Infrastructure at your Service

Last day I did some tests with Oratcptest. This a command line tool which can be used to measure network bandwidth and network latency between client and server.
In this blog I am going to show how to use this tool in a DataGuard environment.
I am using a DataGuard environment with following configuration.


Configuration - PROD

  Protection Mode: MaxAvailability
  Members:
  PROD_SITE1 - Primary database
    PROD_SITE2 - Physical standby database
    PROD_SITE3 - Physical standby database

Fast-Start Failover: DISABLED

Configuration Status:
SUCCESS   (status updated 54 seconds ago)

DGMGRL> show database 'PROD_SITE1' LogXptMode;
  LogXptMode = 'SYNC'
DGMGRL> show database 'PROD_SITE2' LogXptMode;
  LogXptMode = 'SYNC'
DGMGRL> show database 'PROD_SITE3' LogXptMode;
  LogXptMode = 'SYNC'
DGMGRL>

The installation of Oratcptest is very simple. We just have to download on both servers the oratcptest.jar file from the oracle support site. Note that JRE 6 or higher is required. In my case I have JRE 8 on both servers

[oracle@primaserver oratcptest]$ java -version
openjdk version "1.8.0_111"
OpenJDK Runtime Environment (build 1.8.0_111-b15)
OpenJDK 64-Bit Server VM (build 25.111-b15, mixed mode)
[oracle@primaserver oratcptest]$

[oracle@standserver1 oratcptest]$ java -version
openjdk version "1.8.0_181"
OpenJDK Runtime Environment (build 1.8.0_181-b13)
OpenJDK 64-Bit Server VM (build 25.181-b13, mixed mode)
[oracle@standserver1 oratcptest]$

[oracle@standserver2 ~]$ java -version
openjdk version "1.8.0_181"
OpenJDK Runtime Environment (build 1.8.0_181-b13)
OpenJDK 64-Bit Server VM (build 25.181-b13, mixed mode)
[oracle@standserver2 ~]$

We can invoke the help command to see all options available for the oratcptest

[oracle@primaserver oratcptest]$ java -jar oratcptest.jar -help

Now we can assess the network bandwidth for our DataGuard. Note that I am using simple virtual machines. But the steps will be the same on real productions servers.
We first have to determine the highest volume of redo log in my database. Following Oracle query can be used.

select thread#,sequence#,blocks*block_size/1024/1024 MB,(next_time-first_time)*86400 sec, blocks*block_size/1024/1024/((next_time-first_time)*86400) "MB/s" from v$archived_log where ((next_time-first_time)*864000) and first_time between  to_date('2018/08/09 08:00:00','YYYY/MM/DD HH24:MI:SS') and to_date('2018/08/09 11:00:00','YYYY/MM/DD HH24:MI:SS') and dest_id=2 order by first_time;

   THREAD#  SEQUENCE#         MB        SEC       MB/s
---------- ---------- ---------- ---------- ----------
         1        124 .003417969          9 .000379774
         1        125 .271972656        184 .001478112
         1        126 .001953125          5 .000390625
         1        127 11.3662109        915 .012422088
         1        128 10.8466797       6353 .001707332

We can see that the highest value is .012422088 MB/s. The goal is to see if our network bandwidth can support this rate.
As we are using SYNC mode, the primary database will wait for a confirmation from standby databases that they have written the change to disk before informing the application of the commit success.
For SYNC transport we then have to collect the Average redo write size which is calculated using following formula

Average=redo size / redo writes

These metrics can be obtained using an AWR report. In our case the value is

Average=15924844/4015=3967

Now we are going to simulate SYNC writes over the network using Oratcptest. Note I need the location of my standby redo logs

SQL> select member from v$logfile where type='STANDBY';

MEMBER
--------------------------------------------------------------------------------
/u01/app/oracle/oradata/PROD/stredo01.log
/u01/app/oracle/oradata/PROD/stredo02.log
/u01/app/oracle/oradata/PROD/stredo03.log
/u01/app/oracle/oradata/PROD/stredo04.log

From the standby server I can run following command

[oracle@standserver1 oratcptest]$ java -jar oratcptest.jar -server -port=5555 -file=/u01/app/oracle/oradata/PROD/myoratcp.tmp
OraTcpTest server started.

From the primary server

[oracle@primaserver oratcptest]$ java -jar oratcptest.jar standserver1 -port=5555  -write  -mode=sync -length=3967 -duration=10s -interval=2s
[Requesting a test]
        Message payload        = 3967 bytes
        Payload content type   = RANDOM
        Delay between messages = NO
        Number of connections  = 1
        Socket send buffer     = (system default)
        Transport mode         = SYNC
        Disk write             = YES
        Statistics interval    = 2 seconds
        Test duration          = 10 seconds
        Test frequency         = NO
        Network Timeout        = NO
        (1 Mbyte = 1024x1024 bytes)

(14:26:16) The server is ready.
                    Throughput             Latency
(14:26:18)      1.124 Mbytes/s            3.375 ms   (disk-write 2.537 ms)
(14:26:20)      0.813 Mbytes/s            4.668 ms   (disk-write 3.775 ms)
(14:26:22)      1.094 Mbytes/s            3.467 ms   (disk-write 2.773 ms)
(14:26:24)      1.004 Mbytes/s            3.778 ms   (disk-write 2.991 ms)
(14:26:26)      0.560 Mbytes/s            6.779 ms   (disk-write 5.623 ms)
(14:26:26) Test finished.
               Socket send buffer = 11700 bytes
                  Avg. throughput = 0.920 Mbytes/s
                     Avg. latency = 4.126 ms (disk-write 3.280 ms)

[oracle@primaserver oratcptest]$

We can see that the Average throughput is 0.920 M/s which is sufficient to handle our highest peak rate which is .012422088 MB/s. We can also note the latency which includes the time to send the message to the server host, the optional disk write at the server host, and the acknowledgment back to the client process
If we are using ASYNC mode the test will be like

[oracle@primaserver oratcptest]$ java -jar oratcptest.jar standserver1 -port=5555    -mode=async -length=3967 -duration=10s -interval=2s
[Requesting a test]
        Message payload        = 3967 bytes
        Payload content type   = RANDOM
        Delay between messages = NO
        Number of connections  = 1
        Socket send buffer     = (system default)
        Transport mode         = ASYNC
        Disk write             = NO
        Statistics interval    = 2 seconds
        Test duration          = 10 seconds
        Test frequency         = NO
        Network Timeout        = NO
        (1 Mbyte = 1024x1024 bytes)

(14:58:03) The server is ready.
                    Throughput
(14:58:05)     13.897 Mbytes/s
(14:58:09)      5.193 Mbytes/s
(14:58:11)     40.007 Mbytes/s
(14:58:13)     21.475 Mbytes/s
(14:58:13) Test finished.
               Socket send buffer = 210600 bytes
                  Avg. throughput = 16.901 Mbytes/s

[oracle@primaserver oratcptest]$

Conclusion
In this blog we have talked about Oratcptest which is a simple tool which can help to assess our network bandwidth. Oratcptest can help before performing Rman Backups, file copies, a DataGuard environment.
Reference : Oracle Doc ID 2064368.1

 

Leave a Reply

Mouhamadou Diaw
Mouhamadou Diaw

Consultant