Overview
Version
FineBI Version |
6.0 |
Functions
FineBI is an intelligent business analysis and service platform in B/S port. It can be deployed on a server through the web application server, providing a cloud server for companies. Users can visit and use the service platform only by one browser.
FineBI uses the Spider engine, which flexibly supports the analysis of different data magnitudes.
This passage will introduce the requirements of configuring FineBI software and hardware.
FineBI Performance Test Report
The following shows reports of the direct connect data and the extracted data.
Performance test report of the direct connection data:
FineBl V6.0 Direct Connection Performance Test Report.pdf
Performance test report of the extraction data:
Software Environment Recommendation
Software environments suitable for FineBI are as follows:
Company deployment: The Linux system is preferred rather than Windows and virtual machine systems. The Tomcat system is also recommended. You'd better not to deploy directly.
Personal use: There is no requirement for the way to deploy.
Operating systems like Windows, Linux, Mac, Unix, Axi, and IRIX which support 1.8 version of JDK can be used to deploy.
| Operating System | Description |
Database | Some main relational databases like Apache Kylin, Derby, HP Vertica, IBM DB2, Informix, Sql Server, MySQL, Oracle, Pivotal Greenplum Database, Postgresql, ADS, Amazon Redshift, Apache Impala, Apache Phoenix, Gbase 8A, Gbase8S, Gbase 8T, Hadoop Hive, Kingbase, Presto, SAP HANA, SAP Sybase, Spark、Transwarp Inceptor, and HBase and some non-relational databases like MongoDB can be used to deploy. For details, see Support Range of Data Connection. |
App Server | Web app servers like Tomcat, Jboss, Weblogic, and Websphere are supported. |
Browser | Single core: Chrome, Firefox, other browsers that support IE11 and above (including Edge), Safari, and opera. It is recommended to use Google and Firefox browsers in terms of the compatibility of the rendering engine. Dual core: 360 Browser, Sogou Browser, QQ Browser, UC Browser, CM Browser, and Baidu Browser. Only fast mode is supported. Compatibility mode is not supported. |
Note:
1. The Web server of FineBI should not be installed in physical computers or VM virtual machines which run resource-intensive apps at the same time. (For example, the database or app servers.) CPU that allows FineBI to actually take up 80% resources is recommended.
2. The version of Google browser is better in version 70 or above.
3. It is not recommended to use Safari browser if you are not using a Mac computer, because the older version of FineBI may be incompatible with it.
Server Recommendation
You need to first know whether you have used real-time data or extracted data.
Extracted Data
Select configurations according to the project concurrency and the size of metering. Select higher configurations when it falls into multiple configuration intervals.
The concurrency that businessmen often pay attention to refers to the number of online users per hour in scenario one or the total number of users in scenario two.
When the maximum data volume of single form is below 100 million, you can directly use the local disk of the Web server as the data storage medium.
1. Scenario 1: Self-service Analytics for High Daily Active UsersScenario description: BI projects with high daily active users can use self-service analytics to analyze data. (Data cannot be directly obtained from the FineBI query cache.)
Estimate the concurrency according to the node number summary: the number of online users (Y)= 300 * (The number of nodes (X) - 1) + 400.
Disk throughput and bandwidth need to be greater than 100 MB/s(the performance of a regular local hard disk HDD). SSD solid state hard disk is recommended.
Data Volume (Unit: row) | Available Disk Space |
0 to 500 million | 100 to 300 GB |
500 million to 1 billion | 300 to 600 GB |
10 million to 100 million | 600 GB to 1.5 TB |
Note:
1. FineBI is an IO intensive app, which relies heavily on the IO dish. Therefore, it is recommended to use the local disk or the SSD.
2. The read/write speed of disk needs to be at least 100 MB/s. The throughput of IOPS needs to be at least 10 KB.
According to the number of online users per hour during working hours, the reference table is configured as follows:
Number of Daily Active Users | Online (Number of Users Per Hour) | Concurrency (Number of Users Per Second) | Number of Forms-The Size of Forms | Recommended Configuration | Minimum Configuration | ||||
Number of Nodes | JVM/Physical Memory | CPU (2.5 GHz or Above) | Number of Nodes | JVM/Physical Memory | CPU (2.5 GHz or Above) | ||||
500 | < 100 | < 20 | < 100 pieces or < 1 TB
| 1 | 16 GB/32 GB | Eight cores and 16 threads | / | / | / |
1000 | 300 to 1000 | 20 to 70 | < 100 pieces or < 1 TB
| 2 | 16 GB/32 GB | Eight cores and 16 threads | 1 | 32 GB/64 GB | 16 cores and 32 threads |
2000 | 600 to 2000 | 40 to 120 | > 2000 pieces or > 1TB
| 2 | 32 GB/64 GB | 16 cores and 32 threads | 2 | 24 GB/48 GB | 16 cores and 32 threads |
3000 | 900 to 3000 | 50 to 160 | > 4000 pieces or > 2 TB
| 3 | 32 GB/64 GB | 16 cores and 32 threads | 3 | 24 GB/48 GB | 16 cores and 32 threads |
3500 | 1200 to 3500 | 60 to 190 | > 5000 pieces or > 3 TB
| 4 | 32 GB/64 GB | 16 cores and 32 threads | 3 | 32 GB/64 GB | 16 cores and 32 threads |
4000 | 1500 to 4000 | 80 to 220 | > 5000 pieces or > 3 TB
| 5 | 32 GB/64 GB | 16 cores and 32 threads | 4 | 32 GB/64 GB | 16 cores and 32 threads |
Note: JVM memory is not the same as the whole machine memory. It is recommended that JVM memory should account for 2/3 to 3/4 of the total machine memory.
2. Scenario 2: Many Users Viewing Dashboards At the Same TimeScenarios like corresponding users view a report at the same time can use data extraction. It is the cumulative number of visitors (Y) within 5 to10 minutes. (Data can be directly obtained from the FineBI query cache. That is to say, multiple identical calculation requests can be directly obtained and return results.)
When the number of requesting users per second reaches 160, the download speed of the load balancing server needs to reach 100 MB/s.
Estimating concurrent users based on the summary of the number of nodes: the number of concurrent users per 5 minutes (Y) = 380 * the number of nodes (X).
Number of Users Per 5 minutes | Number of Users Per Second | Recommended Configuration | Minimum Configuration | |||||
Number of Nodes | JVM/Physical Memory | CPU (2.5 GHz or Above) | Number of Nodes | JVM/Physical Memory | CPU (2.5 GHz or Above) | |||
< 400 | 40 | 2 | 16 GB/32 GB | Eight cores and 16 threads | 1 | 32 GB/64 GB | 16 cores and 32 threads | |
400 to 800 | 80 | 2 | 32 GB/64 GB | 16 cores and 32 threads | 2 | 24 GB/48 GB | 16 cores and 32 threads | |
800 to 1100 | 110 | 3 | 32 GB/64 GB | 16 cores and 32 threads | 3 | 24 GB/48 GB | 16 cores and 32 threads | |
1100 to 1600 | 160 | 4 | 32 GB/64 GB | 16 cores and 32 threads | 3 | 32 GB/64 GB | 16 cores and 32 threads | |
1600 to 2000 | 190 | 5 | 32 GB/64 GB | 16 cores and 32 threads | 4 | 32 GB/64 GB | 16 cores and 32 threads | |
Note: JVM memory is not the same as the whole machine memory. It is recommended that JVM memory should account for 2/3 to 3/4 of the total machine memory.
Description of user types are shown in the following:
User Type | Description |
Number of Daily Active Users | Within one day, the number of users logging into the BI system |
Number of Online Users | The number of users logging into the BI system at the same time |
The number of concurrent users | The number of users operating the BI system at the same time. That is, the number of users who are sending requests to the server at the same time. That is, the number of requests from the users the server can handle simultaneously. |
Concurrency limit when registering for Lic | For those from IP address, the server gets the IP address from the request as a concurrent key. The main restriction is on the cumulative number of IP in visiting systems. The parameter of Lic has no correlation with the number of users in the following text. |
Direct Connect Data
Description: Provide a recommended configuration based on the test results of the fourth section. The bandwidth between cluster nodes, nodes, and other components is 1000 MB/s.
According to the project concurrency and the calculation capability of the data source DB, select a higher configuration when it falls into multiple configuration intervals.
Editing concurrency cannot directly obtain data from the cache.
The upper limit of concurrent users per second corresponds to the cache hit and the lower limit corresponds to the cache miss. Recommended configuration are as follows:
Number of Daily Active Users | Online | Concurrency | Data Source Calculation Capability Processed Calculations Per Second | Recommended Configuration | Minimum Configuration | ||||||
Number of Nodes | JVM/ Physical Memory | CPU in 2.5 GHz and Above
| Number of Nodes | JVM/ Physical Memory | CPU in 2.5 GHz and Above
| ||||||
500 | < 100 | < 20 | < 10 | 1 | 16 GB/24 GB | Eight cores and 16 threads | 1 | 8 GB/12 GB | Four cores and eight threads | ||
2000 | 100 to 1000 | 40 to 90 | 10 to 20 | 2 | 16 GB/24 GB | 16 cores and 32 threads | 1 | 16 GB/24 GB | 16 cores and 32 threads | ||
3000 | 600 to 1500 | 60 to 130 | ≥ 30 | 3 | 16 GB/24 GB | 16 cores and 32 threads | 2 | 24 GB/48 GB | 16 cores and 32 threads | ||
4000 | 600 to 2000 | 60 to 170 | ≥ 30 | 4 | 16 GB/24 GB | 16 cores and 32 threads | 3 | 24 GB/48 GB | 16 cores and 32 threads | ||
Involving Direct Connect Data and Extracted Data
If you are in a situation where both direct connect data and extracted data are involved, you can configure it as the highest configuration requirements.
Modifying Parameters After Deployment
After deploying FineBI, it is necessary to modify the FineBI configuration parameters.
