Data Mart > MPP > Configuration Requirements > Hardware Requirements and the Common Configuration Ratio Relations

❖Hardware Requirements

The number of entry-level server is several units to dozens of units. The memory of each server is over 32G,CPU is over 4 Processor, and the independent hard disk is over 1TB, with the network card of gigabyte or 10 gigabytes.

After setting up the mart, the transverse expansion can be done by hot plug.

❖Common configuration ration relation

Each CPU Core can assume 20 million rows of data in real time calculations.
Description	If per piece of data is 1000 Bytes,20 million data is 20G of raw data.Suppose the compression ratio of 20G raw data is 1/3 and data loaded by memory is 1/3, then the data in the memory is almost 2G.
The more CPUs, the better performance of system.
Description	For MPP data mart,CPU expansion and performance improvement, is basically linear correlation. Because the system is based on memory calculations, memory matching is also required.
When calculating the CPU, we also need to consider the concurrent amount.
Description	The number of online users, as well as the number of concurrent queries submitted by online users, will vary between CPU and memory requirements. We need to find the number of queries executed per second to estimate the CPU and memory requirements for concurrency. In general, the more concurrent queries, the higher the CPU requirements, the more likely the thermal data will rise, and the temporary memory requirements will rise slightly. The industry formula is generally each concurrency on the original base to append approximately 5% of memory and CPU.
Sufficient memory can help to reduce data exchange.
Description	Each system's thermal data volume is not the same, some 1/5, some 1/3, some 1/10. When there is enough memory, all the hot data is kept in memory, and then the memory is added, which will not have much effect on the performance of the system.
In general, the ratio of raw data to the amount of memory can be defined as 10:1, while the ratio of raw data to CPU core can be defined as 20G:1. In this configuration, the query speed can reach the second level response.
Description	1T of data volume, requires 100G memory, 50 CPU Core. Consider the following configuration: Four 2-channel PC Server, each 2 CPU, and each CPU has 6Core, each memory 24G.
Description	500G of data amount, need almost 50G memory, 24 CPU Core. Consider the following configuration: Three 2-channel PC Server, each 2 CPU, and each CPU has 4Core, each memory 16G.