And it is critical to planning. Isn’t it time that your storage network helped you solve this dilemma? Brocade Gen 6 products are here to help you do it.