Every customer is unique and may even implement the same application differently, therefore, applications define workloads requirements, which may further call out the different resources needed to full-fill it. Some examples include:
The DFabric solution helps data-center infrastructure teams to serve their customers on workloads requirements with ease and resiliency. Any dynamic workload environment where topologies frequently change is the most suited to explore our solution.
- High Performance Compute
- Cloud Data-centers
- Media Entertainment
- Oil gas exploration
- Financial Services
- University Labs
DFabric as a black-box entity, discovers all the resources connected to it and makes them available to the orchestrators for scheduling by creating a shared-pool profile. Any user can select resources by calling-out the specific traits for profiles and see them connected in real-time. This allows resources to locally attached to the Host OS or passed through to the VMs directly.
After the attached resources are no longer needed, they can be detached, refurbished, and made available for the next allocation.
The following are some examples of use cases for different resources allocations:
Use-case#1 - [One to Many]
Request Type - A host (with an iFIC) needing some or all variety of resources to be connected from respective pools (with their own tFIC) and have those resources passed-through to the virtual machines (VMs) running on them. See Diagram 1 below.
Use-case#2 - [Many to One]
Request Type - A pool of devices (with a tFIC) could be connected to multiple hosts (via their respective iFICs) and have individual devices passed through to multiple address domains on different Host OS (and/or VMs). See diagram 2 below.
Compute - Sleds/servers with CPU/DRAM and PCIe form-factor IO slots with iFIC cards.
Graphic - Device chassis with PCIe graphics cards and fabric facing interface via tFIC.
Storage - Device chassis with SSD-JBODs and fabric facing interface via tFIC .
Memory/SCM - Device chassis with SCM-JBOD and fabric facing interface via tFIC.
Accelerator/GPU - Device chassis with GPU or FPGA accelerators for AI/ML/DL applications
Network - Device chassis with network cards (Ethernet, Fiber Channel, or Infiniband) to support north/south and east/west connectivity.
or Subscribe to Stay in Touch
©2020 Drut Technologies Inc. All Rights Reserved.