Maxeler Technologies’ core expertise is delivering complete high-performance dataflow platforms for a wide range of application domains. Maxeler provides a range of dataflow computing platforms for different types of application. This new level of performance comes at a fraction of the cost, power and space compared to using conventional computers. Maxeler dataflow computers bring a novel data-centric computing technology to the forefront of high performance computing. Where computation really matters, the Maxeler approach has the potential to make a significant difference.
Dataflow Computing: At Maxeler, we lead in innovation, development and delivery of High Performance Computing solutions. Our solutions include hardware, software and services to gain at least an order of magnitude advantage in performance per unit of rack space, computations per Watt (green computing), and top price-performance considering total cost of ownership for monolithic applications.
Our solutions exploit dataflow computing – a revolutionary way of performing computation, completely different to computing with conventional CPUs. Dataflow computers focus on optimizing the movement of data in an application and utilize massive parallelism between thousands of tiny ‘dataflow cores’ to provide order of magnitude benefits in performance, space and power consumption. An analogy for moving from control flow to dataflow is the Ford car manufacturing model, where expensive highly-skilled craftsman (control flow CPU cores) are replaced by a factory line, moving cars through a sea of single-skill workers (dataflow cores).
Maxeler provides compute solutions to enable production deployment of dataflow computing, including high-performance compute nodes, compilers and management software. Our technology is in use at FORTUNE 500 companies and universities across the world (see our publications for more information).
Computation: Maxeler’s Multiscale Dataflow Technology delivers unmatched performance per unit space and per Watt for high performance computing applications.
With Maxeler dataflow engines, the algorithm is mapped onto dataflow cores and data streams from memory through the dataflow engine where operations are performed and data is forwarded directly from one dataflow core to another without being written to the off-chip memory. The dataflow structure not only provides high compute performance, but also naturally optimizes the use of memory bandwidth, so even algorithms that are traditionally regarded as memory bound such as sparse matrix solvers can still be accelerated by orders of magnitude.
Maxeler technology has been successfully utilized for large-scale complex Monte Carlo simulations, irregular financial tree-based partial differential equation (PDE) solvers, 3D finite difference (FD) and finite element (FE) solvers, as well as optimization and pattern matching problems. You can see examples of different computational methods in some of our publications:
Data: Maxeler’s dataflow computing platforms are ideal for “big data” problems, with large onboard memories backed by high performance computation capabilities.
The MPC-X series dataflow nodes provide up to 384GB of DRAM and ultrahigh speed connectivity to other nodes, in a dense 1U form factor, allowing problems that might normally use disk or multiple nodes to run in memory on a single node. DFEs can incorporate lossless or lossy data compression into application dataflows, directly multiplying memory capacity and bandwidth and allowing terabytes of data to be held in memory within a single node.
For database or data analytics applications, DFEs can compute bulk data processing such as joining or merging while holding the entire database in memory (possibly in compressed form). Alternatively, DFEs can provide an ultralow latency caching layer (for example a key-value store) between backend storage or database and front-end services.
Connectivity applications are highly efficient on dataflow computing systems. Dataflow engines focus on optimizing movement of data in a system, allowing multiple channels of data to pass through processing elements at full line-rate with predictable low latency. The MPC-N series connectivity nodes provide multiple dataflow engines with data connections.
Dataflow engines can monitor data traffic for applications such as deep packet inspection or lawful intercept, filtering line-rate traffic for data points of interest and sending identified data packets to CPUs for further analysis.
Low latency financial trading applications can utilize our Maximum Performance Trading (MPT) platform to deliver programmable trading applications or Direct Market Access (DMA) gateways with fully in-hardware turnaround times of a few microseconds. Financial exchanges can also exploit the MPT platform for ultralow latency matching engines, market data manager or order routers.