Get the App
SLTechnology News&Howtos  ›  Database  › 

Informix database selection and system planning

Shulou Source: shulou.com Published: 2022-06-01 15:43:58 10月02日 Update

I. Selection of informix business scenarios

1. Financial industry, such as China Construction Bank, PICC, Pudong Development Bank, etc.

2. Communication industry, such as China Mobile, China Unicom and China Telecom

3. Retail industry, such as Walmart, Carrefour, CR Vanguard, etc.

4. E-commerce industry, such as Amazon, etc.

........... All scenarios fit, except for the 100% requirement, which requires very high data consistency.

Second, informix machine equipment selection

1. Selection of machine equipment, roughly evaluated through business scenarios

demand

Indicator Response Time Read and Write Requests MS Level Total Data Returned 1T/1 Year

Number of requests per second 15000 times Data read-write ratio 2:1 Importance core order system, P1 level fault; integral system, P2 level fault Remarks data timeliness, historical order review less, general operation Data within three weeks, data length 1K

Based on the above information, evaluate the system

1 terabyte of data generated in a year

Calculation: Data generated per second:1*1024* 1024 * 1024/(365*24*60*60)=34K/S

15000 requests per second, read/write ratio 2:1

Calculation: Number of reads per second: 15000*2/3=10000 times/s; Number of writes per second: 15000*1/3=5000 times/s

Data record length is 1K

Calculation: According to the above, the data written per second is 34K, and the data recording length is 1K, so the insert per second is 34K/1K=34 times. Since 5000 times are written per second,(update+delete+insert), then calculate (update+delete) 5000-34=4966 times, informix reads and writes are processed in units of page, a page size is 16k,(can also be set to 2k, 8k), then the number of write requests per second is 5000*16=80M/s, the number of read requests per second is 10000*16=160M/s

Processing data up to 20 days

Calculation: Thermal data processed is 1*1024/365*20=56G

Operation millisecond level return

Estimate: millisecond level return, then data needs to be put into memory as much as possible. According to the above information, it can be determined that BUFFERPOOL size is 56G, while other components, OS, Network, Application roughly need 4 G memory, so the memory overallocation is about 60G, according to the overallocation principle, write bandwidth limit is 100MB/S, read bandwidth limit is 200MB/S.

The importance is determined according to the service failure level P1, and HA technology of informix is adopted.

3. Disk selection

HDD: mechanical hard disk, inexpensive, long life

SDD Solid State Drive: Memory particles, slightly expensive, IO read and write high, short life

Tags: Data memory quantity industry system business scenario fault length China processing selection importance information size life bandwidth machine frequency proportion Apple Docker Huawei Linux macOS MariaDB Microsoft MySQL NVidia OPPO Reno Shulou Technology Apple NVidia MariaDB Microsoft