Marvell DragonFly - The I/O acceleration solution that solves the write cliff problem the right way

Back-end data centres increasingly need hyper-scale infrastructure to store, process, and deliver accelerating data demands within ever-shrinking time windows. While CPUs have ridden Moore's law, hard disk drives have not—for database, virtualization, collaboration, and big data applications —the result is a growing storage I/O and latency bottleneck that jeopardizes the service requirements for the next generation of cloud computing.



To meet growing cloud application demands and cut infrastructure costs, public and enterprise clouds are increasingly using virtual machines to consolidate applications onto few servers. But in solving this problem, virtualization has inadvertently exacerbated the storage bottleneck: local DAS and networked storage simply cannot keep up with the flood of random write and read I/O requests in a cost-effective manner. System vendors have tried to address the bottleneck by using SSDs but SSDs have shown dramatic I/O and latency variations under sustained write conditions.



Marvell addresses this problem head-on with the Dragonfly platform, a revolutionary cloud computing I/O acceleration technology platform centred around large-capacity NVRAM. Leveraging intelligent read, write-back caching and the industry’s first write caching on tiered non-volatile memory, DragonFly enables the creation of a next-generation cloud -optimized data centre architecture where data is automatically cached or stored in a low-latency, high-bandwidth NVRAM or NVCACHE card slotted in standard rackmount servers.