Four of today's hottest technologies represent the most powerful drivers for the evolution of Application Management software that we've seen in the last decade: Cloud, Integration, Mobile, and Big Data. Each of these topics is changing the game in terms of application delivery, and generating significant performance and availability risks along the way.
Taken singly and in combination, these technologies are introducing new challenges and exacerbating existing ones.
Some of the risk factors include:
- Massive integration: My recent research shows that “integrated applications” are the most common type of production application, ahead of ERP, BI, custom, and packaged applications. Integration points are fragile, and relatively few integration products have “hooks” capable of sharing management metrics with Application Management tools and consoles.
- Heterogeneous endpoints: Management of mobile endpoints — phones or tablets – is a weak link for most Application Management products, with vendors just starting to provide instrumentation for mobile endpoints.
- Virtual extension of organizational borders: Companies are relying more and more heavily on applications and infrastructure they don’t own. This loss of control is a red flag, particularly as external applications and components integrate to those managed and controlled by IT.
- Abstraction of service delivery from underlying infrastructure: The idea of computing as a utility abstracts the service from the underlying infrastructure, reducing visibility/control over end-to-end execution even more.
- Geographical distribution of application components: Every network manager knows that distance introduces latency. Taking into account the intricacies of network configuration, and the fact that relatively few Application Management products have visibility to the network, again drives home the performance risks of massively distributed systems.
- Heterogeneous hardware and software infrastructure: Application performance is the sum total of the moving parts underlying the application. The more infrastructure and software elements supporting the application, the higher the risk, particularly if the management product lacks visibility to a given technology element.
- “Big Data”: More “moving parts” translates to more metrics. When we look at a transaction traversing 2,000 “hops” (and they do exist), the metrics generated by the underlying systems all become part of event correlation and problem diagnostics. Suddenly, application performance and availability management has become a Big Data problem.
Back in March, I wrote about Hybrid Cloud “FrankenApps” and the fact that, without visibility to off-premise tiers, Application Management and APM toolsets can't adequately support troubleshooting and root cause analysis.
In May, I wrote about Big Data and Hadoop as potential solutions for analyzing the enormous volume of log data and metrics thrown off by massively distributed and integrated systems. Very few vendors as yet are discussing their plans for dealing with the growing sea of metrics underlying application performance, and Big Data analytics could be one option.
Later in May, I wrote about Cloud and Integration in my article entitled Cloud and Integration: The Angelic Child and the Evil Twin. That article made the point that it is nearly impossible to talk about Cloud without taking Integration into account as well.
Research completed in June 2012 confirmed this, as it revealed that nearly 50% of companies have already deployed “Hybrid Cloud” spanning on-and off-premise applications.
(Note: “Hybrid Cloud” is defined in different ways, depending on the analyst or vendor. I've even seen vendors describing the term as spanning on-premise (private) Cloud and non-Cloud environments. Unfortunately, that definition doesn't work for me — managing such environments is relatively trivial compared to managing across organizational boundaries and into the public Cloud).
As a matter of fact, companies are already integrating Cloud in almost every conceivable configuration. For example, 39% are integrating across on-premise and SaaS, and 36% are integrating SaaS to SaaS or on-premise to IaaS.
From an Application Management perspective, you have to scratch your head and ask yourself how companies are monitoring and managing these complex interactions. And from my conversations with end users, I can tell you that most are doing it at a very superficial level. They run synthetic transactions from multiple locations, then triage problems manually by determining how far performance at a given location differs from the norm. Then they use silo tools to sift through the metrics from that location to find the source of the problem (although in many cases, specific sources are never determined).
The gap between the topics I mentioned at the top of this post — Big Data, Mobile, Cloud, and Integration — and the idea of manual triage are at odds with one another. Companies attempting to manage complex use cases in this way will likely experience an escalation in the number of “undiagnosed problems” that are never solved, and which repeat at intermittent intervals.
Later this year, I’ll be publishing Radar Reports on Hybrid Cloud Integration products and Application Performance Management for Hybrid Cloud solutions. Both of these studies will delve more deeply into these topics.
Meanwhile, those of you interested in more integration-related facts and statistics can check out the free and paid versions of my most recent Cloud Integration paper, and a free recorded webinar. All are available on the EMA site.
Rest assured, you haven't heard the end of the “FrankenApp” story — there is still a lot more to come.
Julie Craig is Research Director for Application Management at Enterprise Management Associates (EMA).