If your enterprise system has gotten more complex over the last few years, with new applications and users, it may be time to revisit your hardware configuration to optimize performance.
Companies benefit when they practice taking a holistic view of enterprise systems every two years or so, says Jaqui Lynch, a solutions architect and independent consultant. “In the past, systems may have been set up using best practices at the time, but these may not be the best practices anymore. Technologies change and businesses may need to adapt,” explains Lynch.
Getting a Better ROI
When clients evaluate what they have, they often find cost savings can be realized from moving to the latest hardware, according to Lynch. “Once a system comes out of its three-year maintenance agreement, for example, maintenance costs usually increase. And maintenance is a huge expense. Upgrading the hardware could eliminate a large chunk of that expense,” she adds.
“In the past, systems may have been set up using best practices at the time, but these may not be the best practices anymore. Technologies change and businesses may need to adapt.”
—Jaqui Lynch, solutions architect and independent consultant
For example, some enterprises may determine they’re paying too much for maintenance with an older model POWER6* system. They may be likely to realize a good business case to upgrade to POWER8*, because companies on the newer POWER* releases don’t have to deal with the maintenance issues or the higher power and cooling costs associated with the older technologies.
IBM clients can also realize software benefits. With POWER8, a company can run AIX* 7.2, the newest version, which lets them take advantage of advanced features, such as flash cache, that aren’t available with older POWER configurations.
Though obvious improvements exist for implementing the latest technologies, organizations don’t always need the latest OS to experience upgrade benefits. It’s possible, though not optimal, to run AIX 5.3 on POWER8, for example.
Some IT professionals may be frustrated if it seems the standard answer to resolve a service interruption or performance degradation is to upgrade their firmware or OS.
However, sometimes an upgrade may be all an organization needs to fix a persistent issue. “I once had a client with a hardware problem that was really killing their performance,” Lynch says.
“I looked at their firmware and realized it was eight levels back. Three levels above where they were, the problem they had was addressed.” Once the client made the upgrade, the problem went away and performance improved significantly.
When an organization waits too long to address a minor issue, however, it can lead to a major problem. For example, a previous firmware issue could have been resolved with an upgrade to level v810. But v810 is now withdrawn, Lynch says. A company would need to upgrade to v860 to resolve the issue. “The company wouldn’t just be looking at the routine firmware update that was originally required. Instead, the company would have to change releases, and that’s disruptive,” she adds.
It’s important to make sure you understand how an upgrade would benefit the enterprise before making any decisions about implementation. But keep in mind that it’s also important to stay as current as possible to optimize technology systems. Not every organization needs to be on the leading edge of new enterprise technologies. “But being two to three years behind is unacceptable and risky,” explains Lynch.
Documenting Current Technologies
It’s important for organizations to document their current configurations and then determine if improvements can be made. Lynch recommends using the HMC Scanner tool. In two minutes, the scanner creates a spreadsheet that contains the configuration of servers and LPARs.
“You want to look at a system that’s set up and evaluate it in its current state, and then plan for how to improve it, Lynch says. It’s also a good idea to consider enterprise needs over the next one year, two years and beyond, and plan accordingly. Once an IT professional has analyzed the business system, they can begin to identify adjustments that can be made to enhance performance.
“I often ask a client about what’s more important in their environment. Is I/O more important than network or CPU, for instance?” Lynch says. It’s important to remember that the slots on the server have different priorities, so card placement also directly affects performance, she adds.
Tuning the Network
Another area for administrators to review is network tunables. Many tunables become restricted over time, but not everyone knows they should stop using them. “There are a lot of bad practices out there, especially for those with older systems. There were tunables you set with AIX 5, for example, that you no longer set with AIX 6 or AIX 7,” explains Lynch.
Administrators can avoid this problem by making routine reviews of their configurations, to ensure their tunable guidelines are current.
Even current settings for tunables aren’t necessarily right for larger enterprises. “The default tunables for the network are set for 100 MB network at best, so some network tunables have to be changed. If you don’t, IBM will try to set some of them on the adapter, depending on what version of the operating system you’re on, but they’re not always set for the best settings for a 10 GB network,” says Lynch. So it’s important for IT professionals to educate themselves on the tunables for their systems.
Without setting up tunables, some administrators will run into issues such as buffer shortages, but steps can be taken, based on running commands, to evaluate performance.
Administrators can also tweak the location of assets to improve performance. “I tell clients that if they plan to bring in new cards, for example, maybe we should change the location of the cards to get slightly better performance. Maybe their workloads have grown, but they’re still using an old 4 GB host bus adapter (HBA). They might need to move to 8 GB or 16 GB HBAs to get better performance,” Lynch explains.
In addition, organizations must evaluate their CPU and I/O needs. CPU advancements in speed over the last 10 years mean companies can achieve greater performance. And with virtual I/O servers (VIOS), upgrades enable organizations to take advantage of the newest performance and security patches.
Get Ready to Update
For client systems, Lynch recommends making updates in this order:
Ensure the HMC machine code level is current
Address firmware
Review VIO server
Review the OS
“Check the ‘read me’s’ first though to make sure the combination you put together works,” Lynch advises. “You don’t want to go to a firmware level that won’t work with the OS you have, or go to an HMC machine code level that won’t work with the server level you have.”
It’s important to make sure the combination in each step works together. You don’t want to make any changes without a planning document ahead of time. “Quite often you’re making updates at 2 a.m. or 3 a.m., or on emergency basis. Your team may be tired or stressed, so you want to protect against making mistakes. That’s why planning is critical,” adds Lynch.
“Next, I figure out what I think I want to change, and what I think I’m going to get from the change,” Lynch says. “Then I make the change, measure the results and create the next steps.”
It’s usually difficult to come up with a window to perform the required maintenance, so it’s tempting to fit as much into that window as you can. But that’s almost a guarantee that something will break. Even worse, if something does break, it will be hard to find the cause of the malfunction. A better practice is to find times to address network, storage and systems updates separately.
Companies will also need to make sure they have a valid hardware maintenance agreement to be able to download and install firmware. Likewise, they will need a valid software maintenance agreement for software.
Advanced Features
With newer hardware, corporations can take advantage of new technologies to meet corporate needs. For example, an upgrade to POWER7* or later means the enterprise can run AIX V7.2, which includes server-based flash cache.
Flash cache is designed to boost performance for reads and provide higher throughput with lower latency. It can help reduce response times, improve transaction rates and reduce some applications’ RAM requirements, according to IBM. It’s also beneficial to the performance and scaling of the storage area network (SAN) in that it offloads I/O operations from the SAN. The AIX OS stores a copy of hot data (the OS determines which data is hot) in the flash cache. Writes continue on to the SAN, but reads are done from the flash cache. No changes are required to the application, as the caching is transparent.
In addition to flash cache, you can also use AIX live update for interim fixes in v7.2. New features also include using cluster aware AIX for automation and the dynamic system optimizer, which automatically adjusts some system settings to maximize system efficiency. The DSO is included in the base AIX 7.2.0 and later and is no longer a stand-alone feature.
Routine Review
To optimize system performance, it’s imperative to routinely review enterprise hardware. By fine-tuning the configurations and taking advantage of the newer technologies, companies can make sure they have the capacity to handle increasingly complex business needs.