Background: BridgeMarketing is a technology company dedicated to the electronic advertising industry in New York. Its business includes electronic advertising, third-party data services, and analysis. wang Pei, head of the technical team, graduated from Tsinghua University and Columbia University. before using QData, the data analysis engine used GoogleBigQuery. With the development of business, the following problems were encountered: 1.
Background: Bridge Marketing is a technology company dedicated to the electronic advertising industry in New York. Its business includes electronic advertising, third-party data services, and analysis. wang Pei, head of the technical team, graduated from Tsinghua University and Columbia University. before using QData, the data analysis engine used Google BigQuery. With the development of business, the following problems were encountered: 1.
Background:
Bridge Marketing is a technology company dedicated to the electronic advertising industry in New York. Its business includes electronic advertising, third-party data services and analysis. wang Pei, head of the technical team, graduated from Tsinghua University and Columbia University. before using QData, the data analysis engine used Google BigQuery. With the development of business, the following problems were encountered:
1. Multi-Table Association is often encountered in data analysis. However, when Google BigQuery is associated with two tables, the data volume of small tables cannot exceed 1 MB. Therefore, only simple queries are supported;
2. Data in Google BigQuery does not support modification or deletion. Therefore, each time you need to clear and reload the data, and then upload the data, the entire process is long, which slows down the service response ;?
Test process:
Bridge Marketing designs test scenarios based on its own business. Four most commonly used tables are used in the test. The test scenarios are as follows:
1. Data Loading (data loading, statistical analysis, index creation, etc );
2. Full Table query for a single table (more than 480 Wide Table fields );
3. Multi-table join query. the test process mainly targets QData and Oracle Exadata1/4 and 1/8 configurations of voqutech.
Test conclusion:
1. QData completes all 17 tests;
2. When the hardware configuration of QData is significantly lower than that of Oracle Exadata 1/4 and 1/8, the performance index is significantly better than that of Oracle Exadata 1/8;
3. All services on Google BigQuery will be migrated to QData.
User rating:
Compared with ExaData, QData has the same level of performance both in IOPS and throughput. however, the price of QData is only about one pair th of that of ExaData. it is also very easy to use, maintain and expand. after our boss heard the onsite test result, he said it seems to him that he just hit one million jackpot. he is right, we saved one million dollars by choosing QData.
? Test report
?
1Test Model configuration:
1.1 1/4 configuration of server
Server Type |
QuantityQuantity |
Configuration |
Computing node |
2 |
48 CPU cores/1 TB memory |
Storage Node |
3 |
36 CPU cores/9.6 TB Flash Cache/36x1.2 TB machine Disk |
1.2 1/8 configuration of server
Server Type |
QuantityQuantity |
Configuration |
Computing node |
2 |
24. CPU cores/512 GB memory |
Storage Node |
3 |
36 CPU cores/4.8 TB Flash Cache/18x1.2 TB machine Disk |
1.3 QData Server Configuration
Server Type |
QuantityQuantity |
Configuration |
Computing node |
1 |
32 CPU cores/96 GB memory |
Storage Node |
2 |
32 CPU cores/6.4 TB Flash |
?
2Data Loading Test
2.1 Exadata 1/4
ToUniqueFULLNEWExample,Data exceeds140010 million rows,Capacity28 GB,Data Loading time38.14Second to collect statistics13.79Seconds,Create an index17.96Seconds,Entire time1Minute9.89Seconds.
2.2 Exadata 1/8
2.3 QData
2.4 comparison of test results
From the comparison results, we can see that QData and Exadata 1/4 have comparable performance (or can be said to be slightly ahead). Compared with Exadata 1/8, the performance is improved by about 40%.
?
3Association TestTry
3.1 Test Description
TestMethod |
DescriptionMing |
SELECT 1 |
Full table scan for a single table |
Addr Match |
2. Table Addr-based field Association |
Email Match |
2. Email field-based Association |
SELECT 2 |
4 table join |
Test under loads |
Total time |
3.2 Test Results?
?
According to the test results, QData fully exceeds Exadata 1/8, and is superior to Exadata 1/4 in most test scenarios.
Work silhouette:
QData installation is complete
Test Complete
Opposite the IDC, Manhattan
Original article address: voqutech and QData log on to New York, USA. Thank you for sharing your gratitude.