We had the pleasure of recently presenting at Storage Field Day 23 in San Jose. This was the first appearance by Fungible, and a kind of “coming out party” for our Fungible Storage platform to a wider influencer audience. While we have been busy over the last few years in developing Fungible DPU based technologies, the Fungible Storage Cluster is a foundation for organizations focused on the “cloudification” of the data center, and specifically on delivering storage services to modern applications.
While public cloud has taken a prominent role for organizations looking for agility, scale, and elasticity of resources, customers who have not made the transition to public cloud still struggle to allocate storage services that match the needs of their applications on a granular level. We see that the applications have changed, and are being architected to take advantage of cloud-native architectures, but for many, the existing legacy storage platforms struggle to keep up with the ever-changing requirements where scale, performance, and efficiency are of high value. Ultimately the goal should be to deliver storage as a service tailored to each specific workload or business outcome.
Introducing Fungible Storage Cluster 4.0 and Fungible Storage Initiator
Fungible Storage Cluster is a high-performance, scale-out, block storage platform that allows customers to match the proper storage services to their individual applications and workloads by offering per-volume granularity on services such as durability, compression, encryption, and more. All of this is backed by ultra-low latency by utilizing NVMe/TCP as the storage transportation vehicle. This approach offers increased performance, latency reduction, and a decrease in CPU utilization for common data-centric tasks.
With the 4.0 release of FSC (Fungible Storage Cluster) coupled with Fungible Storage Initiator, customers now have the option of providing storage as a service, tailored to the specific needs of the application while offloading the complex data-centric tasks that consume significant CPU resources.
The 4.0 release adds a few new features to help support organizations looking to cloudify their storage needs.
The RF1 volume is an industry-first feature that combines the performance of a raw NVMe storage volume with the benefits of encryption and compression data services which have traditionally been only available to fully durable volumes. Now storage volumes can be created to offer the efficiency and security benefits at the storage layer and allow customers to make the most of their storage investment while also providing significant performance improvements.
Erasure Coded 2+1 Volumes
Erasure coded storage offers superior data resilience and capacity savings, but has traditionally been difficult to implement at scale while preserving performance due to the complex mathematics involved that require significant CPU overhead to calculate. Given that the Fungible Storage Cluster is a Fungible DPU powered storage array, Erasure Coding is handled in the silicon of the DPU instead of the CPU, and this offers a nearly 10X improvement in performance and efficiency. With FSC 4.0, we are introducing EC 2+1, which allows for greater durability at a smaller entry point, in addition to supporting EC 4+2 and EC 8+2 volumes that we already offered customers
Fungible Accelerator with Storage Initiation
NVMe/TCP is quickly becoming the de facto storage transport of modern applications that need performance and scale. Leveraging the power of the Fungible S1 DPU inside the host compute system, customers now have the option of offloading NVMe/TCP storage operations from the CPU to the Fungible DPU. This offers several major advantages vs doing this work within the native operating system. CPU utilization is significantly decreased, saving up to 50% of CPU cores while performance is increased by 50% over software-based initiation. The reduction in CPU core utilization results in more CPU being available to run applications and workloads instead of servicing host operating system IO needs. The increase in performance results in faster time to data.
Applications have three main storage characteristics that need to be addressed. Performance, Capacity, and Data Services, and customers have been forced into a one-size-fits-all approach to solving their application-focused storage needs. So it begs the question, what if you could allocate the right amount of capacity, performance, and data services at the application layer, and at the same time, significantly reduce the CPU overhead required to do so?
Offering a Granular Approach to Matching Application Requirements and Storage Services
Not all applications have the same storage requirements. This presents a challenge when we look at how many legacy storage products have been designed. Oftentimes customers are given a few knobs to twist to address application requirements, or they are given so many that they need months of education and a PhD in the product’s capabilities to properly address their needs. With Fungible Storage Cluster, we have looked to simplify this process, and give customers control, simplicity, and the confidence to deliver storage as a service to their datacenter.
Fundamental to this approach is the ability to provide granular data service, data resilience, capacity, and performance to match individual application requirements. Some applications need raw performance for transient or ephemeral data, and therefore gain no benefit to additional data services such as compression or encryption. Others may need higher levels of data resilience, moderate capacity, and performance. And some may need encryption, compression, and Quality of Service to properly align the application and storage services. With FSC 4.0, customers get to properly allocate the storage services to match the needs of the application and take a cloud-focused approach to delivering storage services to their organization.
Obviously, a blog post can only go so far, so we ask you to go watch the Storage Field Day 23 presentation here to see how you can cloudify your storage, and watch a demonstration of the Fungible Storage Cluster.