This will be the last in my current series of posts on Network Management, and as I look back on the topics of my previous posts it is clear to me that there is no future for Network Management. So should you make sure to say thanks to our wonderful hosts at Solarwinds while you still can?
Not So Fast…
Yeah, hang on a moment, Solarwinds isn’t going anywhere quite yet (at least, I sure hope not!). I don’t really mean that we will stop managing things. Rather, I suspect we will end up changing the focus, and perhaps the mechanisms, of our management. With that in mind I thought it might be worth opening up a conversation on how things might go in the future. My crystal ball is a little hazy so I’m going to be a bit vague, and I’d like to hear your opinions on how you see things going over the next 5–10 years.
When we talk about Network Management, it’s interesting that we’re focusing on managing just one part of the infrastructure, the part that for most companies is simply a transport for the stuff that makes the money. Mind you, we’re not alone in our myopia; server admins usually have a system in place to monitor their servers, but without the network, those servers aren’t really much use. The Storage team is keen on monitoring utilization, IOPS and similar, but maybe aren’t so concerned with everything else. The Security team monitors their Paranoia Alert and Notification Tracking System (PANTS), but that’s all they see, and the Database team is constantly squinting at their performance metrics and trying to improve them in isolation from everything else. I’m not saying that in the future we’ll stop keeping an eye on all these things, but I do believe we need to step back and start breaking down the barriers between these teams.
Rose Is a Rose Is a Rose Is a Rose
Resources are Resources are Resources are Resources. With the growing vision of a Software Defined Data Center (SDDC), it is becoming obvious that the current divisions that exist in most companies where we separate Network, Server, Database, Storage and Security are just not going to cut it going forward. The bottom line is that these are all just resources that support the applications (you know, the bits that make the money). To keep up with the ever-changing demands of applications, the SDDC starts viewing all these elements as resources to be provisioned on demand, in concert with one another.
From a management perspective this creates an interesting problem, which is that we need to have a wider view of our resources. We try not to allocate VMs on a host that’s already maxed out, and using that same logic, there’s no point allocating storage via an NFS mount on a SAN whose network ports are nearing saturation, or where the latency between the server and the SAN is suboptimal. We absolutely need to continue monitoring elements in the network, but we also need to understand how our resources interact and depend on one another in order to provide the best service to the applications using them. We need to understand not just where an application resides, but where its users and dependencies reside, and our –let’s call them Resource Management– systems will need to be able to monitor the infrastructure and present information and alerts to us in a way that is meaningful not just to a specific element, but to the tenants of that infrastructure. Do you spin up a new virtual firewall or use an existing one? Is a capacity problem one that should be solved by migrating VMs or by upgrading network or storage?
My point is, having separate teams watching over separate management systems is a bit like trying to walk through Grand Central Terminal with your eyes focused firmly on your feet.
“What has been will be again, what has been done will be done again; there is nothing new under the sun.” (Ecclesiastes, 1:9)
What would a Resource Management System (RMS) look like? I wish I knew. Maybe Solarwinds have some insight; they’re already using information gathered by one product to enhance the information in another. At some point, maybe we’ll stop seeing different element management products being sold and instead we’ll be licensing a more holistic product suite that, yes, manages all elements, but whose core benefit is the sum of all those elements. Such a system could actually become part of the larger Software Defined Data center as a provider of information that can be used to make automated deployment decisions.
Alerting needs to get smarter, and I’ve seen many previous attempts –often home grown– to relate an element failure to a service impact. It’s much easier to do this if we begin managing applications in relation to the resources they use rather than managing elements and trying to figure out who uses them afterwards. I think we have the ability to make this happen fairly quickly, but it does require that we start tracking applications, users and dependencies from the get go. Adding these things in afterwards can be a nightmare.
Is this utopia or hell on Earth? How far are we from this goal, or are you already there? Or –and this is entirely possible– am I way off the mark with this entire concept? You are out there using Solarwinds today, so what do you think you’ll be buying and using from Solarwinds in the future?