Data Caching
Data caching primarily addresses two main architectural challenges: the replication and distribution of data across a large scale infrastructure, and near real time localized data processing.
Grids work efficiently when they are compute bound. As they scale, they tend to spend more time waiting for data and less time computing. In smaller grids, tasks typically contain data as well as compute instructions and shared data must be distributed multiple times to each core.
The key to further scalability is to distribute shared data, usually market data, to all nodes independently of the tasks. Databases are wholly inappropriate for this task and Data Caching solutions, in various configurations and topologies, are typically used to reduce the load of data stores and distribute data across compute grids.
Of course, disk based storage comes with an overhead. For real-time and low latency applications this overhead is a distinct disadvantage. Data caching solutions take data storage and processing into distributed, highly available memory, avoiding any disk based reads and writes. Data cache products are found driving websites, best execution engines, order management engines, pricing systems and risk management solutions.
We have strong partnerships in this area, working closely with all three major caching vendors: Oracle (Coherence), VMware (Gemfire) and GigaSpaces.
Some of our recent projects include:
Grids work efficiently when they are compute bound. As they scale, they tend to spend more time waiting for data and less time computing. In smaller grids, tasks typically contain data as well as compute instructions and shared data must be distributed multiple times to each core.
The key to further scalability is to distribute shared data, usually market data, to all nodes independently of the tasks. Databases are wholly inappropriate for this task and Data Caching solutions, in various configurations and topologies, are typically used to reduce the load of data stores and distribute data across compute grids.
Of course, disk based storage comes with an overhead. For real-time and low latency applications this overhead is a distinct disadvantage. Data caching solutions take data storage and processing into distributed, highly available memory, avoiding any disk based reads and writes. Data cache products are found driving websites, best execution engines, order management engines, pricing systems and risk management solutions.
We have strong partnerships in this area, working closely with all three major caching vendors: Oracle (Coherence), VMware (Gemfire) and GigaSpaces.
Some of our recent projects include:
- the construction of market data caches using Gemfire for a number of banks
- the integration of data caching solutions into a pricing engine for an on-line gaming company
- the implementation of a Coherence data caching solution underneath a 2000 node grid to increase efficiency
- the architecture of a GigaSpaces implementation for grid acceleration.
Data Caching
Data caching primarily addresses two main architectural challenges: the replication and distribution of data across a large scale infrastructure, and near real time localized data processing.
Grids work efficiently when they are compute bound. As they scale, they tend to spend more time waiting for data and less time computing. In smaller grids, tasks typically contain data as well as compute instructions and shared data must be distributed multiple times to each core.
The key to further scalability is to distribute shared data, usually market data, to all nodes independently of the tasks. Databases are wholly inappropriate for this task and Data Caching solutions, in various configurations and topologies, are typically used to reduce the load of data stores and distribute data across compute grids.
Of course, disk based storage comes with an overhead. For real-time and low latency applications this overhead is a distinct disadvantage. Data caching solutions take data storage and processing into distributed, highly available memory, avoiding any disk based reads and writes. Data cache products are found driving websites, best execution engines, order management engines, pricing systems and risk management solutions.
We have strong partnerships in this area, working closely with all three major caching vendors: Oracle (Coherence), VMware (Gemfire) and GigaSpaces.
Some of our recent projects include:
Grids work efficiently when they are compute bound. As they scale, they tend to spend more time waiting for data and less time computing. In smaller grids, tasks typically contain data as well as compute instructions and shared data must be distributed multiple times to each core.
The key to further scalability is to distribute shared data, usually market data, to all nodes independently of the tasks. Databases are wholly inappropriate for this task and Data Caching solutions, in various configurations and topologies, are typically used to reduce the load of data stores and distribute data across compute grids.
Of course, disk based storage comes with an overhead. For real-time and low latency applications this overhead is a distinct disadvantage. Data caching solutions take data storage and processing into distributed, highly available memory, avoiding any disk based reads and writes. Data cache products are found driving websites, best execution engines, order management engines, pricing systems and risk management solutions.
We have strong partnerships in this area, working closely with all three major caching vendors: Oracle (Coherence), VMware (Gemfire) and GigaSpaces.
Some of our recent projects include:
- the construction of market data caches using Gemfire for a number of banks
- the integration of data caching solutions into a pricing engine for an on-line gaming company
- the implementation of a Coherence data caching solution underneath a 2000 node grid to increase efficiency
- the architecture of a GigaSpaces implementation for grid acceleration.