Wednesday Jun 17, 2009
With the pending buyout of Sun by Oracle still fresh in our minds, a recent email thread emerged wanting to put together a time-line of technologies created to help Sun servers run Oracle better. This seemed like a great way to start off the Oracle Performance portion of the Performance Technologies group new wiki. I hope you enjoy this overview of the key technologies used to optimize Oracle performance on Sun servers.
Monday Apr 20, 2009
As a Sun engineer who specializes on Oracle performance, I am pretty excited for the future. Sun and Oracle have been working together closely for years, but this will allow for even closer integration and innovation.
Could you imagine... "Dtrace probes for Oracle?" How cool would that be?
Time will tell innovations will come out of this acquisition, but I for one am pretty excited... Certainly, much more so than last week 
Heck, I now get to work with my old buddy Kevin Closson, the Exadata performance guru. Should be a fun ride ahead.
Tuesday Dec 09, 2008
Recently, I was asked to present beginning Oracle analysis techniques to an internal audience of Sun engineers. This presentation was a lot of fun to put together and was well received. After cleaning it up a bit and taking out the boring internal Sun stuff, I thought the presentation might be useful to a larger audience. This presentation focuses on problem statement, environmental, and basic AWR/Statspack analysis.
If you find this useful or have suggestions, drop me a note.
Saturday Nov 08, 2008
Several years ago, I wrote about how Oracle views multi-threaded processors. At the time we were just introducing a dual-core processor. This doubling of the number of cores was presented by Solaris as virtual CPUs and Oracle would automatically size the CPU_COUNT accordingly. But what happens when you introduce a 1RU server that has 128 virtual CPUs?
The UltraSPARC T1/T2/T2+ servers have many threads or virtual CPUs. The CPU_COUNT on these systems is sized no different than before. So, the newly introduced T5540 with 4xUltraSPARC T2+ processors would have 256 threads and CPU_COUNT would be set to 256.
So, what does CPU_COUNT have to do with memory?
Thanks to my friends in the Oracle Real World Performance group, I was made aware that Oracle uses CPU_COUNT to size the minimum amount of SGA allowed. In one particular case, the DBA was trying to allocate 70 database instances on a T5140 with 64GB of memory and 128 virtual CPUs. Needless to say, the SGA_TARGET would have to be set fairly low in-order to accomplish this task. A SGA_TARGET was set to 256MB, but the following error was encountered.
ORA-00821: Specified value of sga_target 256M is too small
After experimentation, they were able to start Oracle with a target of 900MB, but with 70 instances this would not fly. Manually lowering the CPU_COUNT allowed the DBA to use an SGA_TARGET of 256MB. Obviously, this is an extreme case and changing CPU_COUNT was reasonable.
Core and virtual CPU counts have been on the rise for some years now. Combine rising virtual CPU count with the current economic climate and I would suspect that consolidation will be more popular than ever. In general, I would not advocate changing CPU_COUNT manually. If you had one instance on this box, the default be just fine. CPU_COUNT automatically sizes so many other parameters that you should be very careful before making a change.
Monday Sep 22, 2008
The time has come present at Oracle Open World on UltraSPARC CMT performance. I decided to post the final version here in addition to the OOW site. I hope to see you there!
Session ID: S299785
Title: Growing Green databases with Oracle on the UltraSPARC CMT processor
Date: Monday Sept 22nd
Time: 13:00 - 14:00
Place: Moscone South Rm 236
Friday Jun 13, 2008
I am not one to post much in the way of personal information on my blog, but this exception is worth your time.
My son Hayden (13 years old) is participating in his first Olympic triathlon this year (1.5k swim / 28mi bike/ 10k run). He decided to make it a fund raiser for Pancreatic cancer in memory of his Grandmother who suffered from this most lethal of cancers. If you would like to help in anyway, please use this link to donate directly to the Lustgarten Foundation. The Lustgarten Foundation is a foundation dedicated to advance the scientific and medical research related to the diagnosis, treatment, cure and prevention of pancreatic cancer.
Below is an excerpt from Hayden's Letter writing campaign:
Did You Know…
• Pancreatic cancer is the deadliest of all major cancers.
• It kills 50% of those afflicted within 6 months and 97% of those afflicted within 5 years.
• There is no prevention, early diagnoses, or cure.
• 10% of all pancreatic cancers are inherited.
• In 2008 there were 37,680 new cases of pancreatic cancer in the U.S.
• In 2008 there were 34,290 deaths from pancreatic cancer in the U.S.
I chose to raise money for pancreatic cancer research because of its alarming death rate, and because my beloved grandmother, Nancy, passed away from this disease in mid August 2001, having been diagnosed only two months prior. Please help me to raise money for pancreatic cancer research.
Thursday May 29, 2008
This entry continues the Throughput Computing Series to show how a typical DSS operation can be optimized with CMT based servers. The "Create as Select" and "Insert into as Select" operations are quite common in DSS and OLTP environments as well. Unless parallelism is specified, Oracle will single thread these operations. To achieve optimal throughput, these operations can use parallel query and DML operations.
Results
I created a 20GB table on a T5240 server to serve as the source for the "Create as Select" (CAS) operations. The parallelism of the CAS operation was increased until the IO subsystem was maxed out. This resulted in a drop from 25 minutes with no parallelism to 2 minutes 40 seconds with 8 threads...thats nearly a 10x speedup by simply using parallelism built into Oracle!
This server was configured with just two HBAs, one for each the source and destination tables. This limited throughput of CAS operations to 127MB/sec, or one HBA. With this IO configuration, it took only 8 threads to reach maximum throughput. You should experiment to achieve maximum throughput of your IO configuration. If you suspect your IO configuration is not performing up to speed, look into doing some IO micro benchmarking to find the maximum throughput outside of Oracle. A topic for a later discussion
SQL syntax
The following shows how to use parallel DML and parallel query.
## Create as Select ##
##
SQL> alter session enable parallel dml;
SQL> create table abc
parallel (degree 32)
as
select /*+ parallel(gtest, 32) */ * from gtest;
## Insert as Select ##
##
SQL> alter session enable parallel dml;
SQL> insert /*+ parallel(abc,32) */
into abc
select /*+ parallel(gtest,32) */ * from gtest;
Wednesday May 21, 2008
One of the most common ways to improve SQL performance is the use of indexes. While Oracle does have a wide variety of indexes available, these tests focus on the most commonly used B-tree index.
On large tables it is important to ensure indexes get created in a timely fashion, that is why Oracle introduced several features to decrease index creation time:
- "unrecoverable"
This feature prevent the logging of intermediate steps of the index creation process. There is really no value to logging of intermediate steps. Index creation should be thought of as an atomic process - if it fails, you can always start over. If you create indexes as "unrecoverable" they won't be recoverable until a backup is performed on the target tablespace.
- "parallel"
This simply uses parallel query/dml to speed the creation of indexes.
The following index create statement shows how to use the "parallel" and "unrecoverable" features for index creation.
create index gtest_c1 on gtest(idname)
pctfree 30 parallel 64 tablespace glennf_i unrecoverable;
Results
The following test created an non-unique index on varchar(32) column of a 20GB table. Parallelism was increased from 1->64 in order to use the available IO bandwidth. With parallelism of 1 index creation took 34 minutes, while with parallelism of 64 it took only 3 minutes and 45 seconds!
These tests use the same configuration as previous posts regarding Oracle in the
Throughput Computing series.
Wednesday May 14, 2008
A backup window is important to keep in check to ensure time for batch and on-line work. With Oracle RMAN there are several ways to keep backups flowing smoothly. This example shows how you can use multiple channels and parallelism to increase the throughput of backup to the maximum of your IO configuration.
Results
This graph shows scaling in MB/sec based on the # of channels in use. The term "channels" used by Oracle does not have any relation to actual "physical" channels. In Oracle RMAN terms, a channel is simply a "connection" to a database for which to backup data. Data files are assigned to "connections" in a round-robin fashion so as to utilize all connections as evenly as possible.
By configuring a parallelism of 20 with RMAN, I was able to increase throughput from 5->80 MB/sec. Single threaded performance was limited to 5MB/sec mainly due to the high CPU component that comes with using "COMPRESSED" backups. The way to maximize IO throughput with COMPRESSION is to simply add more streams.
RMAN commands to achieve parallelism
I used the following commands to create 20 backup "channels" for RMAN. Notice that they configured to use the same directory, just with different file formats.
RMAN> configure channel 1 device type disk format
'/o6s_data/GLENNF/d2/backup_db_c1%d_S_%s_P_%p_T_%t' MAXPIECESIZE 1024 M;
RMAN> configure channel 2 device type disk format
'/o6s_data/GLENNF/d2/backup_db_c2%d_S_%s_P_%p_T_%t' MAXPIECESIZE 1024 M;
...
...
RMAN> configure channel 20 device type disk format
'/o6s_data/GLENNF/d2/backup_db_c20%d_S_%s_P_%p_T_%t' MAXPIECESIZE 1024 M;
After creating these channels, you must tell RMAN how to connect to these channels:
RMAN> configure channel 1 DEVICE TYPE DISK CONNECT '/as sysdba';
RMAN> configure channel 2 DEVICE TYPE DISK CONNECT '/as sysdba';
...
...
RMAN> configure channel 20 DEVICE TYPE DISK CONNECT '/as sysdba';
Next, you need to tell RMAN to use disk parallelism of 20:
RMAN> CONFIGURE DEVICE TYPE DISK BACKUP TYPE
TO COMPRESSED BACKUPSET PARALLELISM 20;
Finally, let's issue the backup command:
RMAN> BACKUP TABLESPACE GLENNF_RMAN;
Starting backup at 09-MAY-08
allocated channel: ORA_DISK_1
channel ORA_DISK_1: sid=966 devtype=DISK
allocated channel: ORA_DISK_2
channel ORA_DISK_2: sid=952 devtype=DISK
allocated channel: ORA_DISK_3
channel ORA_DISK_3: sid=940 devtype=DISK
allocated channel: ORA_DISK_4
channel ORA_DISK_4: sid=938 devtype=DISK
allocated channel: ORA_DISK_5
channel ORA_DISK_5: sid=939 devtype=DISK
allocated channel: ORA_DISK_6
channel ORA_DISK_6: sid=969 devtype=DISK
allocated channel: ORA_DISK_7
channel ORA_DISK_7: sid=961 devtype=DISK
allocated channel: ORA_DISK_8
channel ORA_DISK_8: sid=963 devtype=DISK
allocated channel: ORA_DISK_9
channel ORA_DISK_9: sid=953 devtype=DISK
allocated channel: ORA_DISK_10
channel ORA_DISK_10: sid=970 devtype=DISK
allocated channel: ORA_DISK_11
channel ORA_DISK_11: sid=920 devtype=DISK
allocated channel: ORA_DISK_12
channel ORA_DISK_12: sid=943 devtype=DISK
allocated channel: ORA_DISK_13
channel ORA_DISK_13: sid=968 devtype=DISK
allocated channel: ORA_DISK_14
channel ORA_DISK_14: sid=929 devtype=DISK
allocated channel: ORA_DISK_15
channel ORA_DISK_15: sid=960 devtype=DISK
allocated channel: ORA_DISK_16
channel ORA_DISK_16: sid=931 devtype=DISK
allocated channel: ORA_DISK_17
channel ORA_DISK_17: sid=927 devtype=DISK
allocated channel: ORA_DISK_18
channel ORA_DISK_18: sid=957 devtype=DISK
allocated channel: ORA_DISK_19
channel ORA_DISK_19: sid=958 devtype=DISK
allocated channel: ORA_DISK_20
channel ORA_DISK_20: sid=964 devtype=DISK
channel ORA_DISK_1: starting compressed full datafile backupset
channel ORA_DISK_1: specifying datafile(s) in backupset
input datafile fno=00068 name=/oracle/O6S/sapraw/glenn1
channel ORA_DISK_1: starting piece 1 at 09-MAY-08
channel ORA_DISK_2: starting compressed full datafile backupset
channel ORA_DISK_2: specifying datafile(s) in backupset
input datafile fno=00069 name=/oracle/O6S/sapraw/glenn2
channel ORA_DISK_2: starting piece 1 at 09-MAY-08
channel ORA_DISK_3: starting compressed full datafile backupset
channel ORA_DISK_3: specifying datafile(s) in backupset
input datafile fno=00070 name=/oracle/O6S/sapraw/glenn3
channel ORA_DISK_3: starting piece 1 at 09-MAY-08
channel ORA_DISK_4: starting compressed full datafile backupset
channel ORA_DISK_4: specifying datafile(s) in backupset
input datafile fno=00071 name=/oracle/O6S/sapraw/glenn4
channel ORA_DISK_4: starting piece 1 at 09-MAY-08
channel ORA_DISK_5: starting compressed full datafile backupset
channel ORA_DISK_5: specifying datafile(s) in backupset
input datafile fno=00072 name=/oracle/O6S/sapraw/glenn5
channel ORA_DISK_5: starting piece 1 at 09-MAY-08
channel ORA_DISK_6: starting compressed full datafile backupset
channel ORA_DISK_6: specifying datafile(s) in backupset
input datafile fno=00073 name=/oracle/O6S/sapraw/glenn6
channel ORA_DISK_6: starting piece 1 at 09-MAY-08
channel ORA_DISK_7: starting compressed full datafile backupset
channel ORA_DISK_7: specifying datafile(s) in backupset
input datafile fno=00074 name=/oracle/O6S/sapraw/glenn7
channel ORA_DISK_7: starting piece 1 at 09-MAY-08
channel ORA_DISK_8: starting compressed full datafile backupset
channel ORA_DISK_8: specifying datafile(s) in backupset
input datafile fno=00075 name=/oracle/O6S/sapraw/glenn8
channel ORA_DISK_8: starting piece 1 at 09-MAY-08
channel ORA_DISK_9: starting compressed full datafile backupset
channel ORA_DISK_9: specifying datafile(s) in backupset
input datafile fno=00076 name=/oracle/O6S/sapraw/glenn9
channel ORA_DISK_9: starting piece 1 at 09-MAY-08
channel ORA_DISK_10: starting compressed full datafile backupset
channel ORA_DISK_10: specifying datafile(s) in backupset
input datafile fno=00077 name=/oracle/O6S/sapraw/glenn10
channel ORA_DISK_10: starting piece 1 at 09-MAY-08
channel ORA_DISK_11: starting compressed full datafile backupset
channel ORA_DISK_11: specifying datafile(s) in backupset
input datafile fno=00078 name=/oracle/O6S/sapraw/glenn11
channel ORA_DISK_11: starting piece 1 at 09-MAY-08
channel ORA_DISK_12: starting compressed full datafile backupset
channel ORA_DISK_12: specifying datafile(s) in backupset
input datafile fno=00079 name=/oracle/O6S/sapraw/glenn12
channel ORA_DISK_12: starting piece 1 at 09-MAY-08
channel ORA_DISK_13: starting compressed full datafile backupset
channel ORA_DISK_13: specifying datafile(s) in backupset
input datafile fno=00080 name=/oracle/O6S/sapraw/glenn13
channel ORA_DISK_13: starting piece 1 at 09-MAY-08
channel ORA_DISK_14: starting compressed full datafile backupset
channel ORA_DISK_14: specifying datafile(s) in backupset
input datafile fno=00081 name=/oracle/O6S/sapraw/glenn14
channel ORA_DISK_14: starting piece 1 at 09-MAY-08
channel ORA_DISK_15: starting compressed full datafile backupset
channel ORA_DISK_15: specifying datafile(s) in backupset
input datafile fno=00082 name=/oracle/O6S/sapraw/glenn15
channel ORA_DISK_15: starting piece 1 at 09-MAY-08
channel ORA_DISK_16: starting compressed full datafile backupset
channel ORA_DISK_16: specifying datafile(s) in backupset
input datafile fno=00083 name=/oracle/O6S/sapraw/glenn16
channel ORA_DISK_16: starting piece 1 at 09-MAY-08
channel ORA_DISK_17: starting compressed full datafile backupset
channel ORA_DISK_17: specifying datafile(s) in backupset
input datafile fno=00084 name=/oracle/O6S/sapraw/glenn17
channel ORA_DISK_17: starting piece 1 at 09-MAY-08
channel ORA_DISK_18: starting compressed full datafile backupset
channel ORA_DISK_18: specifying datafile(s) in backupset
input datafile fno=00085 name=/oracle/O6S/sapraw/glenn18
channel ORA_DISK_18: starting piece 1 at 09-MAY-08
channel ORA_DISK_19: starting compressed full datafile backupset
channel ORA_DISK_19: specifying datafile(s) in backupset
input datafile fno=00086 name=/oracle/O6S/sapraw/glenn19
channel ORA_DISK_19: starting piece 1 at 09-MAY-08
channel ORA_DISK_20: starting compressed full datafile backupset
channel ORA_DISK_20: specifying datafile(s) in backupset
input datafile fno=00087 name=/oracle/O6S/sapraw/glenn20
channel ORA_DISK_20: starting piece 1 at 09-MAY-08
channel ORA_DISK_2: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c2O6S_S_81_P_1_T_654270132 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_2: backup set complete, elapsed time: 00:00:58
channel ORA_DISK_3: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c3O6S_S_82_P_1_T_654270132 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_3: backup set complete, elapsed time: 00:00:58
channel ORA_DISK_4: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c4O6S_S_83_P_1_T_654270132 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_4: backup set complete, elapsed time: 00:00:58
channel ORA_DISK_9: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c9O6S_S_88_P_1_T_654270133 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_9: backup set complete, elapsed time: 00:00:57
channel ORA_DISK_11: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c11O6S_S_90_P_1_T_654270133 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_11: backup set complete, elapsed time: 00:00:57
channel ORA_DISK_12: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c12O6S_S_91_P_1_T_654270133 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_12: backup set complete, elapsed time: 00:00:57
channel ORA_DISK_13: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c13O6S_S_92_P_1_T_654270133 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_13: backup set complete, elapsed time: 00:00:57
channel ORA_DISK_18: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c18O6S_S_97_P_1_T_654270134 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_18: backup set complete, elapsed time: 00:00:56
channel ORA_DISK_20: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c20O6S_S_99_P_1_T_654270135 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_20: backup set complete, elapsed time: 00:00:55
channel ORA_DISK_10: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c10O6S_S_89_P_1_T_654270133 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_10: backup set complete, elapsed time: 00:00:58
channel ORA_DISK_16: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c16O6S_S_95_P_1_T_654270134 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_16: backup set complete, elapsed time: 00:00:57
channel ORA_DISK_1: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c1O6S_S_80_P_1_T_654270132 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_1: backup set complete, elapsed time: 00:01:00
channel ORA_DISK_5: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c5O6S_S_84_P_1_T_654270132 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_5: backup set complete, elapsed time: 00:01:00
channel ORA_DISK_14: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c14O6S_S_93_P_1_T_654270134 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_14: backup set complete, elapsed time: 00:00:58
channel ORA_DISK_7: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c7O6S_S_86_P_1_T_654270132 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_7: backup set complete, elapsed time: 00:01:01
channel ORA_DISK_8: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c8O6S_S_87_P_1_T_654270132 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_8: backup set complete, elapsed time: 00:01:01
channel ORA_DISK_6: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c6O6S_S_85_P_1_T_654270132 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_6: backup set complete, elapsed time: 00:01:04
channel ORA_DISK_15: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c15O6S_S_94_P_1_T_654270134 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_15: backup set complete, elapsed time: 00:01:02
channel ORA_DISK_17: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c17O6S_S_96_P_1_T_654270134 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_17: backup set complete, elapsed time: 00:01:02
channel ORA_DISK_19: finished piece 1 at 09-MAY-08
piece handle=/o6s_data/GLENNF/d2/backup_db_c19O6S_S_98_P_1_T_654270135 tag=TAG20080509T134205 comment=NONE
channel ORA_DISK_19: backup set complete, elapsed time: 00:01:01
Finished backup at 09-MAY-08
Configuration
- T5240 - "Maramba" CMT based server
- 2 x UltraSPARC T2 Plus @ 1.4GHz
- 128GB memory
- 2 x 1GB fiber channel HBA
- 1 x 6140 Storage array with 1 lun per channel.
- Software
- Solaris 10 Update 5
- Oracle 10.2.0.3
- CoolTools
Wednesday May 14, 2008
I just finished reading an excellent blog entry by Andy Harrison that discusses the importance of a good problem statement. After working many problems with shady problem definitions, I will forward this to entry to all violators... So if we can't agree on the problem, we can all agree on the answer
Monday May 12, 2008
A common observation regarding Niagara based servers is that system maintenance or database administration tasks can run slower than previous generations of Sun servers. While single-threaded performance may be less, these maintenance tasks are often able to be parallelized, especially using a database engine as mature as Oracle.
Take for instance the task of gathering schema statistics. Oracle offers many options on how to gather schema statistics, but there are a few ways to reduce overall gather statistics time:
- Increased Parallelism
- Reduced Sample Size
- Concurrency
Oracle has written many articles in
metalink which discuss sample size and the various virtues. There have also been many volumes written on optimizing the Oracle cost based optimizer (CBO).
Jonathan Lewis of who is a member of the famous
Oaktable network has written books and multiple white papers on the topic. You can read these for insight into the Oracle CBO. While a reasonable sample size or the use of the "DBMS_STATS.AUTO_SAMPLE_SIZE" can seriously reduce the gather statistics times, I will leave that up to you to choose the sample size the produces the best plans.
Results
The following graph shows the total run time in seconds of a "GATHER_SCHEMA_STATS" operations at various levels of parallelism and sample size on a simple schema of 130GB. All tests were run on a Maramba T5240 with a 6140 array and two channels.
Note that if higher levels of sampling are required, parallelism can help to significantly reduce the overall runtime of the GATHER_SCHEMA_STATS operation. Of course a smaller sample size can be employed as well.
GATHER_SCHEMA_STATS options
SQL> connect / as sysdba
-- Example with 10 percent with parallel degree 32
--
SQL> EXECUTE SYS.DBMS_STATS.GATHER_SCHEMA_STATS (OWNNAME=>'GLENNF',
ESTIMATE_PERCENT=>10,
DEGREE=>32,
CASCADE=>TRUE);
-- Example with AUTO_SAMPLE_SIZE and parallel degree 32
--
SQL> EXECUTE SYS.DBMS_STATS.GATHER_SCHEMA_STATS (OWNNAME=>'GLENNF',
ESTIMATE_PERCENT=>DBMS_STATS.AUTO_SAMPLE_SIZE,
DEGREE=>32,
CASCADE=>TRUE);
Note that you must have "parallel_max_servers" set to at least the level of parallelism desired for the GATHER_SCHEMA_STATS operation. I typically set it higher to allow for other parallel operations to get servers.
SQL> alter system set parallel_max_servers = 128;
Finally, you can easily run a schema collect on multiple schema's concurrently and in parallel by issuing GATHER_SCHEMA_STATS from multiple sessions and ensuring the level of parallelism is set high enough to accommodate.
Configuration
Monday Apr 21, 2008
Since we just recently announced mutli-chip based CMT servers that provide up to 128 threads in a 1U or 2U box, it seems fitting to pick up this thread on throughput computing.
The key to fully appreciating the CMT architecture with Oracle is to exploit the available threads. As I have spoke about earlier in the "throughput computing series", this can be done through "concurrency", "parallelism", or both. Oracle, being the mature product that it is, can achieve high-levels of parallelism as well as concurrency.
Concurrent processing with Oracle
For examples of concurrent processing with Oracle, look at recent results on the Oracle Ebusiness payroll
benchmark. This shows that using concurrent processes to break up the batch, you can increase batch throughput. By going from 4 to 64 processes, batch time decreased from 31.53 minutes to 4.63 minutes and throughput was increased by 6.8x!
With Oracle's Ebusiness Suite of software, you can increase the number of "concurrent manager" processes to more fully utilize the available threads on the system. Each ISV has different ways of controlling batch distribution and concurrency. You will have to check with your various software vendors to find all the ways to increase concurrency.
Parallelism in Oracle
People often associate parallelism in Oracle with parallel query. In most cases where CMT is involved, I see a lack of understanding of how to achieve parallelism with more basic administrative functions. Oracle excels in providing parallelism for important administrative tasks.
Schema analyze
Index build/rebuild
Parallel loader
Parallel export/import with datapump
While parallelism exists for these administrative tasks, some configuration is required. I will examine the various ways to achieve optimal throughput with CMT based systems on these tasks.
Monday Mar 10, 2008
We recently ran across a pretty strange performance bug with the checksum function of Oracle. This bug (6814520) causes excessive CPU to be used for the checksum routine. The checksum feature of Oracle is enabled by the db_block_checksum=TRUE parameter in Oracle. With the release of Oracle 10gR2, "TRUE" is the default setting. The order of magnitude of CPU overhead depends on the type of Solaris SPARC architecture.
Chip %overhead
---- ---------
SPARC64 250%
USIV 45%
------------------
w/patch 8%
Oracle released a patch via metalink to address this situation. This patch is for 10.2.0.3 installations. The fix will be included in: 11.1.0.7 and 10.2.0.5. A patch for upcoming 10.2.0.4 is in the works.
If you are unsure whether or not you are hitting this bug, you can easily alter this parameter on the fly:
SQL> alter system set db_block_checksum=FALSE
Warning, this will disable the checksum feature and blocks written when this is turned to
FALSE will not contain checksum information.
Regardless of whether you are currently hitting the bug or not, the recommendation is:
INSTALL "6814520" WHEN USING ORACLE 10.2.0.3 !!!
Thursday Feb 14, 2008
I usually really dislike blog entries that have nothing to say other than repackage bug descriptions and offer them up as knowledge, but in this case I have made an exception since the full impact of the bug is not fully described.
There is a fairly nasty Oracle bug with 10.2.0.3 that prevents the use of DirectIO with Solaris. The metalink note
"406472.1" describes the failure modes but fails to mention the performance impact if you use "filesystemio_options=setall" and fail to have the mandatory patch "5752399"
in place.
This was particularly troubling to me since we have been recommending for years the use of the "setall" to ensure all the proper filesystem options are set for optimal performance. I just finished working a customer situation where this patch was not installed and their critical batch run-times were nearly 4x as large... Not a pretty situation.... OK, So bottom line:
MAKE SURE YOU INSTALL "5752399" WHEN USING ORACLE 10.2.0.3 !!!
Wednesday Jan 16, 2008
Most environments have some open source SW that is used as part of the application stack. Depending on the packages, this can take a fair amount of time to configure and compile. To speed the install process, parallelism can easily be used to take advantage of the throughput of CMT servers.
Let us consider the following five open source packages:
- httpd-2.2.6
- mysql-5.1.22-rc
- perl-5.10.0
- postgresql-8.2.4
- ruby-1.8.6
The following experiments will time the installation of these packages in both a serial, parallel, and concurrent fashion.
Parallel builds
After the "configure" phase is complete, these packages are all compiled using
gmake. This is where parallelism within each job can be used to speed the install process. By using the
"gmake -j" option, the level of parallelism can specified for each of the packages. This can dramatically improve the overall compile time as seen below.
- Jobs were ran in a serial fashion but with parallelism within the job itself.
- 79% reduction in compile time at 32 threads/job.
Concurrency and Parallelism
The build process for the various packages are not each able to be parallelized perfectly. In fact, the best overall gain of any of the packages is 6x out of 32. This is where concurrency comes into play. If we start all the compiles at the same time and use parallelism as well, this further reduces the overall build time.
- All 5 jobs were run concurrently with 1 and 32 (threads/job).
- 88% overall reduction in compile time from serial to parallel with concurrency.
- 42% reduction in compile time over parallel jobs ran serially.
Load it up!
Hopefully, this helps to better describe how to achieve better system throughput through parallelism and concurrency. Sun's CMT servers are multi-threaded machines which are capable of a high level of throughput. Whether you are building packages from source or installing pre-build packages, you have to load up the machine to see throughput.
|
|
|
|
