GPU-powered Virtual Device Interface
March 17, 2015, GPU Technology Conference, San Jose, CA—Barry Daniels from exponential-e described a GPU-powered VDI with GPU rendering. This Ethernet-based cloud network for enterprise delivers high bandwidth through smart links.
The company is a spin-out from British Telecom and offers network, cloud, and unified communications services. They are working to address the unique challenges of media companies; power for the servers, costs, capital expenditures, space, capacity, and security. In addition, many of these companies have to make all of the capabilities available to contract workers.
Some companies try to manage the issue by working 24/7 around the world, at the cost and delays of large-scale global transfers. Even though they are in places around the world, they still have problems finding talent. Few are considering public-based, multi-tenant cloud services due to the lack of security, and cannot find mobility solutions.
In the process of developing shared systems, users have to consider shared versus dedicated GPUs for the workstations. The biggest issue is render time. One solution is to use remote render farms that are leased on a monthly basis. The Nvidia Grid appliances are ideal for populating these render farms as they are natively configured for virtual environments.
The proposed configuration is to have the core network connect through a private, secure LAN to the remote render farm. This private layer 2 network allows the user to keep files in the existing datacenter and only requires a local subnet address for the secure connection to the render farm. The system is flexible and allows the user to scale on demand.
The security for this system has the highest government-level certifications on top of those within your own system. Other security features include anti-DDOS tools at all Internet peering points. Special purpose software include Palo Alto and Fortinet protection and purpose-built self-learning, context-aware anti-malware and internal threat security. Future upgrades will include GPU on demand, direct plug-ins for GPU render, render as a local service, security data sharing, and MPAA endorsements.
