AppDynamics Review

It includes transaction correlation, an application flow map, and business transactions. A universal agent might solve deployment and licensing issues.

What is most valuable?

The most valuable features are transaction correlation, application flow map, business transactions, and the key metrics that are displayed on the dashboard.

How has it helped my organization?

It gives a complete, end-to-end visualization of an environment. That's where it is a little easier to analyze any issue in production compared to the other APM tools. That's the key difference between the other tools and AppDynamics.

What needs improvement?

The first thing is that they are going in the right direction. That's the great thing because they're linking IT with business. That's why we mostly like it because the other APM tools are just talking about your IT. They're not linking that context to the business. You have your monitoring; your instrumenting; you're doing a byte-code instrumentation; you're doing a threat analysis. You have enough information. All you need to do is just play around with the data and give the visualization of business. What other APM tools are not giving, AppDynamics is great on that point.

As far as the features that we're expecting, the main thing is the universal agent that I’ve mentioned. They're not clear on what month or what year. I think next year, but they're not clear on the release date. That's one killer that we're really expecting. Because that will save a lot of time for an enterprise like us to go for a massive deployment. That's one of the key features I can say that we're looking forward to.

What do I think about the stability of the solution?

Stability is good so far. We haven't experienced any issues. But after a recent controller upgrade, we experienced a couple of downtimes of the controller, which is not good. On the agent side, not much. We do not see agents stop all of the sudden. We haven't experienced any such things. But in the APM space, the agent is a little tricky, so we have to be a little careful with the agent. With the previous experience around the Dynatrace agent we had, that killed the entire box. The box was completely down.

With this tool, we are taking a few more precautions; meaning, we're not going to production with the agent as of right now. We're putting enough load, enough applications, enough boxes and testing it for 2-3 months. Once we get confident, only then are we planning to go for the production.

Apart from the stability side – as I mentioned, the controller was down a couple of times recently, and the agent is working fine – the agent overhead is not good. It's taking a minimum of 200-300 MB per JVM or per CLR, which is the case with any of the APM tools in the current market. But we expect the overhead to come down. Then, it'll benefit us a lot. For an enterprise like us, we have a lot of shared environments. A box has 50-60 JVMs. A box has 300, 400, 500 virtual machines. In that case, if the overhead is 2-3 personned, we end up killing the box because we have the VM environments.

What do I think about the scalability of the solution?

How we look at scalability is in the controller's scalability. I think AppDynamics is not really there yet. The scalability should be very easy. I think that's what our expectation is. I think it's not even there yet. Controllers won't talk to each other. In a keynote session at a recent NetApp conference, someone was talking about or mentioned controller-to-controller communication. Once that is there, as long as controllers talk to each other, the scalability will become a little easier. That's on the controller side.

On the agent side, the scalability is the main focus area for us now, because we have 100,000 boxes, and we can't really deploy agents app-by-app, machine-by-machine, or manually. We can't really do that. Our approach is automated deployment. But with AppDynamics, the really tricky part is, they expect the application to be modeled in a certain way. They want us to define the app name, tier name, and node name, which is a little tricky.

I can just do mass deployment of agents, but then I have to do configuration also. That is where I think we're a little lagging, so we're working closely with them. We end up developing our own automation scripts to achieve that stage. Again, at that keynote session, people were talking about a universal agent. I think that might really solve the problems from both the deployment and licensing angles.

How is customer service and technical support?

With their support, if you raise a ticket, the response is very good. But the concern is the consulting days. I think they're offering some consulting days. In the first year of a contract, they offer a certain number of consulting days. After that, the consulting days are free. But to book a consultant, I think we need to book the consultant at least one or two weeks in advance. We can't really do that in the enterprise. A lot of things will happen. All of a sudden, we need support. That's a little tricky. We shared the feedback with AppDynamics. One or two days is what we can spend, but one or two weeks is really a problem.

Which solutions did we use previously?

For us, a large enterprise, the audits and the compliance issues; these things are serious concerns. We have 6,000 applications. We have 100,000-plus boxes. If management asks us, "Hey. Can you tell me what happened with a certain number of the boxes? Why did they go down?” Or “Why did certain applications go down? Can you please pull up the reason? Or can you tell me all of the applications a X person has accessed? Which boxes did he touch? Which routers did he touch?" We have no clue in a large enterprise like us. That's where instrumentation is key for us.

Our model is, we're shifting towards platform. Once we shift towards the platform, we want to offer instrumentation as a built-in stack in it. For that, there are two key things. One is explicit instrumentation, and the other is implicit instrumentation. For the explicit instrumentation, we already developed a solution last year. We’re now planning the implicit instrumentation. That's where we did a lot of market research. Our technology labs did a lot of market research. That's when we also went to the Gartner Report. Then, we finally chose AppDynamics.

How was the initial setup?

Initial setup was straightforward. There are two angles to it. The controller setup is pretty straightforward. The agent setup is also straightforward, but only if you are a simple tech startup or you have only one e-commerce application. For those kinds of companies, I think it makes sense. All they need to do is spend 2-3 days to set up everything. But in our case, we have 6,000 applications. I think AppDynamics is expecting the application to be modeled in certain way. I think we were asking them about this as well. I was expecting to get an update at that conference that they are moving away from that application modeling to something else. Once they move to that, I think that is also going to speed up the initial setup process.

Which other solutions did I evaluate?

There weren’t really any other vendors on our shortlist at the time. We are using Introscope extensively in production and we are using Dynatrace extensively in the lower-end environments. If this tool works out well, we're probably going to replace the other two.

As far as the most important criteria when selecting a vendor like AppDynamics, we have different stakeholders. Each stakeholder has their own use cases. The development team expects certain use cases. The support team expects certain use cases. The SWAT team expects certain use cases. Engineering expects certain things. TA expects certain things. We evaluated the tools from all the angles. On top of that, the future is cloud. The future is platform as a service. So, we want a tool that supports that era. That's where AppDynamics is the winner.

What other advice do I have?

The main point is every company is a software company. Invariably, you talk about it. Every button click is important. What if a customer shares feedback with his colleagues, friends or family? Every button click is important. Having said that, you should know what is happening out in your environment, out on your machines, out on your applications. So, application performance monitoring, infrastructure monitoring, and end-user monitoring are definitely very important.

We have our own use cases. According to those use cases, we chose AppDynamics. But whatever the product, don't get married to any product; whether it is CA APM, Dynatrace or AppDynamics. Even now, we're not married to any tool. We will always go with a tool which is going to fit in to our model. That's our message to anybody who's researching this case.

One important thing to note is that my rating doesn't mean AppDynamics is not great. AppDynamics is great. It is going in the right direction. At the conference, the CEO or somebody mentioned that they can't shove this product and develop everything that people are expecting for release by 2020. It's being done in a phased manner, in iterations. So far, whatever they have release to us, that's what the rating is for. That will probably be higher in the coming years. With the features that it has and with the expectations that we have, that's the rating we can say. And on top of that, AppDynamics gets the highest rating of any vendor in the APM space. If I rated the other tools, I'd rate them lower.

Disclosure: I am a real user, and this review is based on my own experience and opinions.
Add a Comment
Sign Up with Email