If you look for it, you'll discover that most everything around us has its own rhythm. There's a rhythm to the oceanic tides, to the music we listen to, and even the traffic we drive through. Inside ourselves, our pulse and heartbeat provide a constant rhythmic pattern that is fundamental to our lives.
Queries and other processing work running on Teradata Vantage platforms tend to have a recognizable rhythm as well. Query arrival rates often ebb and flow based on predictable patterns in user demand. Applications that are high-volume during the day may turn into a trickle at night, and then back to a flood of requests the next day, and so forth.
However, it is not just the flow of work among the different applications that undergoes change depending on the time of day or day or week. The business importance among the different types of work competing for platform resources tends to fluctuate as well.
Teradata Workload Management has been designed to make it easy to adjust to these swings in both processing patterns and business preferences. But before you can roll up your sleeves and define the workload management rules, the first step is to identify the patterns and priority changes that you want to reflect.
Recognizing the changes in user demand and in business requirements, and knowing when they occur, lets you plot out the different processing windows on your platform. A processing window is a discrete interval of time, usually measured in hours, when a specified combination of workload management rules will be enforced. Once the various processing windows have been defined, you can determine what changes in workload management setup need to take place as the clock moves from morning to midday, to afternoon, and then to evening.
State Changes
Refreshing workload management rules when one processing window gives way to another may seem to you like a lot of additional drudge work for the administrator. Especially because such redefinitions may be required several times a day or even in the middle of the night.
And you'd be right.
However, that extra burden on administrators can be avoided by smart use of Teradata Active System Management (TASM), the workload management component available on
Vantage enterprise platforms. TASM allows you to completely automate workload management changes from one processing window to the next. Once priorities and other rules for each processing window have been defined within the Viewpoint portlet, automatic changes can be instituted by means of an action called a "state change."
"States" are a bundling of workload management setup rules for the platform. Although many states may be defined, only one state will be active at a time. Having multiple workload management states is similar to having a different set of traffic rules in place during off-hours compared to rush-hour.
For example, your city might have traffic controls that only allow cars with two or more passengers to be in the carpool lane at rush hour, but during he middle of the day all cars can use that lane. Or busy thoroughfares can be given longer green lights at times when heavy traffic flow is expected, and shorter green lights during quieter times.
Cities don't pay for a traffic engineer to manually adjust the traffic lights each day when rush hour begins or ends. Traffic light behavior changes are programmed into the system and the happen automatically. The same is true with TASM states. Once the rules that apply to each state have been defined, states can be automatically turned on or turned off.
Using TASM, a new state takes over from the previous state by means of a "system event." A system event can be a simple thing like reaching a specified time of day. At 8 AM every weekday morning, for example, a state may become active that results in load jobs being de-prioritized and dashboard queries and other online work being elevated in importance.
System events that usher in a new state can also be triggered by less predictable situations, such as reaching a prescribed threshold of resource usage or losing a hardware component. A system event I've often seen used by Teradata customers is based on platform CPU utilization reaching a high-water mark, such as 95%, and staying there for a sustained period of time. TASM can detect that situation and react with a state change that would typically reduce concurrency levels of the less critical work until the platform cools off a bit.
Once the interaction between system events and associated states have been defined by the administrator, the changes to workload management setup take place automatically, without interrupting or slowing down active work. And states are easy to modify as the business expands globally or new applications come on board.
You can read more about Workload Management available on the Vantage platform in my white paper titled
Born to Be Parallel and Beyond.
User demand will shift, and business requirements will fluctuate. By keeping in tune with the rhythms of your workloads and the patterns in your business requirements, you can define TASM rulesets that support those changes and that can deliver them automatically, without anyone having to touch their keyboard or lift up their phone.
(Author):
Carrie Ballinger
Carrie Ballinger is a Vantage Certified Master with over 30 years with Teradata, currently working in performance optimizations within the engineering organization. During her tenure, she has focused extensively on workload management, MAPS, and other database technologies. Carrie is the author of more than 32 technical Orange Books, the most recent one on Native Object Store, and provides regular technical blog postings here on Teradata.com and on Teradata Community.
View all posts by Carrie Ballinger