Learning From the Machine

Operational listening is a very powerful tool that we make considerable use of at StrikeIron. This is a general philosophy of putting some offerings out on the Web for public consumption with the proper instrumentation in place and then paying attention to what happens. This is opposed to smugly trying to predict the future and expecting the world to change to your requirements, rather than actively measuring the world and changing accordingly. Google for example does a pretty good job of this, as opposed to other technology companies with "five year plans."

For instance, we monitor our system load to the point to where we can find the fifteen minutes of the week where on average we see the absolute least activity and can plan any system maintenance accordingly.

We have built color-coded geographical maps to show us which regions of the world heavier access originates from and allows us to think about what that means for our business and what kinds of opportunities we might have as a result of where we see service adoption (and also where we don't see it).

We look at the individual Web services operations and see which methods are used more than others across thousands of users which aid in the development and architecture of future services.

We monitor service response times and pinpoint and then eradicate any speed bottlenecks as we manage and optimize system resources across our redundant server network via our system administration dashboards.

We can also see what services people are searching our catalogue for and not finding, which of course helps us determine future offerings.

And of course these are all looks and views we can provide to our service provider partners who wish to provide services via our platform and infrastructure, even to private audiences. They too can simultaneously learn from the usage of their services and data sources.

Putting a Web service endpoint up on the Web is not too difficult. But deploying it as part of a commercial and administered platform with all of these learning objects available to it is something else. Getting smarter through real usage data simply means better offerings, better service, and therefore better business for everyone.