System Configuration
Benchmark Application
Test Parameters
INSERT
SELECT
UPDATE
DELETE
MIXED WORKLOAD
MULTICORE SCALABILITY
CompleteDB Embedded Benchmarks |
![]() |
We tested our 64-bit CompleteDB Embedded database using native Direct C++ API. The purpose of the test was to show concurrent transactional processing power of our kernel. The details of the test parameters and the hardware are given below.
System Configuration |
![]() |
Server | Server T710 Dell Server with 2 CPU Sockets |
Number of Processors | 2 |
Processor | E5530 Xeon Processors, 2.4GHz 8M Cache |
Total Cores | 8 |
Total Hardware Threads | 16 |
NUMA Nodes | 2 |
RAM Memory | 98GB |
Operating System | Windows Server 2008 R2, Standard Edition |
Benchmark Application |
![]() |
The test is performed on 14 pinned hardware threads (logical CPU cores) leaving 2 threads for the operating system. The database was completely in-memory and in non-durable mode. Every test was run using two different access methods DirectId and SurrogateId. Ids were randomly shuffled for each SELECT, UPDATE and DELETE operation.
DirectId is an internal database record identifier representing the physical in-memory location where the record resides. It is the fastest method to access a record. SurrogateId is a hash key index used to uniquely identify a row in a database table.
Each modification operation (INSERT, UPDATE, and DELETE) was committed as a separate transaction using READ_COMMITED isolation level. Each SELECT operation was performed as a separate operation.
Test Parameters |
![]() |
Table size (records) | Number of records inserted in a table before test starts |
Number of columns | Number of columns in a test table |
Column size | Size of each column in bytes |
Access method | DirectId or SurrogateId |
Test Type | INSERT, UPDATE, DELETE, SELECT or MIXED WORKLOAD |
CPU Cores | Number of logical CPU cores used |
Transactions per second | Committed transactions per second |
INSERT |
![]() |
In this benchmark we test INSERT only performance.
|
|
||||||||||||||||||||||||||||
|
|
SELECT |
![]() |
In this benchmark we test SELECT only performance. Each column's value was copied into local memory during each select operation.
|
|
||||||||||||||||||||||||||||
|
|
UPDATE |
![]() |
In this benchmark we test UPDATE only performance. Each column's value was changed during each update operation.
|
|
||||||||||||||||||||||||||||
|
|
DELETE |
![]() |
In this benchmark we test DELETE only performance.
|
|
||||||||||||||||||||||||||||
|
|
MIXED WORKLOAD |
![]() |
In this benchmark we test mixed workload performance with the following read/write ratios - 1 delete per 4 updates, per 8 inserts, per 64 selects.
|
|
||||||||||||||||||||||||||||
|
|
MULTICORE SCALABILITY |
![]() |
In this benchmark we test the scalability of mixed workload performance on multiple cores (1, 2, 3 .. 14) with the following read/write ratios - 1 delete per 4 updates, per 8 inserts, per 64 selects.
|
|