Accelerating data analytics workloads for Deutsche Bank
The problem
The bank could not meet regulation deadlines due to an outdated solution of distributed storage workloads.
- The Financial Data Warehouse (FDW) of Deutsche Bank was based on a separate instance of SAS statistical analysis software for each compute node, which communicated independently with external storage.
- This old layout of several different environments that functioned independently was expensive and inefficient, resulting in high licensing costs.
The solution
A common layer for all nodes has been created. Excelero’s NVMesh software-defined storage was laid over the hardware to create a protected storage layer, pooling high speed NVMe drives instead of a legacy appliance.
- Low latency and high throughput allow for finishing analytics workloads much faster
- 72% savings in production run times, i.e. the application runs 3.5 times faster, which means achieving critical regulatory compliance
- High utilization based on distributed storage and efficient CPU consumption, means less cost of hardware
NVMesh Architecture
NVMesh supports the widest selection of protocols and fabrics, including TCP/IP, InfiniBand, RoCE v2, RDMA and NVMe-oF. NVMesh is deployed as a virtual, distributed non-volatile array and supports both converged and disaggregated architectures, giving customers full freedom in their architectural design. A key component of Excelero’s NVMesh is the patented Remote Direct Drive Access (RDDA) functionality, which bypasses the CPU and, as a result, avoids the noisy neighbors effect on the application. The shift of data services from centralized CPU to complete client side distribution enables linear scalability, provides deterministic performance for applications and enables customers to maximize the utilization of their flash drives.
Leveraging standard servers and multiple tiers of flash
Unmodified applications utilize pooled NVMe storage devices across the network at local speeds and latencies
Underlying storage can be presented as block volumes or through integrated file systems
NVMesh abstracts distribute NVMe storage resources and create a single pool of high-performance storage
In-depth, user-friendly interface
To facilitate maintenance, NVMesh features an intuitive GUI and a feature-rich performance analytics engine, which enables pinpointing anomalies quickly and at scale. Customers benefit from elaborate cluster-wide and per-object performance, and utilization statistics that help with monitoring and analysis of the storage environment performance. Administrators can benefit from fully customizable display of detailed metrics of application workloads and datasets.
Deployment in private cloud–
Storage Configurator
NVMesh was designed from the ground up to support any hardware. With the added support for traditional network fabrics and protocols, when enabling NVMe over Ethernet and InfiniBand, customers don’t have to invest in new networking technologies to deploy NVMesh high-performance computing storage.
