AWR – Profiling Database I/O

Oracle Awr (Automatic Workload Repository) statistics, captures and stores fine grained information about file reads and writes (aka i/o), that the database performed, during the course of execution of, application generated database workloads. When analyzing the read and write patterns of the database, it helps a lot to understand what type of activity is generating the reads and writes. With this stored information we can get an indepth understanding of the distribution of random and sequential reads and writes.

I use this information for getting a better understanding of the I/O profile, for my Exadata sizing exercises.

This information can be used to understand clearly how much of the i/o is from Temp activity, Datafile reads and writes, Archivelog writes, log writes, and whether these are small or large reads and writes.

To the best of my understanding the small reads and writes are those < 128k and the large reads and writes are those > 128k.

This information is contained mainly in two awr Views.

Dba_Hist_Iostat_FileType
Dba_Hist_Iostat_Function

Dba_Hist_Iostat_FileType

This view displays the historical i/o statistics by file type. The main filetypes are the following

Archive Log
Archive Log Backup
Control File
Data File
Data File Backup
Data File Copy
Data File Incremental Backup
Data Pump Dump File
Flashback Log
Log File
Other
Temp File

Dba_Hist_Iostat_Function

This view displays the historical i/o statistics by i/o function. The main i/o functions are the following

Recovery
Buffer Cache Reads
Others
RMAN
Streams AQ
Smart Scan
Data Pump
XDB
Direct Writes
DBWR
LGWR
Direct Reads
Archive Manager
ARCH

From everything i have seen sofar, these reads and writes can be directly co-related to the “Physical read total IO requests” and “Physical write total IO requests” system level statistics.

I have written a script that displays information from the above mentioned views and gives a detailed breakdown of i/o gendrated from different aspects of the database activities.
In order to fit in the computer screen real estate, i have actually limited the columns the script displays (So it displays only the file types i am frequently interested in). Please feel free to take the script and modify it to add columns that you want to display.

The full version of the script  awrioftallpct-pub.sql can be found here.

The script accepts the following inputs
– A begin snap id for a snapid range you want to report for
– A End snap id for a snapid range you want to report for
– A Dbid for the database
– The snap interval in seconds (If you have a 30 minute interval input 1800 seconds)

A description of all the column names in the output, broken down by section, is provided in the header section of the script.

There are 6 sections to this script

1) Total Reads + Writes
2) Total Reads
3) Total Writes
4) Read write breakdown for datafiles
5) Data File – Direct Path v/s Buffered Read Write breakdown
6) Read write breakdown for tempfiles

1) Total Reads + Writes

This section displays the number of reads+writes by filetype, and a percentage of reads+writes for each file type, as a percentage of total reads+writes. The last column displays the total reads+writes for all file types. The column DTDP shows the i/o that bypasses flash cache by default and goes directly to spinning disk on Exadata (Temp+Archivelogs+Flashback Logs).

io1-rw

Click on the image to see a larger version

2) Total Reads

This section displays the number of reads by filetype, and a percentage of reads for each file type, as a percentage of total reads. The last column displays the total reads for all file types.

io1-r

Click on the image to see a larger version

3) Total Writes

This section displays the number of writes by filetype, and a percentage of writes for each file type, as a percentage of total writes. The last column displays the total writes for all file types.

io1-w

Click on the image to see a larger version

4) Read write breakdown for datafiles

This section displays the I/O information only pertaining to datafile i/o. It displays the small and large reads and writes and a percentage they constitute of the total reads+writes to datafiles, and a percentage they constitute of the total reads or writes to datafiles. It also displays the total small and large reads and writes and a percentage they constitute of the total reads+writes to datafiles.

io1-dfrw

Click on the image to see a larger version

5) Data File – Direct Path v/s Buffered Read Write breakdown

This section provides a breakdown of I/O by function (As opposed to i/o by filetype in the previous sections). The output shows columns that display the direct path small and large reads and writes, buffered small reads and writes, smart scan small and large reads and other small and large reads and writes.

io3-bf

Click on the image to see a larger version

6) Read write breakdown for tempfiles

This section displays the I/O information only pertaining to tempfile i/o. It displays the small and large reads and writes and a percentage they constitute of the total reads+writes to tempfiles, and a percentage they constitute of the total reads or writes to tempfiles. It also displays the total small and large reads and writes and a percentage they constitute of the total reads+writes to tempfiles.

io3-tf

Click on the image to see a larger version

The full version of the script  awrioftallpct-pub.sql can be found here.

Oracle Active DataGuard – Considerations for the Wide area Network

Oracle customers use Oracle Active Dataguard to create and maintain one or many standby databases that protect their mission critical primary databases from disaster. Typically, in such deployments, the primary databases and standby databases are in geographically separate locations connected via a WAN (Wide Area Network). Log Transport Services, transfers Large volumes of redo logs from the primary location to the standby, using Sql*Net.

We have to ensure that all the components from the source to target are setup correctly to ensure that the data transfer can be done with the best throughput possible. If sufficient network bandwidth is not available with reasonable latencies, then we will start seeing the log transfer and apply,lagging on the standby site (Which is oracle speak for, your primary and standby database is now out of sync from a data perspective).

One key point to keep in mind is that, lower the network round trip time (aka latency), higher your data transfer throughput. Higher the network round trip time (aka latency), lower your data transfer throughput. So it is very important to maintain low round trip times on your Wide area network.

To understand network data transfer throughput, It is important to understand the the concepts of Tcp Window Size and Bandwidth Delay Product (aka BDP).

Tcp Window size is the amount of bytes that can be transmitted without receiving an acknowledgement from the other side. Once Tcp Window size amount of bytes are send, the sender stops sending any more bytes and waits for an acknowledgement from the receiver.

Bandwidth delay product is calculated as the product of the network bandwidth and network round trip time. bdp=network bandwidth*round trip time. This is the amount of data that left the sender before the first acknowledgement was received by the sender. If the senders output bandwidth is stable, and the bandwidth is fully used, then the BDP calculates the number of packets in transit. If we set the Tcp Window size equal to the bandwidth delay product, then in theory we should be able to fully utilize the available bandwidth.

Setup the network

We have to start by setting up the networking components to support the desired/stated bandwidth. So if you have a WAN that is a 10GigE network, all the NIC’s (Network interface cards), ports, switches in the configuration should be configured to support 10GigE full Duplex settings. After setup, run the configuration display utilities and ensure that all these component levels the transfer speeds are set to be 10 GigE. Customers often run into trouble when Auto Negotiation causes some NIC’s to set the transfer speeds to 1GigE because of configuration mismatches.

Use tools like Iperf to test the transfer speeds that your network is capable of achieving.

One important aspect to keep in mind is that it is probable that the WAN is shared by other traffic (e-mail, data replication, san replication). This has two important implications that we should consider.

  • If there is a lot of bandwidth consumption by some of this miscellaneous traffic, round trip times could be degraded periodically on the network.
  • We should be careful in our calculations that we do not completely consume the entire bandwidth for redo transport. (This could impact other processes)
    • So it is important to figure out (Working with the network admins) what the bandwidth entitlements are for redo transport and base our calculations on those numbers.

Caclulate our BDP

Use the following formula to calculate our Bandwidth delay product (BDP)

(bandwidth/8)*rount trip time in seconds.

The network bandwidth is expressed in bits per second, so we divide by 8 to convert to bytes.
Round Trip Time is usually in milli seconds, so we divide by 1000 to convert to seconds.

So for example, if we have a 10Gbit network bandwidth and a 40ms round trip time

BDP=(10000000000/8)*(40/1000) = 50,000,000 bytes.

Setup Sql*Net Parameters

The current recommendations for Dataguard Redo transport are as follows.

Set the SDU size to 65535

  • We can set SDU on a per connection basis using the SDU parameter in the local naming configuration file (TNSNAMES.ORA) and the listener configuration file (LISTENER.ORA)
  • We can set the SDU for all Oracle Net connections with the profile parameter DEFAULT_SDU_SIZE in the SQLNET.ORA file.

Set TCP.NODELAY to YES

To preempt delays in buffer flushing in the TCP protocol stack, disable the TCP Nagle algorithm by setting TCP.NODELAY to YES in the SQLNET.ORA file on both the primary and standby systems.

Setup RECV_BUF_SIZE and SEND_BUF_SIZE

The current recommendation is to set the SEND_BUF_SIZE and RECV_BUF_SIZE parameters (Which are the send and receive socket buffer sizes for SQL*Net) to 3 Times the BDP.

As per the above example we would set them to 50,000,000*3 = 150,000,000

Setup Operating system Kernel Parameters

If you are using the Linux operating system make sure that the values for the following kernel parameters are setup to be higher than the values set for RECV_BUF_SIZE and SEND_BUF_SIZE.

net.core.rmem_max
net.core.wmem_max

Once we have configured the network, operating system and the sql*net, and we have redo transfer, we can perform further network monitoring to see how the network bandwidth is being utilized, and make appropriate adjustments.

Links to helpful Documents

Iperf

How to calcluate Tcp throughput for long distance links (blog)

Oracle Net Performance Tuning (Mos)

Setting Send and Receive Buffer Sizes (Mos)

Tuning Sql*Net peformance (Oracle Docs)

Configuring Oracle Dataguard (Oracle Docs)

Best Practices for Sync Data Transport (White Paper)

Script to compare tkprof output files

I often use the oracle 10046 event tracing mechanism to capture sql’s from a session to identify why certain transactions are running slower in different env’s or at different points in time. Oracle does have a mechanism where you can save the trace information in database tables. One can use the INSERT parameter in tkprof to store the trace information into a database table. Once in the table you can write sql’s that compare multiple runs or multiple statements.

I wrote a python program that compares two different tkprof output files. The files are compared, and the following aspects of each of the sqlid’s in the tkprof output file,s are printed side by side. The output is sorted by the Difference in Elapsed Time, in Descending order.

  • Sql text
  • Plan Hash Value
  • Total Elapsed time
  • Total Logical Reads
  • Total Rows processed
  • Difference in Elapsed Time
  • Difference in Number of Rows processed
  • Difference in Logical reads.

Other columns can be added to this, if you desire.
I use this script output as a quick way to see which sql’s are running slower and are probably candidates for further analysis/tuning.

The sqlid’s from the file provided as the first argument to the script (referred to as the left file) are compared to the same sqlid’s in the file provided as the second argument to the script (referred to as the right file). The following columns are displayed.

sqlid                           sqlid being compared
text                             First 20 chars of the sql text
lplan                           Plan hash value from the left file
rplan                          Plan hash value from the right file
lela                             Total Elapsed time from the left file
rela                            Total Elapsed time from the right file
llreads                       Total Logical reads (query+current) from the left file
rlreads                      Total Logical reads (query+current) from the right file
lrows                         Total rows processed from the left file
rrows                        Total rows processed from the right file
eladiff                        Lela – Rela
rowsdiff                    Lrows – Rrows
lreadsdiff                  Llreads – rlreads

Here is a sample syntax for running the script. (You need the python pandas package to be installed for this to execute successfully)

python ./difftk.py /u01/tkprofout/Newplans.prf /u01/tkprofout/Stage.prf

Here is a sample output

difftk

Click on the image to view a larger version.

The full script is below

#Python script to list differences between sql executions in two tkprof output files
#useful if comparing tkprof from prod and dev for example
#Author : rajeev.ramdas

import sys
import os
import pandas as pd
from pandas import DataFrame

# Define a class to hold the counters for each sqlid
class sqliddet:
     def init(self):
            sqlid=''
            text=''
            plan_hash=''
            tcount=0
            tcpu=0
            tela=0
            tdisk=0
            tquery=0
            tcurr=0
            trows=0

# Define 2 empty dictionaries to store info about each input file
leftsqliddict={}
rightsqliddict={}

# Process each file and add one row per sqlid to the dictionary
# We want to add the row to the dictionary only after the SQLID row and the total row has been read
# So the firstsqlid flag is used to make sure that we do not insert before total is read for the first row.

def processfile(p_file,p_sqliddict):

    myfile=open(p_file,"r")
    line=myfile.readline()
    firstsqlid=True
    linespastsqlid=99
    while line:
        linespastsqlid+=1
        line=myfile.readline()
        if line.startswith('SQL ID'):
            linespastsqlid=0
            if firstsqlid==True:
                firstsqlid=False
            else:
                p_sqliddict[currsqlid.sqlid]=[currsqlid.plan_hash,currsqlid.tcount,currsqlid.tcpu,currsqlid.tela,currsqlid.tdisk,currsqlid.tquery
                            ,currsqlid.tcurr,currsqlid.trows,currsqlid.text]
            currsqlid=sqliddet()
            currsqlid.sqlid=line.split()[2]
            currsqlid.plan_hash=line.split()[5]
        if linespastsqlid==2:
            currsqlid.text=line[0:20]
        if line.startswith('total'):
            a,currsqlid.tcount,currsqlid.tcpu,currsqlid.tela,currsqlid.tdisk,currsqlid.tquery,currsqlid.tcurr,currsqlid.trows=line.split()
        if line.startswith('OVERALL'):
            p_sqliddict[currsqlid.sqlid]=[currsqlid.plan_hash,currsqlid.tcount,currsqlid.tcpu,currsqlid.tela,currsqlid.tdisk,currsqlid.tquery
                       ,currsqlid.tcurr,currsqlid.trows,currsqlid.text]
        continue
    myfile.close()

# Main portion of script
if len(sys.argv) != 3:
   print('Syntax : python ./difftk.py tkprof1.out tkprof2.out')
   sys.exit()

if not os.path.isfile(sys.argv[1]) or not os.path.isfile(sys.argv[2]):
   print ("File Does not Exist")
   sys.exit()

processfile(sys.argv[1],leftsqliddict)
processfile(sys.argv[2],rightsqliddict)

t_difftk_lst=[]

# Match the sqlid's from the file on the left to the file on the right
# Gather up the statistics form both sides, insert into a list
# Transfer the list to a pandas dataframe, add some computed columns

for sqlid,stats in leftsqliddict.items():
    l_totlogical=int(stats[5])+int(stats[6])
    if sqlid in rightsqliddict:
       t_difftk_lst.append([sqlid,stats[8].rstrip(),stats[0],rightsqliddict[sqlid][0],float(stats[3])
                            ,float(rightsqliddict[sqlid][3]),float(l_totlogical),float(rightsqliddict[sqlid][5])+float(rightsqliddict[sqlid][6])
                            ,float(stats[7]),float(rightsqliddict[sqlid][7])
                           ])
    else:
       t_difftk_lst.append([sqlid,stats[8].rstrip(),stats[0],0,float(stats[3]),0
                            ,float(l_totlogical),0,float(stats[7]),0
                           ])

difftk_df=DataFrame(t_difftk_lst,columns=['sqlid','sqltext','lplan','rplan','lela','rela','llreads','rlreads','lrows','rrows'])
difftk_df['eladiff']=difftk_df['lela']-difftk_df['rela']
difftk_df['rowsdiff']=difftk_df['lrows']-difftk_df['rrows']
difftk_df['lreadsdiff']=difftk_df['llreads']-difftk_df['rlreads']

pd.set_option('display.width',1000)
print (difftk_df.sort(columns='eladiff',ascending=False))

Using Pandas for CSV Analysis

As Oracle Dba’s we often come across situations where we are handed CSV (Comma separated values) files, by our managers, or customers, as Raw data, based on which we need to do some work. The first task would be to analyze the file and come up with some summary satistics, so we can quantify the amount of work involved.

When faced with such circumstances, my favorite method is to use sqlloader to upload the file into a database table, and then run sql statements on the data to produce summary info. Most people would probably use excel, formulas, macros and pivot tables to achieve similar results.

In this blog post, i present an alternate method that i’ve been using recently, for csv file summarization.

Pandas is a library written for the Python language for data manipulation and analysis. In order to proceed, first install Python and then install the Python package named ‘pandas’. Pandas is a real good alternative to the R programming language.See my previous post on how to install python and pandas.

For the examples in this post, i am using a Csv file, that has NFL game, play by play statistics for 2014.

Start by invoking the python interactive interpreter.

 

     $ python3
Python 3.4.2 (default, Dec 18 2014, 14:18:16) 
[GCC 4.8.2] on linux
Type "help", "copyright", "credits" or "license" for more information.

First import the following libraries that we are going to use.

     
import pandas as pd
import numpy as np

Read the csv file into a Pandas DataFrame

 

df=pd.DataFrame(pd.read_csv('pbp-2014.csv',header=0))

Check how many rows the dataframe has

 
df.size
2056275

List the columns in the DataFrame

 

list(df)
['GameId', 'GameDate', 'Quarter', 'Minute', 'Second', 'OffenseTeam', 'DefenseTeam', 'Down', 'ToGo', 'YardLine', 'Unnamed: 10', 'SeriesFirstDown', 'Unnamed: 12', 'NextScore', 'Description', 'TeamWin', 'Unnamed: 16', 'Unnamed: 17', 'SeasonYear', 'Yards', 'Formation', 'PlayType', 'IsRush', 'IsPass', 'IsIncomplete', 'IsTouchdown', 'PassType', 'IsSack', 'IsChallenge', 'IsChallengeReversed', 'Challenger', 'IsMeasurement', 'IsInterception', 'IsFumble', 'IsPenalty', 'IsTwoPointConversion', 'IsTwoPointConversionSuccessful', 'RushDirection', 'YardLineFixed', 'YardLineDirection', 'IsPenaltyAccepted', 'PenaltyTeam', 'IsNoPlay', 'PenaltyType', 'PenaltyYards']

Check how many games the dataset covers. The GameId column is a unique identifier that identifies each game. The nunique method returns the number of unique elements in that object.

 

df.GameId.nunique()
256

List details of all games played by the New England Patriots. This command shows how you can provide a filter condition. The filter specifies that all the rows, where the OffensiveTeam is NE or DefensiveTeam is NE be listed.

 

df[(df['OffenseTeam'] == 'NE') | (df['DefenseTeam'] == 'NE')]

Subset a specific set of columns

 

df[['GameId','PlayType','Yards']]

           GameId     PlayType  Yards
0      2014090400     KICK OFF      0
1      2014090400         RUSH      6
2      2014090400         RUSH      3
3      2014090400         RUSH     15
4      2014090400         RUSH      2
5      2014090400         PASS     -2
...
...

Display all Pass and Rush plays Executed by New England. Here we are using a and filter condition to limit the rows to those of New England and the PlayType is either a PASS or a RUSH.

 

df[['GameId','PlayType','Yards']][(df['OffenseTeam'] == 'NE') & (df['PlayType'].isin(['PASS','RUSH']))]   
           GameId PlayType  Yards
1092   2014090705     RUSH      2
1093   2014090705     PASS      4
1094   2014090705     PASS      0
1102   2014090705     PASS      8
1103   2014090705     PASS      8
1104   2014090705     RUSH      4
     

Display the Number of Plays, Total Yards Gained, and Average Yards gained per PASS and RUSH play, per game.

 

df[['GameId','PlayType','Yards']][(df['OffenseTeam'] == 'NE') & (df['PlayType'].isin(['PASS','RUSH']))].groupby(['GameId','PlayType']).agg({'Yards': [np.sum,np.mean],'GameId':[np.size]}) 
                    Yards           GameId
                      sum      mean   size
GameId     PlayType                       
2014090705 PASS       277  4.540984     61
           RUSH       109  5.190476     21
2014091404 PASS       209  8.038462     26
           RUSH       158  4.157895     38
2014092105 PASS       259  6.641026     39
           RUSH        91  2.935484     31
2014092900 PASS       307  9.903226     31
           RUSH        75  4.687500     16
2014100512 PASS       301  7.921053     38
           RUSH       223  5.309524     42
2014101201 PASS       407  9.465116     43
           RUSH        60  2.307692     26
2014101600 PASS       267  6.675000     40
           RUSH        65  4.062500     16
2014102605 PASS       376  9.400000     40
           RUSH       121  3.781250     32
2014110208 PASS       354  6.210526     57
           RUSH        66  2.640000     25
2014111611 PASS       267  8.343750     32
           RUSH       248  6.048780     41
2014112306 PASS       393  6.894737     57
           RUSH        90  4.285714     21
2014113010 PASS       245  6.621622     37
           RUSH        85  5.000000     17
2014120713 PASS       317  6.604167     48
           RUSH        80  3.333333     24
2014121408 PASS       287  7.972222     36
           RUSH        92  3.407407     27
2014122105 PASS       189  5.250000     36
           RUSH        78  4.333333     18
2014122807 PASS       188  5.222222     36
           RUSH        92  4.181818     22


From the above example’s you can see how easy it is to read a csv file, apply filters and summarize the data set using pandas.

Oracle Database In-Memory an introduction Part 2 – What do i need to do, to use the Oracle In-Memory Database ?

Step 1) Define the INMEMORY_SIZE

Customer has to setup the correct value for a database initialization parameter , INMEMORY_SIZE. This parameter specifies the amount of memory, from the SGA, that is to be used for the In-Memory column store. This is a static pool (ie Automatic memory management cannot extend or shrink this area), which means that you have to restart the database if any changes to this parameter needs to take effect. The In-Memory area is sub-divided into two pools: a 1MB pool used to store the actual column formatted data populated into memory, and a 64K pool used to store metadata about the objects that are populated into the IM column store.

 

sho parameter inmemory_size

NAME				     TYPE	 VALUE
------------------------------------ ----------- ------------------------------
inmemory_size			     big integer 500M

select pool,alloc_bytes,used_bytes,populate_status from v$inmemory_area;

POOL			   ALLOC_BYTES USED_BYTES POPULATE_STATUS
-------------------------- ----------- ---------- --------------------------
1MB POOL		     418381824		0 DONE
64KB POOL		     100663296		0 DONE

Step 2) Mark the performance critical objects in your database, with the attribute INMEMORY

select partition_name,bytes/(1024),inmemory,inmemory_compression from dba_segments where
owner = 'SH' and segment_name = 'SALES'

PARTITION_NAME		       BYTES/(1024) INMEMORY INMEMORY_COMPRESS
------------------------------ ------------ -------- -----------------
SALES_Q1_1998			       8192 DISABLED

ALTER TABLE SH.SALES MODIFY PARTITION SALES_Q1_1998 INMEMORY PRIORITY CRITICAL MEMCOMPRESS FOR QUERY HIGH;

Table altered.

select partition_name,bytes/(1024),inmemory,inmemory_compression from dba_segments where
owner = 'SH' and segment_name = 'SALES'  2  ;

PARTITION_NAME		       BYTES/(1024) INMEMORY INMEMORY_COMPRESS
------------------------------ ------------ -------- -----------------
SALES_Q1_1998			       8192 ENABLED  FOR QUERY HIGH


Step 3) Populate the In-Memory datastore

Objects are populated into the In-Memory Datastore, in a prioritized list, immediately after the database is opened, or after they are scanned for the first time. There are 7 levels for the keyword PRIORITY (CRITICAL, HIGH, MEDIUM, LOW, NONE).

The IM column store is populated by a set of background processes referred to as worker processes (ora_w001_orcl). The database is fully active / accessible while this occurs.Each worker process is given a subset of database blocks from the object to populate into the IM column store. Population is a streaming mechanism, simultaneously columnizing and compressing the data. There is a new IMCO (In memory co-ordinator) background process which wakes up every 2 minutes and checks to see if there are any population tasks that need to be completed. Eg: A new object has been marked as InMemory with a PRIORITY other than None.

select v.owner,v.segment_name,v.partition_name,v.bytes orig_size,v.inmemory_size in_mem_size

OWNER	   SEGMENT_NA PARTITION_NAME		      ORIG_SIZE IN_MEM_SIZE
---------- ---------- ------------------------------ ---------- -----------
SH	   SALES      SALES_Q1_1998			8388608     1179648

select * from 
(
select /*+ full(sales) */ channel_id,count(*)
from sh.sales partition (sales_q1_1998)
group by channel_id
order by count(*) desc
)
where rownum < 6

CHANNEL_ID   COUNT(*)
---------- ----------
	 3	32796
	 2	 6602
	 4	 3926
	 9	  363

Elapsed: 00:00:00.09

select * from table (dbms_xplan.display_cursor());

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------
SQL_ID	40pjk921r3jrc, child number 0
-------------------------------------
select * from ( select /*+ full(sales) */ channel_id,count(*) from
sh.sales partition (sales_q1_1998) group by channel_id order by
count(*) desc ) where rownum < 6

Plan hash value: 2962696457

---------------------------------------------------------------------------------------------------------
| Id  | Operation			| Name	| Rows	| Bytes | Cost (%CPU)| Time	| Pstart| Pstop |
---------------------------------------------------------------------------------------------------------

PLAN_TABLE_OUTPUT
---------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT		|	|	|	|    12 (100)|		|	|	|
|*  1 |  COUNT STOPKEY			|	|	|	|	     |		|	|	|
|   2 |   VIEW				|	|     4 |   104 |    12  (34)| 00:00:01 |	|	|
|*  3 |    SORT ORDER BY STOPKEY	|	|     4 |    12 |    12  (34)| 00:00:01 |	|	|
|   4 |     HASH GROUP BY		|	|     4 |    12 |    12  (34)| 00:00:01 |	|	|
|   5 |      PARTITION RANGE SINGLE	|	| 43687 |   127K|     9  (12)| 00:00:01 |     5 |     5 |
|   6 |       TABLE ACCESS INMEMORY FULL| SALES | 43687 |   127K|     9  (12)| 00:00:01 |     5 |     5 |
---------------------------------------------------------------------------------------------------------

For much more in-depth technical details of the Oracle Database In-Memory, please see this whitepaper.

Oracle Database In-Memory an introduction Part 1 – What is Oracle Database In-Memory, and how is it different ?

On July 22nd 2014, Oracle corporation, announced Oracle Database 12c’s latest patch release 12.1.0.2. This latest patch release includes the new Oracle Database In-Memory functionality.
The Oracle Database In-Memory enables a single database to efficiently support mixed workloads. It uses a “dual-format” architecture, that retains the Record-Setting OLTP performance of the oracle databases, while simultaneously supporting real-time analytics and reporting. This is achieved by retaining the traditional oracle memory architecture, but adding a new purely in-memory column format (Automatically created and maintained by oracle), optimized for analytical processing.

So now you have the data stored in the Oracle database in your database files, in a row format, and for any of the objects marked as INMEMORY, oracle creates an In-Memory column store, where the data resides in a column format.The IM column store does not replace the buffer cache, but acts as a supplement, so that data can now be stored in memory in both a row and a column format.

Ok why the dual format, one would ask ?

The Row format is retained as is, so that there is no compromise/degradation in the OLTP performance of the database. In a Row format database each row is made up of multiple columns, with each column representing an attribute about that record. A column format database stores each of the attributes about the transaction in a separate column structure. A column format is ideal for Analytics, but is not very efficient in processing DML requests like insert, update and deletes (Which operates on the whole row). Oracle Database In-Memory (Database In-Memory) provides the best of both worlds by allowing data to be simultaneously populated in both an in-memory row format (the buffer cache) and a new in-memory column format (The In-Memory Store).

cncpt_vm_379

The picture above shows the In-Memory area in the SGA

cncpt_vm_378

The above picture shows an example of the Column Store.

No changes are necessary to your existing applications in order to take advantage of the Oracle Database In-Memory option. Any query that will benefit from the In-Memory column store will be automatically directed there, by the Optimizer. The In-Memory store is kept transactionally consistent with the buffer cache. There are numerous optimizations that have been implemented that speed up this data access in the In-Memory store. All the database functionality that Oracle has built over the last 30 years, continues to work in this new version.

It is normal that folks would go out and compare, Oracle Databaes In-Memory, with other In Memory Database products available in the Market today. So let us explore the differences with some of those products next.

TimesTen

Oracle TimesTen In-Memory Database is a,row oriented, memory optimized relational database, that resides entirely in the physical memory of a server. Oracle TimesTen In-Memory Database operates on databases that fit entirely in physical memory using standard SQL interfaces. In contrast, the Oracle Database In-Memory, only stores selected objects, in memory, in a column format (As opposed to a row format in Oracle TimesTen), the remaining objects continue to exist on the storage subsystem.

SAP HANA

SAP HANA is an in-memory, column-oriented relational database. So the entire database has to reside in physical memory of the server. As mentioned earlier in this article, OLTP transactions have some disadvantages while operating on column stores.

Todays databases can be in the 10’s or 100’s of TeraBytes. Storing this entire data in physical memory can be an expensive proposition, not to mention that, it is quite prevalent that ,only a small subset of this data, ends up having daily usage. This is where Oracle’s approach of storing only selected objects in memory, has significant benefits.

Here is a link to the Oracle Database In-Memory Launch.

Exadata Database Machine Specifications – Quick Reference

With the different combinations of the Oracle Exadata database machines and storage expansion racks available to customers, it is hard for me to remember, each system’s specifications.

So i have created a JavaScript page, that can be used to quickly lookup the specifications of the current generation (x3-8 and x4-2) of Exadata database machines and storage expansion racks.

The page can be accessed from a web browser from your desktop,laptop, or favorite mobile device at http://dbastreet.com/exaspecs

You can narrow the results down by the exadata rack size, type of disk, and type of specification (cpu, storage or performance).

Here is a screenshot

exaspecs

Setting up a Python 2.7.6 Virtual Env for Python development

Python is an excellent language to learn, for DBA’s who want to automate all the repetitive tasks, they need to perform. Once you start using python it is likely that you want to setup multiple environments with different versions of Python and libraries, based on the project you are working on. Virtualenv is a tool to create isolated python environments.

Below are the steps that i followed, to install a brand new working Python 2.7.6 environment with the following packages.

 
           SQLAlchemy    - Object Relational Mapper and SQL Toolkit for Python
           numpy         - Fundamental package for scientific computing
           matplotlib    - Python 2D plotting library
           ipython       - Interactive Python Shell
           pandas        - Python Data Analysis Library 
           Flask         - An easy to use Python Lightweight Micro Framework

This installation is performed on Ubuntu Linux, and i have already installed the libsqlite3-dev package and the oracle instant client.

Install Python 2.7.6

Download Python-2.7.6.tgz from http://www.python.org/getit

Install python 2.7.6 to your directory of choice.

		tar -xvf Python-2.7.6.tgz
		cd Python-2.7.6/

		./configure --prefix=/u01/Rk/Apps/Python/Python276
		make
		make install

Now you have python 2.7.6 installed into the /u01/Rk/Apps/Python/Python276 directory. (You will have a python binary in /u01/Rk/Apps/Python/Python276/bin)

Download virtualenv

curl -O https://pypi.python.org/packages/source/v/virtualenv/virtualenv-1.11.1.tar.gz

Install virtualenv

tar -xzvf virtualenv-1.11.1.tar.gz
cd virtualenv-1.11.1/
/u01/Rk/Apps/Python/Python276/bin/python virtualenv.py /u01/Rk/Apps/Python/p276env1

Activate the virtualenv

           . /u01/Rk/Apps/Python/p276env1/bin/activate

Install the additional Python Modules you need

	   pip install SQLAlchemy
	   pip install numpy
	   pip install matplotlib
	   pip install ipython
	   pip install pyzmq
	   pip install tornado
	   pip install jinja2
	   pip install pandas
	   pip install Flask
	   pip install Flask-SQLAlchemy

Install cx_Oracle

Ensure that the oracle instant client is installed, and the environment variables ORACLE_HOME and LD_LIBRARY_PATH are setup correctly.

Download cx_Oracle (a python extension module that allows access to oracle.) source from from http://cx-oracle.sourceforge.net/

tar -xzvf cx_Oracle-5.1.2.tar.gz
cd  cx_Oracle-5.1.2/
python setup.py install

Setup an alias (In your .bash_profile) to simplify invoking the virtualenv every time you want to use it.

alias p276env1='. /u01/Rk/Apps/Python/p276env1/bin/activate'

Now, anytime you want to execute a python program in this environment, you can invoke the Linux command line and

p276env1
python

OraChk Collection Manager

OraChk (Previously known as Raccheck) is a utility from oracle to perform configuration checks on Oracle database platforms, and report on configurations that do not match oracle’s best practices. OraChk has the ability to upload the results of its checks into an oracle database. Details about this utility can be found in Mos Note 1268927.1

Oracle has now released  OraChk Collection Manager which is a companion application to OraChk, which has an Oracle Application Express, Front End which can be used  as a dashboard, in which customers can track their ORAchk, RACcheck and Exachk collection data in one easy to use interface.

Details about downloading and using “OraChk Collection Manager” can be found in Mos Note 1602329.1

Exadata X4-2 Whats New – A Technical Review

 

On Dec 11 2013 Oracle corporation announced the general availability of, the 5th generation of the Oracle Exadata Database Machine X4. This new version of the database machine introduces new hardware and software to accelerate performance, increase capacity, and improve efficiency and quality of service for database deployments.

In this blog post i’ll review in detail the enhancements in the Oracle Exadata x4-2 database machine.

The 2 socket version of the Database machine, X4-2 gets new database servers and new storage servers. The 8 socket version of the database machine, still called the x3-8, gets the new storage servers but keeps the same database serves from the previous release. Hence the Lack of change in Name to the X4-8. The name stays as X3-8.

Improvements in the database servers (X4-2)

Cpu Improvements

The database servers in the x4-2 are the Sun X4-2  servers. They come with the 2 Twelve-Core Intel® Xeon® E5-2697 v2 Processors (Ivy Bridge) (2.7GHz), with turbo boost enabled by default. This cpu at times can clock upto 3.5Ghz.

Comparing this with the previous version x3-2, this is a 50% increase in number of cores per database server. The x3-2 had 16 cores per node and the x4-2 has 24 cores per node. In a x3-2 full rack there were 128 cores and in the x4-2 full rack there are 192 cores. That gives 64 more cores in the x4-2 full rack compared to the x3-2 full rack.

Memory Improvements

Each database server has 256Gb of Dram. This is optionally expandable to 512Gb per node.So you can have either 2Tb or 4Tb of Ram in a full rack of x4-2.

Storage Improvements

The database servers now have 4, 600Gb hard disks in them (Internal storage, used for O/S, Oracle Binaries, Logs etc). The x3-2 used to have 300Gb disks. The disk controller batteries are online replaceable

Network Improvements

The database servers have 2 X Infiniband 4 X QDR (40Gb/S) Ports (PCIe 3.0). Both ports are Active. The improvements in the x4-2 are that these are now PCIe 3.0, and that the ports are used active active. (The ports were used Active/Passive in the previous release)

Improvements in the storage servers (Exadata cells x4-2)

Cpu Improvements

Each x4-2 exadata cells come with 2, Six-Core Intel® Xeon® E5-2630 v2 Processors (2.6 GHz).

Memory Improvements

Each x4-2 cell, now has 96Gb of Ram. The x3-2 used to have 64 Gb of Ram.

Flash Cache Improvements

The Exadata x4-2 cells, now have the F80 PCIe Flash cards. There are 4, 800Gb F80 flash cards in each cell. So each cell has 3.2Tb of flash cache. In an Oracle Exadata database machine x4-2 full rack there is 44TB of flash cache (Used to be 22Tb in the x3-2). The Oracle Exadata database machine x4-2 full rack now provides, 2.66Million read iops , 1.96Million write iops from the flash cache. This is 70% more flash iops than the previous generation.

Storage Improvements

The Exadata cell x4-2 High Performance version, now uses 1.2Tb 10k Rpm High Performance SAS disks.

  • A full rack of High Performance disks gives 200TB of raw space, 90TB of usable space with normal mirroring and 60TB of usable space with High Mirroring

The Exadata cell x4-2 High Capacity version, now uses 4Tb 7.2k Rpm High Capacity SAS disks.

  • A full rack of High Capacity disks gives 672TB of raw space, 300TB of usable space with normal mirroring and 200TB of usable space with High Mirroring

The Disk controller batteries are online replacable.

Network Improvements

The Exadata storage servers have 2 X Infiniband 4 X QDR (40Gb/S) Ports (PCIe 3.0). Both ports are Active. The improvements in the x4-2 are that these are now PCIe 3.0, and that the ports are used active active. (The ports were used Active/Passive in the previous release)

 

Software Improvement Highlights

The new x4-2’s get a new version of the exadata storage server software version 11.2.3.3.0 (This version of the cell software can be installed on the v2,x2,x3 and x4 platforms).

Exadata Flash Cache Improvements

Data written to the flash cache is now compressed. This compression is done by the controller (Hence near Zero overhead). This data is automatically decompressed when it is read from the flash cache. Up to 2X more data fits in smart flash cache (If you get 2x compression of the data, you could fit upto 80TB of data on the flash cache), so flash hit rates will improve and performance will improve for large data sets. This feature can be turned on in both x3-2 and x4-2 storage servers. Customers need to license the Advanced Compression option to use this new feature.

Enhancements to the smart flash cache software, enables exadata software to understand database table and partition scans and automatically caches them when it makes sense (This will help eliminate the need to specify CELL FLASH CACHE KEEP).

Exadata network resource management

With a new version of the Infiniband switch firmware 2.1.3-4, Exadata network resource management now prioritizes messages through the entire infiniband fabric. Latency sensitive messages like redo log writes are prioritized over batch, reporting and backup messages.

Infiniband Active Active Ports

Double-ported infiniband PCIe-3.0 Cards used in the database servers and storage servers, implement active-active mode usage of the infiniband ports, providing a 80GigaBits Per Second network bandwidth (Used to be 40Gigbits ber second on the x3-2, since it was Active Passive bonding).

The Rdbms software and the clusterware software already had the ability to send packets via multipe interfaces. Enhancements have been done to the RDS kernel drivers, which now have the ability to sense if one of the ports is down, and route the network traffic through the surviving port. On the x4-2’s when active active Infiniband networking is setup you will not see the bondib0 interface, instead you will see a ib0 and ib1.

Miscellaneious Info

  • The power,cooling and airflow requirements remain similar to that of the x3-2.
  • The storage expansion racks have also been refreshed and provides the increased flash cache and increased disk space.
  • A single Database Machine configuration can have servers and storage from different generations V2, X2, X3, X4.
  • Databases and Clusters can span across multiple hardware generations.
  • The half and full x4-2 racks, do not ship the spine switch anymore. The storage expansion racks still ship with the spine switch.
  • A new One Command available through patch 17784784 has the new Exadata Deployment assistant that supports the x4-2’s.

Datasheets and White Papers

Datasheet: Oracle Exadata database machine x4-2

Datasheet: Oracle Exadata database machine x3-8

Datasheet: Oracle Exadata storage expansion Rack x4-2

Datasheet: Oracle Exadata storage server x4-2

A Technical Overview of the Oracle Database Machine and Exadata Storage Server (Updated for x4-2)

Exadata Smart Flash Cache (Updated For X4-2)