Although the future of enterprise storage is almost inevitably flash, a significant percentage of early adopters report that the newer technology fails to deliver on its promises. Indeed, as many as 16% report that their flash arrays have performed below expectations.
However, the fault does not lie with flash technology itself. Often the poor performance can be traced back to the application in use.
Legacy apps – the unavoidable bottleneck
The reality is that very few older applications are able to take full advantage of the speed and capabilities of flash drives. The underlying operating system may be lightning quick, but the older code base of your legacy apps “under delivers” on expectation.
There are two key reasons why deploying all flash arrays (AFAs) makes little sense for legacy applications:
1. Limited support for parallelism
The true power of flash drives lies in the NVMe and PCIe interfaces which support massive bandwidth throughput. Even more importantly, these newer interfaces support large numbers of simultaneous I/O requests.
Legacy applications built in the era of SATA and SAS cannot take advantage of this parallelism because older drive technologies only allow for a limited number of storage request queues. Once those queues are full, the application reaches its I/O performance ceiling.
On the hardware front, connecting flash drives to SAS/SATA backplanes results in minimal performance improvement because the interfaces still cannot support parallelism.
2. Existing infrastructure must be redesigned
Even where AFAs are deployed on NVMe backbones, the rest of your data centre is still designed around optimizing the performance of spinning disk storage. Over-provisioning and caching can increase SAS performance, but the additional networking overheads reduce the gains from NVMe.
Obviously adopting AFAs provides an opportunity to free up space in your data centres – but it also adds significantly to the cost of deployment. And without recoding your legacy applications, the performance gains will still underwhelm.
Maintaining legacy hardware for legacy applications
In many cases, it makes better financial sense to retain legacy applications of legacy hardware. Not only will these applications continue to perform optimally, but your business has the option of making savings elsewhere – such as the adoption of third party support and maintenance services.
The money saved by re-using legacy and post warranty hardware can then be reinvested in NVMe AFA storage for new applications. Or even to cover the costs of redesigning legacy apps so that they are finally able to take advantage of increased bandwidth and parallelism.
To learn more about keeping your legacy storage systems running, and the savings your business could make, please get in touch.
Although the future of enterprise storage is almost inevitably flash, a significant percentage of early adopters report that the newer technology fails to deliver on its promises. Indeed, as many as 16% report that their flash arrays have performed below expectations. However, the fault does not lie with flash technology itself. Often the poor performance can be traced back to the application in use. Legacy apps – the unavoidable bottleneck The reality is that very few older applications are able to take full advantage of the speed and capabilities of flash drives. The underlying operating system may be lightning quick, but the older code base of your legacy apps “under delivers” on expectation. There are two key reasons why deploying all flash arrays (AFAs) makes little sense for legacy applications: 1. Limited support for parallelism The true power of flash drives lies in the NVMe and PCIe interfaces which support massive bandwidth throughput. Even more importantly, these newer interfaces support large numbers of simultaneous I/O requests. Legacy applications built in the era of SATA and SAS cannot take advantage of this parallelism because older drive technologies only allow for a limited number of storage request queues. Once those queues are full, the application reaches its I/O performance ceiling. On the hardware front, connecting flash drives to SAS/SATA backplanes results in minimal performance improvement because the interfaces still cannot support parallelism. 2. Existing infrastructure must be redesigned Even where AFAs are deployed on NVMe backbones, the rest of your data centre is still designed around optimizing the performance of spinning disk storage. Over-provisioning and caching can increase SAS performance, but the additional networking overheads reduce the gains from NVMe. Obviously adopting AFAs provides an opportunity to free up space in your data centres – but it also adds significantly to the cost of deployment. And without recoding your legacy applications, the performance gains will still underwhelm. Maintaining legacy hardware for legacy applications In many cases, it makes better financial sense to retain legacy applications of legacy hardware. Not only will these applications continue to perform optimally, but your business has the option of making savings elsewhere – such as the adoption of third party support and maintenance services. The money saved by re-using legacy and post warranty hardware can then be reinvested in NVMe AFA storage for new applications. Or even to cover the costs of redesigning legacy apps so that they are finally able to take advantage of increased bandwidth and parallelism. To learn more about keeping your legacy storage systems running, and the savings your business could make, please get in touch.