Operational Competency. Tim O'Reilly, You Like Totes Read My Mind!
Growing Pains, Operations, and Massages
My friends and I have parts of our digital lives spread out over various hosts and service providers. In both personal and professional use, it hasn't gone unnoticed that many places are struggling to keep up with operational issues. Even the
posterchild of scalability, Flickr, is regularly down for maintenance. (With Flickr it's not that disruptive, because they usually give tons of notice, and have a cute message up during downtime. ) But it's indicative of the state of things when the people who literally wrote the book on the subject, are regularly down.
I don't mean to attack Flickr. They probably handle operational issues better than anybody else. But what's going right now is that it's easier to
create a tool than it is to master its
implementation. And tools and services are getting adopted and integrated into our lives faster than we're learning how to use them, support them,... or live without them.
RUBY ON RAILS
I know of one web-hosting provider who is a leader in Ruby-on-Rails hosting. Though very technically sophisticated and literate, their service-level lately has been horrible for all web-related services, not just RoR. They do concede a change in performance and availability, but blame this on their users not using Ruby on Rails "correctly," hijacking resources from everybody else! This explanation is acceptable with a one-off fluke runaway process. But when the service-level for users and servers has been suffering for months on end, it's time to re-think what's going on! If it's clear that certain Ruby on Rails tacts can kill a server, then tailor adminstration for such occurences. Clear-cut = controllable. But it's
not clearcut. It's
not predictable, and in the case of this service provider, they should admit that the "right" way to use Ruby on Rails, for both developer and administrator, is still being worked out.
I'm not suggesting Ruby On Rails is a bad tool, simply that the pace of development with the tool has lapped the pace of admistration of the tool.
OPERATIONS: THE NEW SECRET SAUCE
This week, Tim O'Reilly wrote
a great post about operational/ deployment challenges of web services, and how operational performance could end up being a market differentiator.
One of his primary angles was comparing open source software to commercial software, in terms of operational options/ tools/ culture. He left the post undecided on whether any differences truly existed. But I was more interested in his phrase, "operational competency," a deceptively simple idea.
Whether software is for free, or for pay; the time, energy, and experience it takes to wire it for your needs is a crucial part of its utility to you. When we start to use and play with tools and services before we achieve operational competency, we need to understand the consquences. I guess this is typical early-adopter risks kind of stuff.
What can we do about these lopsided
production:
operational-competency relationships?
Related links:
TECHNORATI TAGS: scaling, performance, operations, sysadmin, oreilly, service, servers, deployment, Tim OReilly