manifold CLOUD is a live production software that runs on COTS FPGA programmable acceleration cards (PAC).
At its core, manifold CLOUD is a service-oriented software that utilizes an on-demand configurable pool of shared resources allocated within a private cloud environment.
Hardware resources (PAC) are pooled together in clusters which can be thought of as Virtual Private Clouds or, simply, a broadcast production.
Typically, a cluster has a fixed purpose for a set period of time, such as the “6 ‘o’clock news" or the “Sunday football game.”
Multiple clusters can be operated simultaneously, each with different services utilizing a shared hardware resource pool from one or more data centers.
Users operate manifold CLOUD services through a single-sign-on secure web UI which facilitates access to clusters.
All manifold CLOUD services are performed with guaranteed sub-frame latency just as live event operators are used to.
COTS FPGA accelerators use 90% less power than comparable CPU based servers.
If you're serious about green gas emissions you can't find a more efficient solution!
Utilizing the latest generation COTS FPGA Programmable Acceleration Cards offers up to 1.6Tbps of processing per RU. That's enough to process up to 1024 HD signals!
And because it's networked if you need more just add another server!
manifold CLOUD is built upon the time multiplexing of services using FPGA High Bandwidth Memory (HBM).
In difference to all other broadcast products on the market manifold CLOUD does not have a fixed amount of senders and receivers.
The only limitation is the amount of bandwidth allocated to the system!
A manifold CLOUD cluster has two main components - sources and services. Sources are audio, video, and metadata flows in either IP or SDI and are allocated to a cluster via NMOS or API.
manifold CLOUD services are live production functionality such as compression, multi-viewing, routing, audio and video mixing, color correction and color space conversion, etc. For example, a service could be a multiviewer output head, an up/down/cross converter instance, or a JPEG-XS encoder. A cluster contains all the services required for a particular production.
The capacity and performance of a manifold CLOUD system takes full advantage of the fast development cycles of COTS FPGA chip and server manufacturers while at the same time lowering the capex investment required.
manifold CLOUD scales with the total aggregated bandwidth of all the accelerators in a cluster.
Services are generic in that they run on any supported hardware regardless of manufacturer and are automatically instantiated on assigned PAC by manifold CLOUD.
manifold CLOUD is inherently resilient and will reroute services as required upon failure.
manifold CLOUD supports COTS FPGA Programmable Accelerator Cards from several vendors already, with more to be announced.
As of the current generation, these accelerators provide up to 400Gbps of processing per card. With 4 cards per server this allows for up to 1.6 Tbps of media processing per RU.
Utilizing these accelerators manifold CLOUD can for example process up to 512 x 3G multiviewer sources and 256 x 3G heads in one RU, and it continues to scale linearly as more compute is added!
prodesign FALCON 1SM21 Acceleration Card
The FALCON acceleration card offers Intel® Stratix® 10 FPGA technology with high throughput HBM2 memory integration. The board is based on PCIe Gen3x16 and off ers four 100GE QSFP-DD front ports supporting 8x 100Gbit/s or 4x 200Gbit/s
Thomas-Krenn.AG is a leading European manufacturer of customized server systems and our partner for integrated server and FPGA solutions.
BittWare - TeraBox servers
BittWare’s TeraBox™ range of certified server platforms feature the latest FPGA accelerators enabling customers to develop and deploy quicker with reduced risk and cost.