You know how it feels when you find a great sale on a new TV, buy without measuring your living room and learn that it won’t comfortably fit anywhere. Or buy that new car and realize that its real pain to park, and it barely even fits in your garage.
Nobody wants to feel like they should have known better before making a purchase decision. Often, it’s more of an inconvenience than a real mistake: pants and TVs can be returned. Unfortunately, products such as cars and data-storage systems don’t offer this luxury.
To maintain a data center that balances high availability, consistent performance and low costs, you first need to understand your organization’s IT infrastructure and how workloads are interacting with it before buying new technology. This knowledge comes from acquiring and analyzing I/O workload profiles and testing them under the range of conditions that your company could face. Below are three tips for successfully using I/O profiling to better understand storage deployments, get storage costs under control and achieve the performance your system needs—and deserves.
1. Use synthetic texts as contextual clues.
To get a complete picture of your storage environment, you must use the best workload I/O profiling tools. That means going beyond some of the more common synthetic testing options.
Synthetic benchmark testing is available from a few organizations, such as the Storage Performance Council (SPC). These tests offer some basic insight and can be a useful first step in getting a storage project off the ground. But they are limited because they really only represent one or a small number of generic workloads. These benchmarks are run under ideal circumstances by the vendors, so they don’t necessarily represent a realistic view of a storage environment or of your specific application workload profiles.
Use synthetic benchmarks to establish baseline comparative results, but don’t rely on them for purchasing or deployment decisions. When your infrastructure and business are at stake, run tests that are specific to your company, on the basis of its particular needs.
2. Customize testing scenarios to make informed storage decisions.
I/O profiling helps storage admins use live production workload data in test environments and thereby gain insights into systems already in use. Your team can closely mirror the real conditions of your IT environment, build an accurate workload profile and introduce control and real-world consequences into every test—in addition to eliminating vendor bias.
You can also use workload I/O profiling to track specific metrics, including the following:
- Scale: The number of transactions that a storage solution must handle rarely stays the same over time. Testing what happens as transactions increase or under worst-case loading parameters can help your storage team determine the limits of your current hardware.
- Failure: It’s an unfortunate fact of life: storage technology fails. So, testing what happens during failure, and how performance is affected as data is being rebuilt afterwards, is an important part of any test plan.
- Media: Many storage solutions today are combinations of different technologies. Testing different media configurations against your environment’s conditions can help your company purchase the right amount, and right combination, of storage and thus prevent costly mistakes such as over- and underbuying.
- Time: It’s rare that a storage system is taxed the exact same amount at all times. I/O profiling allows you to vary profiles to account for different usage rates, such as varying block sizes and file sizes.
Using these metrics, literally hundreds of scenario combinations are possible. Once you have them, you can compare them to different storage solutions, gaining critical information about how each option will perform.
3. React in real time and change your testing strategy on the fly.
The scenarios we’ve discussed so far involve purchasing new storage equipment. But you must take another factor into account: what if you need to change or upgrade your current storage array? This situation may arise for several reasons.
- Updates to system operating software and microcode: Vendors are constantly updating their code in an attempt to make storage faster, more functional and more efficient. But anyone who’s dealt with a software update knows that it can have unintended consequences. By using testing hardware running recent updates, I/O profiling and workload generators can shake out potential problems before they affect your actual systems.
- Media upgrades: As with software updates, physical storage media are subject to the cycle of firmware upgrades. Each upgrade has the potential to change and disrupt how a device behaves. I/O profiling can identify potential changes in device function before it affects your company’s live storage.
- Additional features: Storage products are constantly adding new features—think deduplication and compression facilities. These features have great economic benefits in dollars per gigabyte, but they also may degrade performance. Storage I/O performance testing can help you be sure. In a test environment, you can run simulated production workloads against the systems with new features turned on or off, then see what works best for your particular situation.
Virtually all organizations rely on storage solutions to keep business running. Although new features and technology have made storage more reliable and efficient, they haven’t made it any less complex. If anything, storage complexity has grown over the past 20 years, as businesses that once had terabytes of information now have petabytes.
Making investments in storage based on estimates about your company’s needs and environmental requirements is unacceptable—even if those guesses are educated. Every year, organizations spend up to 25 percent of IT budgets on data storage. Being over- or underprovisioned by 50% to 100% on storage owing to an inaccurate understanding of workload performance requirements can easily happen. This situation can certainly be expensive for IT organizations and the company as whole.
When it’s time to act on your research and purchase storage, detailed workload I/O profiling based on your production environment is critical. It can help storage architects and engineers align their buying decisions with the top issues facing the industry today—which include performance, capital expenses, scale and manageability. Storage performance testing and analytics also prepares your team to manage the many changes your storage environment will undergo as your company evolves, and to increase the value of your infrastructure over time.
About the Author
Len Rosenthal is chief marketing officer at Virtual Instruments. With more than 30 years of experience at leading public and privately held IT-infrastructure companies, Len has held executive and senior positions at Load DynamiX, Panasas, QLogic, Inktomi, HP and more.