VMware redefines its DevOps strategy for availability monitoring; so should you
Application Performance Management (APM) is an utter waste of effort–or an absolute necessity, depending on how you go about it. As “‘Performance’ means something new in a rationalized cloud architecture” last week described, “APM” covers a too-broad range, and you owe it to yourself to analyze your own use of APM very carefully.
One of the authors in last week’s focus, Bernd Harzog, is back with another instructive piece, “VMware vFabric APM End of Life“. This one packs in considerably more than is first apparent. It begins as a business report on EMC’s decision to retire its APM product, “vFabric APM“. This end-of-life means that VMware and other EMC units can offer only third-party products when an EMC proposal includes APM functionality.
That’s a legitimate business choice, and Harzog explores its background and consequences usefully. After concluding that “[k]illing vFabric APM was therefore a logical part of the process of rationalizing the VMware product line that has been going on since the arrival of the new CEO, Pat Gelsinger“, though, he launches what could well stand on its own as a valuable guide to the varieties of APM. Harzog makes a point which “IT Ops” frequently emphasizes: that the value of modern APM products is entirely different from that of the leaders just a few years ago. Harzog makes this especially concrete in “VMWare vFabric APM End of Life”: “mission-critical” extends far beyond one “flagship” Java-coded application running on a small nucleus of enterprise-class servers. Now, the emphasis is on agility, collaboration, and leveraging assets, and APM must automatically discover, account for a variety of implementation languages, deploy rapidly, diagnose at least partially, and update “hands-off”. It’s a tall order. While Harzog structures his six-step analysis process as advice to EMC customers ‘orphaned’ by the recent announcement, the counsel is valuable reading for anyone in DevOps, that is, with responsibility for availability and recovery. I particularly applaud his emphasis on practical definitions: “Any application worth yelling over is worth being monitored with APM solution.”
If you’re thinking about or doing APM the way it was done in 2005-2007, you’re probably just paying large licensing fees to achieve very little. The latest generation of APM products, though, can help you transition to the software-defined datacenter (SDDC) that has become the imperative goal for so many IT (information technology) shops.