Top Ten Db2 Performance Tips - No. 7 Configuration Parameters and Tuning

Configuration parameters play a crucial role in optimizing the performance of an IBM Db2 database environment. By understanding the impact of these parameters and tuning them appropriately, database administrators can fine-tune the system to achieve optimal performance. In this blog post, we will explore key configuration parameters and discuss recommended settings and tuning options.

The first thing to understand is that the actual parameters, their specific names, and how they are implemented and modified will differ between Db2 LUW and Db2 for z/OS. This blog post will talk mostly in generalities but the ideas resonate across all Db2 database management systems (and, indeed, even across other relational DBMS platforms).

The first type of parameters are used to configure Database Manager options. The database manager configuration parameters govern the overall behavior of the Db2 instance. Parameters such as database shared memory, lock list, and log buffer size have a significant impact on performance. Adjusting the shared memory size to accommodate the workload, appropriately sizing the lock list to handle concurrent transactions, and setting an optimal log buffer size based on the transaction rate are essential considerations.

Next we have the Buffer Pool parameters. Buffer pools act as a cache for frequently accessed data pages, reducing disk I/O and improving query performance. The size and configuration of buffer pools are critical for efficient memory utilization. Allocating an appropriate amount of memory to buffer pools based on workload characteristics and monitoring buffer pool hit ratios helps optimize performance. 

Here are some crucial considerations for configuring Db2 for z/OS buffer pools:

• Data Access Patterns: Understand the access patterns of your applications. Buffer pool configuration should align with the types of queries and transactions that are frequently executed. If your workload involves mostly random access, a larger buffer pool may be required. If it involves mostly sequential access, specific settings to favor sequential reads may be required.
• Buffer Pool Sizing: Determine the appropriate size for each buffer pool. Consider the amount of available memory on your system, the size of the database objects, and the anticipated workload. Larger buffer pools can improve performance, but it's essential to balance memory usage across other system components.
• Multiple Buffer Pools: Db2 for z/OS allows you to create multiple buffer pools. Consider segregating data based on access patterns or table spaces. For example, you could use separate buffer pools for frequently accessed tables and indexes versus less frequently accessed ones.
• Page Sizes: Db2 for z/OS supports various page sizes. Ensure that the buffer pool page size matches the page size of the objects being cached. Using the correct page size can reduce internal I/O operations and improve efficiency.
• Thresholds and Actions: Set appropriate thresholds for buffer pool thresholds, such as the high water mark and low water mark. Define actions to be taken when these thresholds are crossed, such as dynamically adjusting the buffer pool size or issuing alerts.
• Asynchronous I/O: Enable asynchronous I/O to allow Db2 to overlap I/O operations with processing. This can improve performance by reducing wait times caused by synchronous I/O operations.
• Monitor and Analyze: Regularly monitor buffer pool performance using Db2's monitoring tools. Analyze statistics and performance metrics to identify areas for improvement or potential issues.
• Buffer Pool Replacement Policies: Understand and configure the buffer pool replacement policies (e.g., LRU, MRU, FIFO) based on your workload characteristics. Different policies prioritize different data pages for retention in the buffer pool.
• Maintenance: Regularly review and adjust buffer pool configurations based on changing workloads, data growth, and hardware upgrades.

Also, Db2 for z/OS offers the following buffer pool tuning “knobs” that can be used to configure buffer pools to the type of processing they support:

• DWQT (deferred write threshold) –  expressed as a percentage of the virtual buffer pool that might be occupied by unavailable pages. When this threshold is reached Db2 will start to schedule write I/Os to externalize data. The default is 30%, which may be too high for many shops.

• VDWQT (vertical deferred write threshold) – basically the same as DWQT, but for individual data sets. The default is 10%, which once again is quite likely to be too high for many shops.

• VPSEQT (sequential steal threshold) – expressed as a percentage of the buffer pool that can be occupied by sequentially accessed pages. Tune buffer pools for sequential access such as scans and sorting by modifying VPSEQT to a larger value. The default is 80%.

• VPPSEQT (parallel sequential steal threshold) – the sequential steal threshold for parallel operations; the default value is 50%.

Additionally, adjusting parameters like page cleaning and prefetch size can further enhance buffer pool performance.

Remember that buffer pool configuration is not a one-size-fits-all approach. It requires a deep understanding of your specific workload, database design, and hardware capabilities. Regular monitoring and tuning are essential to maintain optimal performance over time.

There are other Memory settings that are nearly as critical as buffer pools, too. One example is the Db2 for z/OS EDM pool. EDM stands for Environmental Descriptor Manager. The EDM pool is used by Db2 to control programs as they execute. It will contain structures that house the access paths of the SQL statements for running programs. 

Actually, the EDM pool is a group that encompasses several other pools, all of which can be configured separately, including skeleton pools, DBD pool, sort pool, and RID pool. The RID pool is used by DB2 to sort RIDs (record identifiers) for List Prefetch, Multiple Index Access, and Hybrid Join access paths.  RID pool failures can cause performance degradation as alternate access paths are invoked, such as scans, and the CPU invested up to the point of the failure is wasted.

Db2 for z/OS also enables a Dynamic Statement Cache to be configured and tuned. It permits dynamic SQL prepare information to be cached in memory to optimize dynamic SQL. 

Another consideration to keep in mind is Query Parallelism, which refers to the ability of Db2 to execute multiple query tasks concurrently, leveraging multiple CPU cores. Enabling parallelism can significantly speed up query execution for resource-intensive workloads. The degree of parallelism, controlled by parameters like DFT_DEGREE and NUM_INIT_AGENTS, should be set based on the available hardware resources and workload characteristics. Careful tuning of these parameters ensures optimal parallel query execution without overloading the system.

Tuning Db2 Sort operations is also critical. During query execution sorting can consume significant memory resources. The sort heap is responsible for allocating memory for sort operations. Tuning the Db2 LUW SORTHEAP parameter to an appropriate size based on the complexity of sort operations and available memory helps avoid  excessive disk I/O and improves query performance. Regular monitoring and adjustment of SORTHEAP ensure efficient memory utilization for sort operations.

Statement Concentration is another configuration parameter to consider. It controls the consolidation of multiple SQL statements into a single unit of work. Enabling statement concentration reduces the overhead associated with parsing and optimizing individual statements, improving overall performance. By setting appropriate values for parameters like STMT_CONC and STMTHEAP, administrators can optimize statement concentration based on the workload and resource availability.

Connection and Maximum settings are also crucial to consider. For example, in Db2 for z/OS MAXDBAT sets the maximum number of database agents and Db2 LUW uses MAXAPPLS to define the maximum number of concurrent application connections. And an example of a setting that can control maximums is DSMAX (Db2 for z/OS) that can be set between 1 and 200000; it controls the maximum number of underlying data sets that Db2 can have open at any point.

It is important to note that tuning these configuration parameters should be done carefully and based on workload analysis and performance monitoring. The optimal settings may vary depending on the specific requirements and characteristics of the database environment. Regular monitoring of performance metrics, workload patterns, and system behavior is crucial to identify areas for tuning and ensure continued optimization.

Summary

In conclusion, configuration parameter tuning is a critical aspect of optimizing the performance of an IBM Db2 database. By understanding the impact of key parameters and adjusting them based on workload characteristics, administrators can fine-tune the system for optimal performance. 

We have only taken a high-level look at several configuration considerations and parameters in this blog post. But keep in mind that the configuration parameters and their settings contribute to an efficient and high-performing Db2 environment. Regular monitoring and tuning of these parameters help achieve optimal performance and responsiveness in the database system.

Top 10 Db2 Performance Tips - No. 6 Monitoring and Performance Metrics

Monitoring and measuring performance metrics are essential practices for maintaining and optimizing the performance of an IBM Db2 environment. By regularly monitoring and analyzing performance data, DBAs can identify bottlenecks, proactively address performance issues, and make informed decisions to improve overall system efficiency.

It is important to monitor and measure performance metrics in order to gain insights into the behavior of the applications and databases in use at your site. By examining their behavior and identifying areas for improvement, you can improve the overall satisfaction of your customers and end users. 

Performance metrics provide valuable information about resource utilization, query response times, disk I/O, CPU usage, and other critical aspects of database performance. By tracking these metrics over time, administrators can detect patterns, identify trends, and pinpoint potential performance bottlenecks.

A Strategy

The first part of any Db2 performance management strategy should be to provide a comprehensive approach to the monitoring of the Db2 subsystems operating at your site. This approach involves monitoring not only the threads accessing Db2 and the SQL they issue, but also the Db2 address spaces. You can accomplish this task in three ways:

• Batch reports run against DB2 trace records. While DB2 is running, you can activate traces that accumulate information, which can be used to monitor both the performance of the DB2 subsystem and the applications being run.

• Online access to DB2 trace information and DB2 control blocks. This type of monitoring also can provide information on DB2 and its subordinate applications.

• Sampling DB2 application programs as they run and analyzing which portions of the code use the most resources.

Using all three approaches can be a reasonable approach if you have analyzed the type of workloads in use and which types of monitoring are most effective for each. 
You also need to establish a strategy for your organization's tuning goals. And your tuning goals should be set using the discipline of service level management (SLM). A service level is a measure of operational behavior. SLM ensures applications behave accordingly by applying resources to those applications based on their importance to the organization. Depending on the needs of the organization, SLM can focus on availability, performance, or both. In terms of availability, the service level can be defined as “99.95% up time, during the hours of 9:00 AM to 10:00 PM on weekdays.” Of course, a service level can be more specific, stating “average response time for transactions will be two seconds or less for workloads of strategic external clients.”

For a service level agreement (SLA) to be successful, all of the parties involved must agree upon stated objectives for availability and performance. The end users must be satisfied with the performance of their applications, and the DBAs and technicians must be content with their ability to manage the system to the objectives. Compromise is essential to reach a useful SLA.

Furthermore, the objectives of the SLA must be defined and measurable. For example, in the earlier SLA you must define what a “strategic client” is and differentiate strategic from nonstrategic. Although doing this may be difficult, failing to do so can result in worthless SLAs that cannot be achieved.

In the end, if you do not identify service levels for each transaction, then you will always be managing to an unidentified requirement. Without a predefined and agreed upon SLA, how will the DBA and the end users know whether an application is performing adequately? Without SLAs, business users and DBAs might have different expectations, resulting in unsatisfied business executives and frustrated DBAs. Not a good situation.

Db2 Traces

One of the first types of performance metrics to consider is monitoring based on reading trace information. You can think of a Db2 trace as a window into the performance characteristics of aspects of your Db2 workload. Db2 traces record diagnostic information describing particular events. As Db2 operates, it writes trace information that can be read and analyzed to obtain performance information.

Db2 provides six types of traces, and each describes information about the Db2 environment:

• Accounting - Records performance information about the execution of DB2 application programs
• Audit - Provides information about DB2 DDL, security, ­utilities, and data modification
• Global - Provides information for the servicing of DB2
• Monitor - Records data useful for online monitoring of the DB2 subsystem and DB2 application programs
• Performance - Collects detailed data about DB2 events, enabling database and performance analysts to pinpoint the causes of performance problems
• Statistics - Records information regarding the DB2 subsystem’s use of resources

You can start Db2 traces in two ways: by specifying the appropriate DSNZPARMs at Db2 startup or by using the -START TRACE command to initiate specific traces when Db2 is already running.

Each trace is broken down further into classes, each of which provides information about aspects of that trace. Additional informatiohn about the classes for each type of trace can be found here, here, and here.

You should understand what type of information is traced and the approximate overhead of each trace class before starting any of the Db2 traces. Some traces are better left off until or unless a performance problem is occurring, at which point, the trace can be started to capture details about the situation. Others are better to have turned on before problems occur. Keep in mind that you will have some trace classes (and IFCIDs) that are always started, and other that are only used in emergencies.

What are IFCIDs?

Each trace class is associated with specific trace events known as Instrumentation Facility Component Identifier (IFCIDs), pronounced “if-kid.” An IFCID defines a record that represents a trace event. IFCIDs are the single smallest unit of tracing that can be invoked
by Db2.

In some cases, it can make sense to avoid activating trace classes altogether and start traces specifying only the IFCIDs needed. This way, you can reduce the overhead associated with tracing by recording only the trace events needed. 

There are several hundred different IFCIDs. Most IFCIDs contain data fields that describe events pertinent to the event being traced. Some IFCIDs have no data; instead they merely mark a specific time. Certain trace events of extended durations require a pair of IFCIDs: one for the beginning of the event and another for the end. These pairs enable the computation of elapsed times. Other trace events that are not as lengthy require only a single IFCID. 

You can find the IFCIDs associated with each trace class in the IBM Db2Command Reference manual in the section describing the -START TRACE command. But that manual does not describe the purpose of each IFCID. A list describing each IFCID can be found in the data set named  SDSNIVPD(DSNWMSGS), which is part of the Db2 installation. 

Db2 Performance Monitors

Several popular performance monitoring solutions are available for Db2 for z/OS. IBM’s Omegamon, BMC Software’s MainView, Broadcom’s Sysview, and Rocket Software's TMON are the leading performance monitors. When selecting a performance monitor, be sure to examine the online components as well as the batch reporting capabilities of the monitor. 

An online performance monitor is a tool that provides real-time reporting on Db2 performance statistics as Db2 operates. In contrast, a batch performance monitor reads previously generated trace records from an input data set. Most performance monitors today can be used to serve both purposes.

With online DB2 performance monitors, you can perform proactive performance management tasks. In other words, you can set up the monitor such that when it detects a problem it alerts a DBA and possibly takes actions on its own to resolve the problem.

The most common way to provide online performance monitoring capabilities is by online access to DB2 trace information in the MONITOR trace class. Some online DB2 performance monitors also provide direct access to Db2 performance data by reading the control blocks of the Db2 and application address spaces. This type of monitoring provides a “window” to up-to-the-minute performance statistics while Db2 runs. This information is important if a quick reaction to performance problems is required.

Most online Db2 performance monitors provide a menu-driven interface accessible from TSO or VTAM. It enables online performance monitors to start and stop traces as needed based on the menu options chosen by the user. Consequently, you can reduce overhead and diminish the learning curve involved in understanding Db2 traces and their correspondence to performance reports.

Following are some typical uses of online performance monitors. Many online performance monitors can establish effective exception-based monitoring. When specified performance thresholds are reached, triggers can offer notification and take action. For example, you could set a trigger when the number of lock suspensions for the TXN00002 plan is reached; when the trigger is activated, a message is sent to the console, and a batch report is generated to provide accounting detail information for the plan. You can set any number of triggers for many thresholds. Following are some examples of thresholds you might choose to set:

• When a buffer pool threshold is reached (PREFETCH DISABLED, DEFERRED WRITE THRESHOLD, or DM CRITICAL THRESHOLD).
• For critical transactions, when predefined performance objectives are not met. For example, if TXN00001 requires subsecond response time, set a trigger to notify a DBA when the transaction receives a class 1 accounting elapsed time exceeding 1 second by more than 25%.
• Many types of thresholds can be established. Most online monitors support this capability. As such, you can customize the thresholds for the needs of your DB2 environment.

Most online performance monitors can produce real-time EXPLAINs for long-running SQL statements. If an SQL statement takes a significant amount of time to process, an analyst can display the SQL statement as it executes and dynamically issue an EXPLAIN for the statement. Even as the statement executes, an understanding of why it takes so long to run can be achieved. A complete discussion of the EXPLAIN statement is provided in the next chapter.

Some online performance monitors provide historical trending. These monitors track performance statistics and store them in DB2 tables or in VSAM files with a timestamp. They also provide the capability to query these stores of performance data to assist in the following:

• Analyzing recent history. Most SQL statements execute quickly, making difficult the job of capturing and displaying information about the SQL statement as it executes. However, you might not want to wait until the SMF data is available to run a batch report. Quick access to recent past-performance data in these external data stores provides a type of online monitoring that is as close to real time as is usually needed.
• Determining performance trends, such as a transaction steadily increasing in its CPU consumption or elapsed time.
• Performing capacity planning based on a snapshot of the recent performance of DB2 applications.

Some monitors also run when Db2 is down to provide access to the historical data accumulated by the monitor.

A final benefit of online DB2 performance monitors is their capability to interface with other z/OS monitors for IMS/TM, CICS, z/OS, or VTAM. This way, an analyst gets a view of the entire spectrum of system performance. Understanding and analyzing the data from each of these monitors, however, requires a different skill. Quite often, one person cannot master all these monitors

Conclusion 

Although this blog entry was brief, and there are many additional aspects to Db2 performance monitoring, monitoring and measuring performance metrics should be viewed as a vital requirement for all organizations using Db2 databases. By collecting and analyzing performance data, DBAs can detect performance bottlenecks, identify areas for improvement, and make informed decisions to enhance overall system efficiency.