Abstract: Today we are going to take you through the data to learn how GaussDB(for Cassandra) challenges the ultra-low price and high performance at the same time.

This article is from the Huawei cloud community “Data Proof, Huawei Cloud GaussDB(for Cassandra) performance improvement behind the things, by Gauss Cassandra official.

In our daily philosophy, the pursuit of cost performance is the most common, but do you know what to buy low configuration can enjoy high performance, low latency, ultra low price database? Today we are going to take you through the data to see what GaussDB(for Cassandra) can do to challenge the ultra-low price and ensure high performance at the same time!

As we all know, databases are needed where there is data. GaussDB(forCassandra) is best used in massive data scenarios such as logistics, content distribution, and live video.

This time, I will test the new 2U8G small size of GaussDB(for Cassandra) and compare it with the open source Cassandra. Come and have a look with me.

1. Test tools

2. Data model

3. Test environment

4. Test steps

4.1 Creating a table model

  • Create a keyspace:

    create keyspace ycsb WITH REPLICATION = {‘class’ :’SimpleStrategy’, ‘replication_factor’: 3 };

  • Enter the keyspace:

    use ycsb;

  • Create the usertable:

    create table usertable (y_id varchar primary key, field0 varchar, field1 varchar, field2 varchar, field3 varchar, field4 varchar, field5 varchar, field6 varchar, field7 varchar, field8 varchar,field9 varchar);

4.2 Data Preparation Phase

nohup ./bin/ycsb loadcassandra-cql -P workloads/workload-insert-mostly -threads $THREAD_COUNT-s > $LOG_FILE 2>&1 &
Copy the code

4.3 Pressure test of four YCSB business models

5. Test results

The pressure test is divided into two groups. One group has 80% CPU load and the other group has 50% CPU load

5.1 Indicator description of test results

OPS: number of operations per second insert_Lat(us) : Average delay per INSERT operation (microseconds) insert95Lat(us) : number of microseconds in which 95% of insert operations are performed 99% of insert operations in how many microseconds update_Lat(US) : Average latency for each UPDATE operation (microseconds) update95Lat(us):95% Microseconds or less for update operations update99Lat(us):99% Microseconds or less for update operations read_Lat(us) : Average latency for each READ operation (microseconds) read95Lat(us) : ≤ 95% Microseconds for read operations read99Lat(us) : ≤ 99% Microseconds for read operationsCopy the code

5.2 80% CPU Load

5.3 50% CPU Load

5.3.1 Comparison with open source

6, 2U8G cost

For example, buy an instance with 2U8G 3-node storage of 160GB on huawei cloud. The options are as follows:

The above on-demand rate is 5.84/ hour and the annual cost is 22,534 yuan

7,

  • Among the four general service models of YCSB, the newly launched 2U8G GaussDB(for Cassandra) has an OPS of 15000+ when the CPU is 80% fully loaded, and 8000+ when the CPU is 50% normally loaded

  • A high performance

The new 2U8G GaussDB(forCassandra) provides the same processing capacity as Cassandra 4U16G under normal CPU load, and the number of operations per second is 50% higher than that of Cassandra 4U16G in read and multi-service scenarios

  • Low latency

The average write latency is within 1.7ms, and the average read latency is within 3ms

The new GaussDB(forCassandra) 2U8G is also good

Buying advice:

GaussDB(for Cassandra) provides twice the performance of open-source databases and requires only one third of the self-built storage space, which helps customers save costs. For example: self-built 8 U32 GB x 3 data quantity: 90 GB (three copies), purchase GaussDB(for Cassandra) optional 4U16 GB x 3 nodes Data volume: 30 GB (three copies for DFV storage)

Author: Huawei Cloud GaussDB(for Cassandra) team

Product page: www.huaweicloud.com/product/gau…

Click to follow, the first time to learn about Huawei cloud fresh technology ~