It may seem that Managed Service Providers (MSPs) run their IT infrastructure just like everyone else. For the most part, that is true. The difference with most MSPs is that their business model demands that they get the most they can out of what they have; that includes hardware, software and their people – actually, that isn’t that different from the rest of us. So I wanted to look at one segment of IT in light of MSPs needs; Systems Management, and lay out what I thought were six crucial items for them in particular. But even if you’re not an MSP, most still apply.
1. Know what your customers are experiencing
It is crucial for any MSP, whether you are offering servers and storage in an IaaS offering, or a full-blown system in a SaaS offering, for you to keep track of what your customers are experiencing. Primarily, this deals with basic performance and availability around what you are offering:
- Can I get to the services I need?
- Are the services performing to our agreed on level, or what I perceive is acceptable?
It is very important to have a pulse on this information so that you can quickly react to problems; hopefully before they’ve experienced it themselves. What? How can you know about an end user problem before they do?
This is done by using active monitors; sometimes known as synthetic monitors. Their purpose is to execute transactions from a robotic end user computer, much like a real user would. They won’t catch everything, but it is a key part of every systems management toolkit. And, the good thing is, you can have these robots scattered around various parts of your network which gives you the perspective of the different geographies. It also helps in making quick isolation decisions to nail down the failing component.
But even beyond that, it is good to have an idea of frustrations your customers are experiencing. Are they having trouble performing a routine task, or finding information that they need? This can be measured by examining the user interface paths that customers are taking. For example, are they using the help screens? If so, what are they looking for?
2. Make sure alerts really are problems
Which is worse, not getting alerts when there are problems, or getting alerts when there are not problems? I think most of us would agree that you would rather be chasing a phantom problem, than being clueless when a real problem arises. But, there’s a point where you spend too much time chasing after problems that aren’t. And that’s when flashing icons start getting ignored; which is really dangerous. It’s like having a fire alarm that malfunctions to the degree that you begin ignoring it.
So what’s the solution? Use a mix of baselining (by component) along with dynamic thresholding? This way, over time you learn what is normal; by different days of the week/month, times of day, etc. And then you can set your thresholds (or have them set for you) accordingly. Is it foolproof? No; but it’s a lot better than the old way.
3. Elastic environments call for elastic management
Elastic; the word brings out thoughts of Cloud Computing. But it’s not only cloud. Virtualization is even more prevalent today and brings up equal challenges when it comes to systems management.
The point here is that servers can be added to or taken away from an application’s fabric in a heartbeat. And when that happens, your systems management solution needs to be in touch with these changes and do what’s needed to ensure that it is monitoring and managing these components appropriately.
4. Automate, automate, automate!
If it’s one thing that MSPs seemed to be thinking alike on, it’s the fact that they need to invest as much as possible in automating the common tasks of their internal processes. To MSPs, automation helps several ways:
- Increases flexibility
- Improves expense management
- Keeps the customers happy
If you really looked hard, you would probably find that the administrators and other IT staff perform the same tasks throughout a typical day; to the degree that they forget that their even doing it. Even if it’s a small task that takes 5 minutes for them to do, if it’s repeated enough over the course of time, it takes a lot of time; and interrupts them from more important jobs.
Automation is an even more important ingredient to any self-service framework. The goal for self-service should be to satisfy your customer request without any staff intervention and to do it as quickly as possible.
A good IT Process Automation (ITPA) tool can help you automate these tasks. And forget the whole idea of “lights out operation” that used to be preached. Automated tasks should be built around and in between the reality that there needs to be approvals and inputs from one or more people in the organization along the way. You’ve seen “The Terminator”; we’ve got to keep our eyes on these things.
5. Be able to look into the future; just a little…
There are two areas where being able to look into the future come in handy: detecting possible problems before they surface, and planning for changes in business needs.
Determining problems prior to their occurrence involves watching trends in resource utilization, workloads, transactions, response times, etc. Watching metrics over time is not foolproof, but it will give you a pretty good idea of when utilization problems will occur. Other, more complex, problems cannot be determined this way. This will need functionality which completely understands dependencies between workloads and the various components that feed them.
Now let’s look at the planning portion. Although the term “Capacity Planning” is a bit passé, there is a need for MSPs to be able to predict what is going to happen in the future. And I’m not talking about the traditional IT modeling tool that takes 6-12 months before it can help you, and is as fragile as a Faberge egg when it comes to configuration changes. But this functionality is needed in a couple of different ways:
- Based on current trends, where will bottlenecks be in the future and approximately when will this occur?
- Allow me to perform “What-if’s” with the configuration. Common areas to do this would be:
- Changing workloads
- Changing capacities of various components, such as networks, disk space, servers, etc.
6. Bring your systems and processes together
A single company almost never uses tools from just one vendor. They may have one tool for networks, another for SAN or databases, and yet another that works as a high level, general-purpose management framework.
So, from an enterprise standpoint, the desire would be to bring the metrics and events from all of these components together under one common dashboard. A typical catch phrase for this technology is a Manager of Managers; or MOM. MOMs allow the customer to generate higher level alerts, get a better feel for the overall enterprise, etc. One of the keys to a good MOM is to ensure that it has adapters to the most common frameworks (especially the ones you use).
But another reason to “bring things together” is to actually make them work together (“what a concept!”). Many times your ITPA tool will perform these functions. For example, you may need something that will tie your event engine into your ticketing system; and your backup system to your CMDB. How you do that, whether with the MOM, ITPA, or another tool, it doesn’t matter. Just do it.
To reiterate a bit, MSPs typically run their business on the notion of doing the very best job for as many customers with the fewest resources in the shortest amount of time; to the degree that no single customer of theirs could do a better job themselves. And it’s been my experience that when you find certain industries that have greater demands expectations of their software solutions, as MSPs do, other industries many times find that they are also moving in that same direction.
Mar 15 2012, 08:07 AM
Filed under: appmanager, Systems Management, dynamic thresholds, service level agreements, virtualization, NetIQ, Cloud computing, MSP, Aegis, IT Process Automation, AppManager Performance Profiler, Operations Center, Novell, SaaS, IaaS, ITPA, SLAs, Managed Service Provider, Manager of Managers, Customer Experience, Capacity Planning, Attenda, MOM, Ted Ernst, Data Center Solutions