This document is applicable to historical FineBI 6.0.x projects deployed on non-O&M platforms.
For details about FineBI 6.1 projects deployed on the O&M platform, see FineOps Help Doc.
To meet enterprises' business requirements on high availability and concurrency, FanRuan supports both standalone and cluster deployment methods.
Also, the corresponding deployment architecture is recommended based on the specific business load situation to ensure the normal and stable business running.
This document describes various recommended optimal deployment solutions based on the number of concurrent users and server resources.
Deployment solution description:
(1) The deployment solutions listed in this document are all based on the configuration requirements when the server is dedicated to the project and components. (Namely, no other content is installed on the server.)
· The FanRuan standalone architecture consists of the project node and the configuration database (external FineDB).
· The FanRuan cluster architecture consists of project nodes, the load balancer, the state server, the file server, and the configuration database (external FineDB). Different application components in the architecture need to be selected and matched for different scenarios.
· During actual deployment, you can freely select the software and hardware in the range supported by the technology roadmap.
(2) Considering that server resources of some users are limited:
· For project node deployment solutions, this document describes recommended and minimum configuration requirements respectively. If resources are abundant, meet the recommended configuration requirements. If resources are insufficient, at least meet the minimum configuration requirements.
· For cluster deployment solutions, this document provides both high availability and non-high availability deployment solutions. If resources are abundant, you are advised to use the high availability deployment solution. If resources are insufficient, the non-high availability deployment solution can be considered.
(3) For standalone projects, unavailability is unavoidable if a low-possibility fault occurs. If high availability is required, you are advised to preferentially select the cluster architecture for deployment.
System performance description:
(1) Project node quantity:
· In a cluster scenario, the more the nodes are, the larger number of concurrent users the system can handle, and the more similar the average template response time is.
· The number of concurrent users the system can handle in a standalone project is only one half that in a dual-node cluster, and one third that in a three-node cluster.
(2) Template performance:
· In a standalone scenario, the larger the JVM memory is, the shorter the template response time is.
· In a cluster scenario, if the number of concurrent users is similar in each node, the average template response time is similar.
· The performance of preview requiring platform authentication is worse than that of preview not requiring platform authentication. This is because different users cannot share the same result cache.
Server type
Linux server
System version
Ubuntu 18.04.4 and later versions
System kernel
3.10 and later versions
System architecture
x86_64
Available disk space
Configured according to the data volume for the server where each node is located
At least 40 GB for the root directory
The following table lists the available disk space when the data volume is 0 to 5 million rows.
0 – 5 million
100 – 300 GB
5 – 10 million
300 – 600 GB
10 – 100 million
600 GB – 1.5 TB
Middleware
Tomcat 8.5.57 and later versions (the latest version recommended)
JDK 1.8 (8u221 and later versions; the latest version recommended)
Disk type
SSD
FineBI is an intensive I/O application, which relies heavily on the disk I/O. Therefore, you are advised to use the local disk or the SSD.
Disk read/write speed
Above 100 Mbps
IOPS
Above 10K
You can deploy only one project node of the cluster on one server. Therefore, you need to prepare as many servers as the number of project nodes.
The following describes user dimensions.
Number of daily active users
Number of users who log in to the FineBI project within one day
Number of online users
Number of logged-in users in FineBI at a certain moment
Number of concurrent users
Number of users who perform operation in FineBI at a certain moment.
It represents how many users send requests to the server at the same time, namely the maximum number of users that can send requests for the server to handle at the same time.
Concurrency limit during license registration
Maximum number of system-accessed IP addresses obtained by the server from requests (as concurrent keys).
The concurrency limit mainly restricts the cumulative number of IP addresses visiting the system. This parameter for license registration has no correlation with the number of users in the following text.
This document provides recommended configurations for three scenarios. If your FineBI project involves multiple scenarios, select the highest configuration to deploy your project.
You need to first check whether your data is real-time or extracted one according to Difference Between Direct-Connected Data and Extracted Data.
This scenario prevails when only direct-connected data (rather than extracted data) is used.
The bandwidth between nodes in the cluster or nodes and other components is 1000 Mb/s.
If multiple configuration options are available, select a higher configuration based on the project concurrency and calculation capability of the data source DB.
Users with concurrent edits are all ones obtaining data not from the cache.
For concurrent users per second, you need to determine whether all of them obtain data from the cache. If so, check whether the number of concurrent users per second reaches the upper limit. If not, check whether the number of concurrent users per second reaches the lower limit.
Number of Online Users
Per Hour
Number of Concurrent Users
Per Second
Data Source Calculation Capability
Number of Processed Calculations Per Second
500
< 100
< 20
< 10
Standalone configuration, in which nodes should meet the following requirements:
CPU: 8 cores; 16 threads; 2.5 GHz
JVM memory: 16 GB
Physical memory: 24 GB
CPU: 4 cores; 8 threads; 2.5 GHz
JVM memory: 8 GB
Physical memory: 12 GB
2000
100 – 1000
40 – 90
10 – 40
10 – 20
Two-node cluster configuration, in which each node should meet the following configuration:
CPU: 16 cores; 32 threads; 2.5 GHz
Physical memory: 32 GB
3000
600 – 1500
60 – 130
30 – 60
≥ 30
Three-node cluster configuration, in which each node should meet the following requirements:
JVM memory: 24 GB
Physical memory: 48 GB
4000
600 – 2000
60 – 170
60 – 80
Four-node cluster configuration, in which each node should meet the following requirements:
This scenario prevails for self-service data analysis of FineBI projects with a high number of daily active users (namely users who obtain data not from the FineBI cache).
Concurrency can be estimated based on the total number of nodes as follows: Number of online users (Y) = 300 * (Number of nodes (X) - 1) + 400.
Disk throughput and bandwidth need to be greater than 100 MB/s (namely performance of normal HDDs). SSDs are recommended.
The JVM memory is not the same as the total device memory. You are advised to set the JVM memory to 2/3 to 3/4 of the total device memory.
< 100 tables or < 1 TB
300 – 1000
20 – 70
JVM memory: 32 GB
Physical memory: 64 GB
40 – 120
> 2000 tables or > 1 TB
900 – 3000
50 – 160
50 – 80
> 4000 tables or > 2 TB
JVMmemory: 32 GB
Physicalmemory: 64 GB
Two-nodecluster configuration, in which each node should meet the followingconfiguration:
1200 – 3500
60 – 190
70 – 100
> 5000 tables or > 3 TB
1500 – 4000
80 – 220
80 – 200
Five-node cluster configuration, in which each node should meet the following requirements:
This scenario prevails when users view data such as reports at the same time. In this scenario, the system generally calculates the cumulative number of visitors (Y) within 5 to 10 minutes (namely the number of users who obtain data totally from the FineBI cache, among which a user with multiple calculation requests is returned by the engine as only one result).
Concurrency can be estimated based on the total number of nodes as follows: Number of concurrent users per 5 minutes (Y) = 380 * Number of nodes (X).
When the number of request users per second reaches 160, the download speed of the load-balancing server needs to reach 100 MB/s.
Number of Users
Per 5 Minutes
< 400
40
400 – 800
80
800 – 1100
110
1100 – 1600
160
1600 – 2000
190
Applicable object:
Standalone project
· Application server: one node
· FineDB configuration database: MySQL 5
Parameter configuration:
The FanRuan standalone architecture consists of the project node and the configuration database (external FineDB).
Database (FineDB) Server Configuration
Server quantity
1
If conditions permit, you are advised to deploy the external configuration database on one separate server.
If conditions are limited, you can deploy the external configuration database and the FineBI project on the same server.
But at least ensure that the server is dedicated to the FineBI project and the database (namely, no longer deployed with other applications).
Operating system
System version: Ubuntu 18.04.4 and later versions
System kernel version: 3.10 and later versions
System architecture version: x86_64
Network requirement
(1) You are advised to deploy the configuration database and the application project on the same network segment to avoid problems such as network fluctuations.
(2) If the configuration database and the application project are in a public network environment, the bandwidth needs to be above 10 Mbps.
(3) The network needs to be smooth between the configuration database and the application project and their ports need to be accessible to each other.
Database quantity
Database type
MySQL 8 database
Database version: versions later than MySQL 8.0.20
Database driver version: 5.1.49 (compatible with MySQL 5/8)
Database character set: utf8
Database sorting rule: utf8_bin
Database name: only numbers, letters, underscores, and periods (.) supported
Database permission: create, delete, alter, update, select, insert, and index
CPU
Above 2.5 GHz
8 cores; 16 threads
Physical memory
8 GB
More than 200 GB
Network speed
100 Mbps
100 MB/s
Standard direct-connected/extracted cluster
· Application server: at least two nodes
· State server: Redis standalone
· File server: SFTP
· Load balancer: NGINX
· Configuration database: MySQL
The FanRuan cluster architecture consists of project nodes, the load balancer, the state server, the file server, and the configuration database (external FineDB).
Sharing Requirement
The requirements listed in this section are ones that must be met by every type of cluster components.
If conditions permit, you are advised to deploy the load balancer, the state server, the file server, and the external configuration database on one server separately.
If conditions are limited, ensure that these components are deployed on at least one separate server (exclusively occupied by these components and not shared with the FineBI project).
GNU Compiler Collection (GCC)
The deployment of both Redis and NGINX in the Linux system relies on GCC.
Ensure that the system has the GCC environment.
Query command: gcc -v
Installation command: yum install gcc gcc-c++
(1) You are advised to deploy each component and the application project on the same network segment to avoid problems such as network fluctuations.
(2) If each component and the application project are in a public network environment, the bandwidth needs to be above 10 Mbps.
(3) The network needs to be smooth between each component and the application project and their ports need to be accessible to each other.
State Server
Deployment solution
Redis standalone
JVM/Physical memory
4 GB/8 GB
More than 100 GB
File Server
SFTP
500 GB – 1 TB
Expand the capacity according to usage.
Notes
The FTP server can be installed in the Linux system only by the user with highest permissions (root). Otherwise, the installation cannot proceed. If the FTP server does not need to be installed, this requirement can be ignored.
Load Balancer
NGINX
1.21 and later versions recommended (preferentially the latest version)
Highly-available extracted cluster
· Status server: Redis cluster with three master nodes and three slave nodes
· File server: NAS
· Load balancer: NGINX + Keepalived
· Configuration database: PostgreSQL
Redis cluster with three master nodes and three slave nodes
If conditions permit, prepare six servers on each of which one node is deployed.
If conditions are limited, prepare three servers on each of which one master node and one slave node are deployed.
NAS
2
Keepalived + NGINX
NGINX 1.21 and later versions recommended (preferentially the latest version)
Highly-available database in the master-slave mode
滑鼠選中內容，快速回饋問題
滑鼠選中存在疑惑的內容，即可快速回饋問題，我們將會跟進處理。
不再提示
10s後關閉
Submitted successfully
Network busy