YCSB benchmark results for ApsaraDB for MongoDB across physical server and cloud disk architectures, covering multiple instance specifications and read/write ratios.
Test environment
|
Parameter |
ECS for stress testing |
ApsaraDB for MongoDB (Physical Server) |
ApsaraDB for MongoDB (Cloud Disk) |
|
Region and zone |
China (Hangzhou) Zone H |
China (Hangzhou) Zone H |
China (Hangzhou) Zone H |
|
Network type |
VPC |
VPC |
VPC |
|
Instance family |
c6e (Compute-optimized, enhanced performance) |
General-purpose and dedicated |
Dedicated |
|
Instance type |
ecs.c6e.2xlarge (8 cores, 16 GB) |
||
|
Storage type |
PL1 ESSD cloud disk |
Local SSD |
PL1 ESSD cloud disk |
|
Storage space |
- |
500 GB |
500 GB |
|
Topology |
- |
Standard three-node replica set (one primary, one secondary, and one hidden node) |
Standard three-node replica set (one primary, one secondary, and one hidden node) |
|
Image version |
Alibaba Cloud Linux 3.2104 LTS 64-bit |
3.10.0-327.ali2017.alios7.x86_64 |
4.19.81-17.2.al7.x86_64 |
|
Instance kernel version |
- |
MongoDB 4.2 (baseline minor version 4.2.13) |
MongoDB 4.4 (baseline minor version 4.4.1) |
|
Consistency configuration |
- |
Default read/write concern |
Default read/write concern |
The ECS and ApsaraDB for MongoDB instances are in the same region and zone. Round-trip time (RTT): approximately 0.104 ms.
Test tool
This test uses YCSB 0.17.0 for benchmarking.
YCSB is a Java-based benchmarking tool for multiple databases.
Test procedure
-
Add the primary private IP address of the ECS instance to the ApsaraDB for MongoDB whitelist. Modify a whitelist.
NoteYou can log on to the ECS console and view the primary private IP address of the ECS instance in the Network Information section of the Instance Details page.
-
Connect to the ECS instance. Connect to an ECS instance.
-
Load test data using the YCSB tool.
./bin/ycsb load mongodb -s -p workload=site.ycsb.workloads.CoreWorkload -p recordcount=1000000 -p mongodb.url="mongodb://test:****@dds-bp13e84d11****.mongodb.rds.aliyuncs.com:3717/admin" -threads 8Modify the following parameters:
-
recordcount=1000000: The total number of records to load into the ApsaraDB for MongoDB instance. -
mongodb.url="mongodb://test:****@dds-bp13e84d11****.mongodb.rds.aliyuncs.com:3717/admin": The endpoint of the ApsaraDB for MongoDB instance. In this example, the database account is test and the database is admin.NoteYou can log on to the ApsaraDB for MongoDB console and view the endpoint in the Private Connection - VPC section on the Database Connection page.
-
threads 8: The number of concurrent threads for the client.
-
-
Run the following command to perform stress testing.
./bin/ycsb run mongodb -s -p workload=site.ycsb.workloads.CoreWorkload -p recordcount=1000000 -p operationcount=5000000 -p insertproportion=0 -p readproportion=50 -p updateproportion=50 -p requestdistribution=zipfian -p mongodb.url="mongodb://test:****@dds-bp13e84d11****.mongodb.rds.aliyuncs.com:3717/admin" -threads 8Modify the following parameters:
recordcount=1000000: The total number of data records loaded to the ApsaraDB for MongoDB instance.-
operationcount=5000000: The number of read and write operations to perform. -
insertproportion=0: The proportion of insert operations. -
readproportion=50: The proportion of read operations. -
updateproportion=50: The proportion of update operations. mongodb.url="mongodb://test:****@dds-bp13e84d11****.mongodb.rds.aliyuncs.com:3717/admin": The endpoint of the ApsaraDB for MongoDB instance. In this example, the database account is test and the database is admin.NoteYou can log on to the ApsaraDB for MongoDB console and then view instance endpoints in the Internal Connections - VPC section of the Database Connections page.
threads 8: The number of concurrent threads on the client.
Test results
Scenarios
-
r:w=100:0: 100% read, 0% write. -
r:w=95:5: 95% read, 5% write. -
r:w=50:50: 50% read, 50% write. -
r:w=5:95: 5% read, 95% write. -
r:w=0:100: 0% read, 100% write.
Parameters
-
recordcount: Total records loaded into the ApsaraDB for MongoDB instance. -
operationcount: Total read and write operations. -
threads: Concurrent client threads. -
throughput: Operations per second (OPS). -
RAL: Average read latency in microseconds (us). -
WAL: Average write latency in microseconds (us).
If RAL or WAL is 0, no read or write operation is performed.
The following tables and charts compare physical server and cloud disk architectures at various read/write ratios.
Physical server instance (MongoDB 4.2)
Cloud disk architecture instance (MongoDB 4.4)
Performance comparison
-
Instance type: 4 cores, 8 GB memory
-
Instance type: 8 cores, 16 GB memory
-
Instance type: 8 cores, 32 GB memory
-
Instance type: 16 cores, 64 GB memory
-
Instance type: 2 cores, 16 GB memory
-
Instance type: 4 cores, 32 GB memory
Test result analysis
-
For
CPU:Memory=1:2, physical servers outperform cloud disk by 7%-50%, most notably at 4 cores, 8 GB. -
For
CPU:Memory=1:4, the gap narrows to about 10%. Physical servers suit write-heavy workloads; cloud disk suits read-heavy workloads. -
For
CPU:Memory=1:8:-
Physical server and cloud disk deliver nearly identical performance.
-
As memory increases, cloud disk significantly outperforms physical server, most notably at 4 cores, 32 GB.
-
The cloud disk architecture offers the following advantages:
-
Dedicated resources eliminate contention from shared physical servers.
-
Independent compute and storage scaling without data migration.
-
Supports ESSD PL1, PL2, and PL3 performance levels with cost-effective pay-as-you-go billing.