By Jim Bahn, Senior Director of Product Marketing
Flash doesn’t solve all performance or availability problems. Below are six reasons those issues can outlast a flash installation, and how to manage them over time.
- Networked storage is a complex beast. If you’ve put flash on your laptop, you might have solved most of your performance problems. But according to analysts, our experience over the past eight years and that of major storage vendors (like Nimble and NetApp) who keep track of things like this, less than 50 percent of performance problems reported by customers wind up being due to the storage array. Poorly behaving apps, hot spots, oversubscribed ports, cache configuration issues, underprovisioned server CPU or memory, imbalanced and bully VMs, rogue clients, suboptimal queue depth settings, out-of-spec hardware components that are about to fail, and dozens of other factors all make up the other half of performance and availability issues. How do you find or avoid those problems? You need to monitor your infrastructure.
- Flash doesn’t help with availability even when it does help with performance. If your multi-pathing fails or fails to handle your workload, flash won’t help.
- Once you deploy flash, you still need to determine the best apps to put on it. Are you planning to put every app on flash? Probably not. Which apps will benefit and which will not? What are the cost and performance tradeoffs? Our Load DynamiX Enterprise customers are answering this question intelligently – with data on their own requirements.
- After deploying flash, you still can benefit from optimizing your configurations. How should you set queue depths? How should you load balance ports or VMs? With proper load balancing, you can extend the life of your current configuration by quarters or even years. We have customers who’ve seen payback in less than a year simply by better utilizing their existing infrastructure.
- If you are running flash, and you’re getting good performance, at what point of your workload growth will latency begin to creep back up? In our testing, we’ve seen competitive arrays that deliver 1ms latency but fail to break 10ms latency after the workload increases. What would it be worth if you could test the effect of increasing workloads ahead of time and know, today, exactly when your performance will begin to exceed SLAs?
- Once you deploy flash, you might want to test new firmware releases before deploying into production, especially if you are running mission critical apps. Flash performance can be taken to its knees by a poorly applied software update. Again, our customers claim that Load DynamiX Enterprise is the best solution for change validation.
Hundreds of our customers have moved to flash, but they use our testing and monitoring solutions more than ever, as more pressure is being applied from greater VM density, and more financial pressure is being applied to IT budgets.
Want to learn more about how applications and infrastructure can work better together? Contact us and let’s have a conversation.