4.16 PERFORMANCE
4.16.1 LCU Performance
This section shows some performance figures of the LCU measured with LCC NOV96 on a Motorola MVME167 33MHz processor board running VxWorks 5.2. Times marked with * are measured with VxWorks 5.1.1 and LCC JUL95.
Database Access
| Database read/write |
Access time (microseconds) |
| Symbolic address one level hierarchy (time depends also on the size of the database) scalar attribute |
1000 |
| Additional time per hierarchy level |
+ 250 to 450 |
| Additional time for vector attribute |
+ 100 |
| Additional time for table attribute |
+ 300 |
| Direct address scalar attribute |
100 |
| Direct address vector attribute |
110 |
| Direct address table attribute |
120 |
I/O Signal Access, direct addressing
| Action |
Time (microseconds) |
| Analog write |
300 |
| Analog read |
480 |
| Digital write |
240 |
| Digital read |
215 |
Message System: command received from another node
| Action |
Time (microseconds) |
| Command sent to CI function, returning reply, total time |
6200* |
| Command sent to CI which spawns a task. The task returns reply |
8100* |
| Command sent to CI which spawn task. The task sends a command to a device, waits for the reply from the device and returns reply to the originator of the command |
17200*
|
The times measured on the message system starts for commands coming from an external node when the networking software receives the first interrupt from the network interface and ends when the reply has been delivered to the network interface.
Message system send command to local process
| Action |
Time (microseconds) |
| Command sent to local process, no CI, small buffer (< 100 bytes) |
1500* |
| Reply returned to sending process, small buffer (< 100 bytes) |
800* |
| Command sent to local process, no CI, large buffer (8000 bytes) |
3500*
|
| Reply returned to sending process, large buffer (8000 bytes) The figure corresponds to a data transfer rate of 3.2 MBytes/second |
2500* |
| Send command LCCGVER from process to lccServer and receive reply |
3600 |
| Send command LCCSHUT from process to lccServer and receive reply (lcc state LOADED, three devices) |
12000 |
The times is measured for sending from the time the sending process calls msgSendCommand until the receiving process has returned from msgRecvMsg. The time to return reply is measured from the time the process sending the reply enters msgSendReply until the receiving process returns from msgRecvMsg. The total round-trip delay can be calculated by adding the time to send a command and the time to return a reply
Data transfer rate from LCU to WS
| Mode |
Transfer rate in kBytes/second |
| Message system, running qsemu on WS over Ethernet |
800 |
| Message system, running qsemu on WS over FDDI |
1100* |
| Raw socket connection over Ethernet |
1100*
|
| Raw socket connection over FDDI |
2500* |
4.16.2 Workstation performance
The WS performances have been estimated using the CCS stress tests on a quiet HP Workstation of the 9000 family, model C110 running HP-UX 10.10 and the NOV96 release of the VLT software.
The stress test being still under development, the measurements do not yet address all values.
The tables will be completed when further values will become available.
Database Read Access Time
| Database read |
Access time (microseconds) |
| Symbolic address one level hierarchy (time depends also on the size of the database) scalar attribute |
160 |
| Vector attribute access |
190 |
| Table attribute access |
200 |
| Direct address scalar attribute |
30 |
| Direct address vector attribute |
33 |
| Direct address table attribute |
35 |
Message System, local Command/Reply
| Message system Action |
Time (microseconds) |
| Command sent to local process, small buffer (< 100 bytes) |
1300 |
| Reply returned to sending process, small buffer (< 100 bytes) |
280 |
| Reply returned to sending process, large buffer (8000 bytes) The figure corresponds to a data transfer rate of 3.2 MBytes/second |
930 |
