Comparison of some memory products

an Oracle ExaData product introduction 1 hardware and software integrated machine

The lower level uses the common Oracle software, so his reliability and security are protected

Hardware: Use larger memory and larger flash to store

Configuration sub one-eighth with/One-fourth with/full

For example, full with 4T memory (storing thermal data), 22T flash memory (storage temperature data) 2 Framework principle:

1) using InfiniBand to support multiple concurrent links "transfer Cable" technology, improve the speed of reading data from the hard disk,

A link 500m/s,12 bar is 6g/s, the common use of Pci-x bus. Speed limit in 1g/s

2) adopts mixed column compression: Compression is faster to pass, compressed and delivers faster than Hana

Exadata is read into memory when data is used.

3) predicate filtering is the use of the where statement to only pass the required data to memory

4) Store the index, if the stored field is a date or shaping, it will automatically establish the corresponding index.

Every 1M of data is indexed, the index holds the maximum value and the minimum value two Oracle times ten 1 is a memory database

Very fast.

Scenario 2 is insufficient for real-time trading types

1) storage data can not be too large, preferably not more than 2G

2) does not support complex structure SQL statements, (like stored procedures, etc. or triggers) three SAP HANA 1 Features:

1) Put all the table data in the back to the memory, and then the project needs data directly from the memory to get the data

2) Massively parallel processing technology, which is processed by multiple independent CPUs in parallel

3) Use a data dictionary, so that the in-memory data is much lower than the actual storage capacity, the approximate compression ratio is 1:8

4) SAP supports both row and column storage

5) After Hana is used, Hana completely replaces the original table

6) SQL statements that support SQL 92 and SQL 99 standards

7) If it appears

8) power failure and other abnormal conditions, the method used is

Log record + disk record for data recovery 2 insufficient:

1) long implementation cycle

2) The application platform needs two times development

3) In the case of a more complex logic operation, the effect is generally

Four summary:

Memory technology can be considered from the following aspects 1 using memory compression technology, saving memory space 2 using a data dictionary for field mapping, compressing memory data 3 using the WHERE clause to filter conditions, reducing unnecessary memory data 4 Open source free memory database with big data, high concurrent access support

.....