RE: Oracle Win NT Configuration Standards

From: Satar Naghshineh <Satar.Naghshineh_at_irvine.mellesgriot.com>
Date: Thu, 21 Sep 2000 17:05:17 -0700
Message-Id: <10626.117614@fatcity.com>

This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible.

------_=_NextPart_001_01C02428.D041DA30
Content-Type: text/plain;

charset="iso-8859-1"

Thanks Chris for the reply, and no, I didn't perceive your initial response to be negative. After taking the time to read Sean's paper (thanks to the person who posted it on the list), I noticed it was a standard install he wanted his company to use for their clients. I do not know if his company has a standard application and this document is based on this application or if this document was a general reference.

Anyway...
1. Agreed, but I thought the documentation came with the software. I have two NT databases (one Intel and one Dec Alpha) and they both came with the OFA documentation.

2. Agreed. The ideal scenario is to have one I/O operation have it's own dedicated Controller, Disk Bus and as many disks as feasible to separate the operation into many disk spindles. If this is a read process, then this should happen only once and any future read I/O operations should be read from memory. By the way, if anyone has this type of set-up please let me know and I'll send you my resume. ;)

3. That I understand. I was referring to the attempt to standardize things like "The system tablespace must not be greater than 500MB", which I thought he might have stated, but did not. His paper did a good job on detailing this aspect.

4. Agreed. Here's one for you (using your example from the document). If I had 500MB redo logs that were frequently changing and/or some other accessed data that was residing on the same disk, wouldn't be better to have them run on RAID 0+1 than on no RAID at all?

I was asked to create a Documented procedure on Oracle installs (basically what Sean is trying to accomplish) and I couldn't make my boss understand that every install is different because of the application(s) and the business requirements.

The very fact that Sean included the "deviation" section implies that this document cannot be a standard, but only a reference. It was a good paper and it did touch on some important theories, but whether these theories are applicable in every install, is the question on hand.

Regards,
Satar

P.S. No one knows everything about Oracle, if they claim that they do, I will have to question their sanity (or lack of social life). ;)

> -----Original Message-----
> From: Bowes, Chris [SMTP:Chris.Bowes_at_kosa.com]
> Sent: Thursday, September 21, 2000 3:35 PM
> To: Multiple recipients of list ORACLE-L
> Subject: RE: Oracle Win NT Configuration Standards
>
> Hi Satar.
>
> Just a few thoughts. First, no, your answer didn't sound mean. I
> hope my original reply to yours (as well as this one) didn't either. To
> discuss your points:
>
> 1. yes OFA addresses naming issues. However, an OFA paper for NT has been
> a scarce commodity. A chunk of his paper was on OFA for NT and what it
> meant to/for him.
>
> 2. Yes, you would think that common sense does dictate that you spread
> things out. Unfortuneately there are too many times when I have dealt
> with hardware folks who can get two 30g drives cheaper than six 10g drives
> and say "There, you have 60g. Go to town!" and think they did me a
> favor... His paper dealt with generalities like "keep indexes separate
> from table data" and keep system separate from tables and indexes" and
> used a 7 disc approach for his example. Based upon that approach, this
> should be applicable to most scenarios. Granted not all. But when you
> get down to a 2 or 3 disc option, you've pretty much killed the IO
> performance anyway.
>
> 3. You can make some iron-clad rules and enforceable ideals. His paper
> was saying to put tables in a tablespace named DATA##_<SID>.dbf, put
> indexes into INDX##_<SID>.dbf. Put Snapshots into a tablespace of it's
> own, nothing but system stuff in system. Things like that, you can
> standardize. You publish it as a standard and make people justify not
> following it. As DBA's we are here to be the gatekeeper to make sure user
> tables are not created in the system tablespace or that indexes don't go
> in the same tablespace as the table it was created for.
>
> 4. Agreed that RAID levels cannot be made standard for everything, but
> this isn't what Sean was doing. He was doing what you said at the end of
> #4. He was listing a file type (like redo logs) and listing the preferred
> RAID solution for them in order of desirability (For redo logs, he listed
> no raid first, then raid 0, then raid 1 then raid 5).
>
>
> I understand that no 2 installs are alike and that dynamic solutions to
> the situations at hand are needed. However, you can publish a set of
> guidelines and say, "if you violate them, you had better have a good
> reason". That's what separates us from the average Joe. We know what we
> are doing (at least we think we do :) ) and we enforce what we believe to
> be in the best interests of the system at hand. If we can't set a
> standard or do what is right for the system we are admin'ing, then we're
> just those wierd people who mutter at their screens and squirrel tapes for
> rainy days... The key paragraph to Sean's paper was the opening one: "The
> purpose of this document is to list the configuration standard guidelines
> which Organon (Ireland) Ltd., hereafter referred to as O(I)L, will
> normally apply to installations of Oracle software on it's servers. A
> secondary function of the document is to provide vendors in advance with a
> view to incorporating these standards during planning phase of projects
> where the application software requires Oracle." Basically "this is my
> companies standard and since I am telling you in advance, you had better
> follow it or have a real good reason why you didn't."
>
>
> I highly recommend Sean's excellent paper for reading. Even if you don't
> think it applies to you, it is good thinking material. Take care
> everyone. Have a good day.
>
>
>
> --Chris
> Chris.Bowes_at_Kosa.com
>
>
>
> -----Original Message-----
> From: Satar Naghshineh
> [mailto:Satar.Naghshineh_at_irvine.mellesgriot.com]
> Sent: Wednesday, September 20, 2000 9:40 PM
> To: Multiple recipients of list ORACLE-L
> Subject: RE: Oracle Win NT Configuration Standards
>
>
>
> I must admit, I am at a disadvantage because I didn't read his
> paper. But here are some questions I have in regards to the ability to
> standardize installations:
>
> 1. Doesn't OFA address issues in naming conventions?
>
> 2. Doesn't common sense dictate that you want to spread I/O out
> evenly? How can you have a standard on "file locations" when I might have
> two hard drives and someone else have 20?
>
> 3. "what goes in which tablespace" Assuming that you are meaning
> "objects in Tablespaces", cannot be standardized, not even SYSTEM.
>
> 4. "what RAID levels are preferred", There is no standards in RAID
> preference. One might choose not to run on a RAID system. Sean might tell
> us of the different RAID systems and the pro and cons of them based on the
> type of database of a certain set of data, but he can't say if you run an
> OLTP database you should run RAID 50.
>
> Again, naming conventions are covered by OFA, if you need a copy,
> order it from Oracle. There can be no standard database install. If you
> believe in a standard base install, you probably believe in an answer to
> someone's question on "How can I improve my performance on Oracle" or "How
> do I backup my database".
>
> I've been known to sound mean at times, and I want to let you know
> that is not my intention. My argument is that there can be no standard
> base install. If there was a way, then every Joe would be an Oracle DBA
> and I wouldn't have to work so hard.
>
> Regards,
> Satar
>
> P.S. No offense to Sean either, I don't know the guy, but I
> appreciate his kind intentions.
>
>
>

------_=_NextPart_001_01C02428.D041DA30
Content-Type: text/html;

charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =

charset=3Diso-8859-1">
<META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version = 5.5.2650.12">

<TITLE>RE: Oracle Win NT Configuration Standards</TITLE>
</HEAD>
<BODY>

Thanks Chris for the = reply, and no, I didn't perceive your initial response to be negative. = After taking the time to read Sean's paper (thanks to the person who = posted it on the list), I noticed it was a standard install he wanted = his company to use for their clients. I do not know if his company has = a standard application and this document is based on this application = or if this document was a general reference.

Anyway... 1. Agreed, but I = thought the documentation came with the software. I have two NT = databases (one Intel and one Dec Alpha) and they both came with the OFA = documentation.

2. Agreed. The ideal = scenario is to have one I/O operation have it's own dedicated = Controller, Disk Bus and as many disks as feasible  to separate = the operation into many disk spindles. If this is a read process, then = this should happen only once and any future read I/O operations should = be read from memory. By the way, if anyone has this type of set-up = please let me know and I'll send you my resume.  ;)

3. That I = understand. I was referring to the attempt to standardize things like = "The system tablespace must not be greater than 500MB", which = I thought he might have stated, but did not. His paper did a good job = on detailing this aspect.

4. Agreed. Here's = one for you (using your example from the document). If I had 500MB redo = logs that were frequently changing and/or some other accessed data that = was residing on the same disk, wouldn't be better to have them run on = RAID 0+1 than on no RAID at all?

I was asked to = create a Documented procedure on Oracle installs (basically what Sean = is trying to accomplish) and I couldn't make my boss understand that = every install is different because of the application(s) and the = business requirements.

The very fact that = Sean included the "deviation" section implies that this = document cannot be a standard, but only a reference. It was a good = paper and it did touch on some important theories, but whether these = theories are applicable in every install, is the question on hand. =

<P><FONT COLOR=3D"#0000FF" SIZE=3D2 FACE=3D"Arial">Regards,</FONT>
<BR><FONT COLOR=3D"#0000FF" SIZE=3D2 FACE=3D"Arial">Satar</FONT>
</P>

P.S.  No one = knows everything about Oracle, if they claim that they do, I will have = to question their sanity (or lack of social life). ;)

<BR>
<BR>
<UL>
<P><FONT SIZE=3D1 FACE=3D"Arial">-----Original Message-----</FONT>
<BR><B><FONT SIZE=3D1 FACE=3D"Arial">From:&nbsp;&nbsp;</FONT></B> <FONT =

SIZE=3D1 FACE=3D"Arial">Bowes, Chris [SMTP:Chris.Bowes_at_kosa.com] Sent:   Thursday, September 21, 2000 3:35 PM To:     Multiple recipients of list ORACLE-L Subject:       = RE: Oracle Win NT Configuration = Standards

<P><FONT COLOR=3D"#0000FF" SIZE=3D2 FACE=3D"Arial">Hi Satar.=A0</FONT>=20
<BR><FONT FACE=3D"Arial">=A0</FONT>
<BR><FONT COLOR=3D"#0000FF" SIZE=3D2 FACE=3D"Arial">=A0=A0=A0=A0=A0Just =

a few thoughts.=A0 First, no, your answer didn't sound mean.=A0 I hope = my original reply to yours (as well as this one) didn't either.=A0 To = discuss your points:

=A0
 1. yes OFA = addresses naming issues.=A0 However, an OFA=A0paper for NT has been a = scarce commodity.=A0 A chunk of his paper was on OFA for NT and what it = meant to/for him.=A0=A0

=A0
 2.=A0 Yes, you = would think that common sense does dictate that you spread things = out.=A0 Unfortuneately there are too many times when I have dealt with = hardware folks who=A0can get=A0two 30g drives cheaper than=A0six=A010g = drives and say "There, you have=A060g.=A0 Go to town!" and = think they did=A0me=A0a favor...=A0 His paper dealt with generalities = like "keep indexes separate from=A0 table data" and keep = system separate from tables and indexes" and used a 7 disc = approach for his example.=A0 Based upon that approach, this should be = applicable to most scenarios.=A0 Granted not all.=A0 But when you get = down to a 2 or 3 disc option, you've pretty much killed the IO = performance anyway.=A0

=A0
 3.=A0=A0You can = make some iron-clad rules and enforceable ideals.=A0 His paper was = saying to put tables in a tablespace named DATA##_<SID>.dbf, put = indexes into INDX##_<SID>.dbf.=A0 Put Snapshots into a tablespace = of it's own, nothing but system stuff in system.=A0 Things like that, = you can standardize.=A0 You publish it as a standard and make people = justify not following it.=A0 As DBA's we are here to be the gatekeeper = to make sure user tables are not created in the system tablespace or = that indexes don't go in the same tablespace as the table it was = created for.=A0

=A0
 4.=A0 Agreed that = RAID levels cannot be made standard for everything, but this isn't what = Sean was doing.=A0 He was doing what you said at the end of #4.=A0 He = was listing a file type (like redo logs) and listing the preferred RAID = solution for them in order of desirability (For redo logs, he listed no = raid first, then raid 0, then raid 1 then raid 5).=A0

<P><FONT FACE=3D"Arial">=A0</FONT>
<BR><FONT FACE=3D"Arial">=A0</FONT>
<BR><FONT COLOR=3D"#0000FF" SIZE=3D2 FACE=3D"Arial">I understand that =

no 2 installs are alike and that dynamic solutions to the situations at = hand are needed.=A0 However, you can publish a set of guidelines and = say, "if you violate them, you had better have a good = reason".=A0=A0 That's what separates us from the average Joe.=A0 = We know what we are doing (at least we think we do :)=A0 ) and we = enforce what we believe to be in the best interests of the system at = hand.=A0 If=A0we can't set a standard or do what is right for=A0the = system we are admin'ing, then we're just those wierd people who mutter = at=A0their screens and squirrel tapes for=A0rainy days...=A0 = The=A0key=A0paragraph to Sean's paper was the opening=A0one: "The = purpose of this document is to list the configuration standard = guidelines which Organon (Ireland) Ltd., hereafter referred to as = O(I)L, will normally apply to installations of Oracle software on it's = servers.=A0 A secondary function of the document is to provide vendors = in advance with a view to incorporating these standards during planning = phase of projects where the application software requires = Oracle."=A0=A0 Basically "this is my companies standard and = since I am telling you in advance, you had better follow it or have a = real good reason why you didn't."

<P><FONT FACE=3D"Arial">=A0</FONT>
<BR><FONT FACE=3D"Arial">=A0</FONT>
<BR><FONT COLOR=3D"#0000FF" SIZE=3D2 FACE=3D"Arial">I highly recommend =

Sean's excellent paper for reading.=A0 Even if you don't think it = applies to you, it is good thinking material.=A0 Take care everyone.=A0 = Have a good day.

=A0

--Chris 
Chris.Bowes_at_Kosa.com

<P><FONT FACE=3D"Arial">=A0</FONT>
</P>
<UL>
<P><FONT SIZE=3D2 FACE=3D"Times New Roman">-----Original =

Message----- 
From: Satar Naghshineh [<A = HREF=3D"mailto:Satar.Naghshineh_at_irvine.mellesgriot.com">mailto:Satar.Nag= hshineh_at_irvine.mellesgriot.com</A>] 
Sent: Wednesday, September 20, 2000 9:40 = PM 
To: Multiple recipients of list = ORACLE-L 
Subject: RE: = Oracle Win NT Configuration Standards

<BR>
</FONT>
</P>

I must admit, I am = at a disadvantage because I didn't read his paper. But here are some = questions I have in regards to the ability to standardize = installations:

1. Doesn't OFA = address issues in naming conventions?=20

2. Doesn't common = sense dictate that you want to spread I/O out evenly? How can you have = a standard on "file locations" when I might have two hard = drives and someone else have 20?

<P><FONT COLOR=3D"#0000FF" SIZE=3D2 FACE=3D"Arial">3. &quot;what goes =
in which tablespace&quot; Assuming that you are meaning &quot;objects =
in Tablespaces&quot;, cannot be standardized, not even SYSTEM.</FONT> =

4. "what RAID = levels are preferred", There is no standards in RAID preference. = One might choose not to run on a RAID system. Sean might tell us of the = different RAID systems and the pro and cons of them based on the type = of database of a certain set of data, but he can't say if you run an = OLTP database you should run RAID 50.

Again, naming = conventions are covered by OFA, if you need a copy, order it from = Oracle. There can be no standard database install. If you believe in a = standard base install, you probably believe in an answer to someone's = question on "How can I improve my performance on Oracle" or = "How do I backup my database".

I've been known to = sound mean at times, and I want to let you know that is not my = intention. My argument is that there can be no standard base install. = If there was a way, then every Joe would be an Oracle DBA and I = wouldn't have to work so hard.

Regards, Received on Thu Sep 21 2000 - 19:05:17 CDT