BMC Updates iSeries Management and Planning Products
April 5, 2005 Alex Woodie
With 331 different products for every major platform, it’s sometimes easy to forget that BMC Software is actively developing OS/400 software, too. In fact, the Houston, Texas, software giant has some pretty sophisticated products, namely PATROL for iSeries–Predict, a capacity planning tool, which IBM sells with its own part number, as well as PATROL for iSeries, a systems management product. BMC recently introduced new releases of both of these products.
If you’re looking for “lights out” automation of your multi-platform data center, BMC’s PATROL series should be on your short list. The PATROL series allows administrators to monitor many servers from a single console, as well as build in the capability for servers to take corrective action for simple types of problems. IBM calls this “autonomic computing,” but the folks at BMC just call it “automating repetitive tasks.”
Mark Banta, the lead developer for PATROL for iSeries, says the software provides automation capabilities that aren’t found in OS/400. “It’s all automated and cyclical. You don’t have to remember to run a report, once the template has been created,” he says. “We want to get to where 99 percent of what customers need to have taken care of, gets taken care of automatically, and provide drill-down analysis for the other 1 percent.” Judging from a recent demo by Banta, that’s a fairly accurate statement. The product’s functionality is deep and multifaceted, although it’s somewhat complex, which reflects its enterprise roots.
Administrators interact with PATROL for iSeries from the same Windows, Web, or Unix-based console they use for other PATROL products for MVS, Unix, Windows, Linux, Domino, and WebSphereMQ, although the Windows console is by far the most popular console, Banta says. There is no greenscreen option, and BMC customers don’t ask for one.
PATROL for iSeries installs an agent to collect performance and state data from the OS/400 server. The product uses the concept of Application Classes to segment the OS/400-specific operational areas it monitors. There are 15 Application Classes providing real-time data and graphing for such things as processor utilization and response times, QSYSOPR and job message queues, ASPs and main memory usage, and use of library objects and failed logons.
Each of these Application Classes has a corresponding Knowledge Module that provides the best out-of-the-box settings based on BMC’s iSeries knowledge. Administrators can customize the Knowledge Modules with their own thresholds and commands, through several mouse clicks in the main tree view and subsequent dialog boxes. For example, if DASD usage goes past 80 percent, the product can be set up to immediately send an email to the head administrator alerting him of the condition. It can also be set up to cascade additional messages to other operators if the condition is not responded to.
In late March, BMC introduced a new release of PATROL for iSeries, version 3.6, that includes new Application Classes for monitoring history logs, the IFS system, and the state of the Internet connection on the iSeries server.
The capability to monitor the QHST file allows users to detect messages arriving that match specified sets of selection criteria, including message IDs, minimum severity, message types, and message text. The matching messages are counted, allowing thresholds and recovery actions to be configured for the arrival of important messages, and made available in ad hoc and drill-down reports. The product provides out-of-the-box tracking of abnormally ended jobs and system value changes.
Support for IFS monitoring in version 3.6 brings some semblance of control over Windows and Unix applications that may be running on the iSeries server, and the degree to which they’re consuming iSeries system resources. While the monitoring provides more information about what’s happening in the IFS than it previously did, there is still not the depth and quality of system information available for tracking native OS/400 objects, commands, and concepts.
Finally, version 3.6 adds NETSTAT Monitoring, which monitors TCP/IP network status for an iSeries. Information on each of the major NETSTAT areas (interfaces, routes, and connections) is available and can be independently turned off or on, BMC says.
Lastly, the security architecture was also rebuilt with version 3.6 to support the SSL structure used by other PATROL products. Licenses for PATROL for iSeries are tier-based and cost between $1,350 and $15,190 per monitored server. Users must also purchase a PATROL Console, which cost $3,000 each.
PATROL for iSeries–Predict
For years, BMC developed the Best/1 capacity planning tool, which IBM sold as its own product. With OS/400 V5R2, that changed, and IBM started offering a new capacity planning tool, called PATROL for iSeries–Predict (or iPredict, which is a less awkward name, although not an official one). This product was modeled after a similar Windows planning tool that BMC had developed, and since that product was at version 7, iPredict debuted as a version 7.1 release. In late February, BMC started shipping the second major release of the product, version 7.1.1.
iPredict is used primarily by IBM business partners and iSeries resellers to accurately size new boxes, upgrades, and server consolidation projects. The software is also used by some larger, multiplatform enterprise shops, for the same reason they would choose the PATROL products–the capacity planning process and interface are similar across all of BMC’s Predict products.
Unlike the PATROL for iSeries product, iPredict utilizes a native 5250 interface as well as a Windows-based console (which is included in the price of the package). The greenscreen interface is used to select the various performance data (which is collected by IBM’s data collection service) that is sent to the Windows console for analysis.
The basis of the planning process in iPredict is the Job Classification Member, which is used to create a usage model that is based on actual workloads, preferably collected during peak hours. “When you’re doing capacity planning, you don’t want to deal with the system as a whole,” says Jerik Tornheim, BMC’s development manager for the product. “You want to separate the work you’re doing into workloads and these workloads should have some sort of meaning for you.” Examples of these workloads could be printing, WebSphere, compiling, and so forth. Tornheim’s demo had a workload titled “DataMirror.”
Once a certain time-slice of performance data for a certain application, user ID, subsystem, jobs or other category has been collected and downloaded to the console, iPredict users can do all sorts of things to get an accurate picture of their use of various iSeries resources. The most important metrics to watch (and therefore to base one’s upgrade or purchase decisions on) are the use of batch and interactive processor capacity, I/O reads, transaction rates, and response times, although there are many other metrics available in this sophisticated tool.
Version 7.1.1 Enhancements
Before doing “What If” modeling to see how certain workloads would run on different iSeries systems, or how upgrading the memory, DASD, disk arms, or I/O processors will change the utilization of resources, an iPredict user needs to do a little calibrating to account for some things that IBM doesn’t collect. This is one of the enhancements in version 7.1.1. “There are all kinds of things, that IBM doesn’t pick up in its data,” but affect response times, such as tape loads, Tornheim says.
Another enhancement in the new release is the capability to model the performance of partitions with shared processors. This product functions on a partition basis, because that is how IBM collects its performance data. Obviously, if a partition is only using 40 percent of a given processor, that partition’s processor utilization rate is going to be different than if it were using 80 percent of that processor.
Version 7.1.1 also accounts for the impact of cache rebuilding in shared processor configurations. Since there is a small–but potentially significant–penalty for sharing a processor among partitions due to the requirement to clear and rebuild the processor cache as it shifts workloads, this needs to be accurately reflected in any capacity planning scenario. “There is generally a 5 percent overhead for each partition that can fit into the processors,” Tornheim says. “Things are not linear.”
The new release also better supports the modeling of constrained, or single-threaded, workloads, where the number of concurrent transactions or threads is limited. This option is available under the Advanced Tuning menu selection.
Finally, BMC has improved the way users select their disk and RAID configurations with this release. It has also delivered new Model Summary and Period-Model Summary reports that include in a single report the predicted values for a number of variables, including the computer name, the CPU model, the memory size, processor information, the number of disks, the average disk utilization, and the total I/O count. The CPU utilization, queue length, throughput, and average response times are also included in these reports, for both interactive and batch.
PATROL for iSeries–Predict version 7.1.1 is available now. Pricing is tier-based and ranges from $350 to $12,000. For more information, visit www.bmc.com.