I recently expounded on my disappointment with cloud computing services that fail to recognize that server metrics are not necessarily enough to properly auto-scale applications in “I Find Your Lack of Win Disturbing”. One of the (very few) frustrating things about working for F5 is that we’re doing so much in imageso many different areas of application delivery that sometimes I’m not aware that we have a solution to something that’s a broader problem until I say “I wish …” (I guess in a way that’s kind of cool in and of itself, right?)

Such is apparently the case with auto-scaling and application metrics. I know we integrate with IIS and Apache and Oracle and a host of other web and application servers to collect very detailed and specific application metrics, but what I didn’t know was how well integrated we’ve gotten these with our management solution. Shortly after posting I got an e-mail from Joel Hendrickson, one of our senior software engineers, who pointed out that “all of the ingredients inGrandma’s Auto-Scaling Recipe’ and much more are available when using the F5MP [F5 Management Pack].”

Joel says, “I think you’re essentially saying that hardware-derived metrics are too simplistic for decisions such as scale-out, and that integrating/aggregating data from the various ‘authoritative sources’ in application is key to making informed decisions.”

Yes, that’s exactly what I was saying, only not quite so well. Joel went on to direct my attention to one of his recent blog posts on the subject, detailing how the F5MP does exactly that.

Given that Joel already did such an excellent job of explaining the solution and what it can do, I’ve summarized the main metrics available here but will let you peruse his blog entry for the meaty details (including some very nice network diagrams) and links to download the extension (it’s free!), video tutorials, and the F5 Management Pack Application Designer Wiki Documentation.