22. Workload Management Number of Incoming Queries Reject Filter Delay Throttle Reject Throttle Exception Reject Outside Of SLA Queries After Filters and Throttles Exception Reclass
Bad queries come in two forms: abusive or inefficient. Although the optimizer’s query rewrite capability catches and corrects a large percentage of bad queries, it cannot catch them all. With just a little tuning, many of these queries will run much faster and use a fraction of the original resources. Other workloads that disrupt the data warehouse include jobs that lock tables or records. These update or data loading jobs can block use of a table or collection of records for a long time. If the job runs for 2 hours, that’s how long locks are held. And while those locks are held, no other user can proceed if those tables or records are needed. Another kind of problem workload is the one that updates the data warehouse for 30-120 minutes, then crashes and rolls back. Such a job would hold locks for a log time and then longer still as the rollback occurs. Last we note that users can abuse the system. While we know the Teradata System can support these users and get the work done, a little planning or education may help with these workloads too. There are many batch jobs and DBA maintenance tasks that should run late at night but people insist on running them during peak processing time slots. These jobs consume huge amounts of system resources, effectively blocking others from using the resource. Another example is the user who submits 20 long running jobs into the system then goes to lunch figuring the work will be done when they get back,. But for the users who are working through lunch, the 20 large queries hog system resources and slow down the system overall. What normally happens is that long running or abusive workloads gobble all the resources leaving very little for the short running tasks. This occurs because each time the big workload asks for CPU or memory or disk use, it asks for a lot of it. So when the little tasks that run 1-2 seconds show up looking for resources, there are none left. This is what causes the 1 second query to take 5 minutes. Another aspect of running bad workloads and good ones together is congestion. There is often just too much work in the system at any given moment to run efficiently. Said differently, if there are 5 big workloads running and each wants 30% of the system, they all slow down to a crawl as the system attempts to serve them all. A better solution is to delay two of the workloads, let the first three finish quickly, then run the remaining two tasks when resources become less congested. Although being delayed may frustrate a user submitting tasks, it can actually occur that they get their work done faster when the congestion [Enter any extra notes here; leave the item ID line at the bottom] Avitage Item ID: {{E82248AB-ACC8-4DB1-B8FC-ACC90DD5538C}}
For performance SLAs Logarithmic scale Watch for peaks
Viewpoint is a web portal application framework with primary focus on Teradata Systems Management functionality that is integrated into the Teradata platform It supports Systems Management via a web browser and extensible to Teradata EDW end users and management allowing them to understand the state of the system and make intelligent decisions about their work day Viewpoint does NOT compete with Enterprise class portal products. It should not be used as a full portal infrastructure across an enterprise. While it can do many of the same things as WebSphere Portal, Oracle Portal, SAP NetWeaver, and so on, Teradata is not in the business of providing extensive SOA based portlet factories, federated portals, and other sophisticated portal integration capabilities. For example, Teradata will not supply a collaboration subsystem, rules engine, or a personalization engine as do some enterprise portal vendors. [Enter any extra notes here; leave the item ID line at the bottom] Avitage Item ID: {{0C45E380-AF09-4895-A958-9A4B112092FC}}
Filtered queries gives the DBA a list view of all queries or sessions running on a particular system. The dba can easily switch back and forth to different systems and also drill down into an individual session for more detailed information. [Enter any extra notes here; leave the item ID line at the bottom] Avitage Item ID: {{B24B2154-404D-4212-B2CA-BB25B4DEF77D}}