We live in a data-driven world. As data continues to grow exponentially so does the importance of databases, which help organizations deal with the chaos. At the same time, organizations are looking for ways to reduce cost and simplify the infrastructure that power these databases. And that’s where hyperconverged infrastructure comes in.
Hyperconverged infrastructure has been a great technology for use cases such as virtual desktop infrastructure (VDI) and small-scale, lightly-loaded applications. Organizations have enjoyed the significant operational efficiency and cost benefits. But what about enterprise database systems and associated applications that demand high levels of I/O and have a difficult time scaling beyond a single host? How do you take advantage of hyperconverged infrastructure for these?
Boosting Hyperconvergence Performance
With hyperconverged infrastructure systems, there is an interesting performance characteristic imposed due to the architecture: the same nodes that generate I/O load have to handle that I/O load. Now even though there are a number of areas that can impact overall workload performance, application performance bottlenecks are typically I/O driven.
The most popular way to gain an instant speed boost is by replacing spinning disks with shiny new flash devices. You can go all flash, but the media type is often not the problem. And don’t forget a 4TB SSD can cost you 10x more than a 4TB enterprise hard disk drive. Also, keep in mind that the storage takes data only as quickly as it’s received from the CPU (or from the host). If you’ve got bottlenecks that are not on the storage device, but are somewhere else in the system, you still have to figure out ways on how to clear those bottlenecks before you actually can get all the benefits you expect to see even if you go all flash. There’s probably nothing worse than to go all flash and only see a modest performance increase because there was some other type of bottleneck in the environment.
Scaling up is another way to deal with workload performance challenges, but by itself it’s not sufficient. You can add all the RAM, flash, and CPU in the world and it might still not be enough. Not all workloads can take advantage of the additional horsepower due to bottlenecks elsewhere in the system.
Parallel I/O for the Win
Ultimately, what you need to worry about is latency – nothing is a workload killer like latency. All the other metrics in your system might look fine on your end, but if the end user is experiencing a lag, it’s going to be a problem. The results of poor latency are pretty significant and this is where software can help.
DataCore can help you solve some of the performance challenges with databases and associated applications by sending data down to the storage device in the node via multiple channels using our Parallel I/O technology, which is included in our Hyperconverged Virtual SAN software. This technology helps write data far more quickly and get far more performance out of the overall solution. But, don’t take my word for it – hear it directly from one of our customers in this webinar, A Case Study on Improving SQL Server Performance 20x with Hyperconverged.
To learn more about our hyperconverged and SDS solutions, request a 15-minute live demo today or download the whitepaper, Leveraging Hyperconvergence for Databases and Tier 1 Enterprise Applications!
