Virtualized & Accelerated Cloud Infrastructure for Big Data workloads
Recorded Jul 25 12:00 am
About this Webinar
Whether for private clouds, public clouds or a multi-user environment, IT operators require an infrastructure solution that can instantly spin-up data-intensive workloads and that can scale to support multiple big data workloads or tenants. Attala Systems and Intel have partnered to create a solution that delivers composable storage with local NVMe performance for big data workloads like Apache Hadoop and Spark. The performance available to direct-attached PCIe flash is welcome, but cloud total cost of ownership (TCO) is better when storage is shared to many hosts while offering the elasticity to meet dynamic tenant demands. The maturing of the NVMe over Fabrics (NVMe-oF) specification enables local NVMe performance over a network. With remote direct memory access (RDMA), NVMe actions and data can be transported over a fabric such as Ethernet. This brief illustrates how a complex big data analytics workload is well served by Attala Composable Storage with Intel® 3D NAND SSDs. This workload—which once struggled to scale to support concurrent customers when serviced by an iSCSI appliance or enterprise storage area network (SAN), is shown to match local NVMe performance and scale independently for two customers with no degradation in application runtime when using the NVMe-oF target.
Taufik Ma, CEO, Co-founder Attala Systems; & Colin Cunningham, Data Scientist, Intel