Large RAC cluster advice needed - 16 nodes for 120,000 TPS?

So we are doing some modeling on a system that we are getting ready to build. Our current estimate indicates we will need to plan for about 120K transactions per second. Another guy did the research and thinks we need to plan to have 16 nodes of IBM 3850s which will be 256 cores and about 256gigs of ram to support the TPS. This sounded a little over the top so I decided to do some of my own research and I found:
http://www.oracle.com/us/corporate/press/018144
And that indicates a HP with two Intel® Xeon® x5570 quad-core 2.93 GHz processors was able to do 10,529 TPS. This makes me feel like the hardware that was suggested (sixteen x3850s) is way overkill.
Does anyone have any ideas or whitepapers on this sort of planning?
Edited by: bolesj on May 7, 2009 10:42 AM

Hi,
Did u thought about using database machine (Oracle/Hp)?
http://www.oracleracsig.org/pls/apex/f?p=105:206:3383923510465954::NO
Or take o brief look on webseminar presented by Bohan Chen (Yahoo). He uses a RAC system with 16 nodes.
http://www.oracleracsig.org/pls/apex/f?p=105:206:3383923510465954::NO
Regards,
Rodrigo Mufalani
http://mufalani.blogspot.com

Similar Messages

Advice needed on monitor for Photoshop and Lightroom use

Hi. I am aserious amature photographer wishing to move to the next level and sell some of my work.
I just had a custom pc built to work with the new copies of Photoshop CS5 and Lightroom 3 I bought (lots or RAM and HD space, ssd, etc....)
The last piece of my system is to purchase a monitor. I want to be somewhere in the better then Best Buy but less than NEC/Eizo range in price, or between $500 - $900. I have worked with cameras since the early 80's and moved to digital several years ago but the only post processing I have done is with Photoshop Elements. I would be doing mostly prints to sell but also need to have a web site to do so. Will also use the pc for daily net surfing... but do not game or watch a lot of video on the pc.
Being really new to this whole process I have a few questions.
The first thing I need to decide is whether I need to look for a wide gamut display or not.
I realize the whole chain must be 10 bit (Adobe -OS - graphics driver - graphics card - display port.
I have Adobe Photoshop CS5 and Lightroom 3, Win7 64 bit, Zotac ZT 50701 10M video card (which uses GeForce GTX 560 fermi and an nvidia chipset. It does have displayport). I am having a hard time determining whether my video card actually supports wide gamut (10bit).
Standard vs Wide Gamut? Is wide gamut important enough to deal with the issues it brings (calibration, viewing things other than PS and LR or color managed, which appear to be rare?) Is sRGB good enough for most prints (don't do fine art, mostly nature and portraits but starting to do some HRD things). If wide gamut is the way to go I have no problem with that and have the time to learn about calibration, color management, etc... But I also want to make sure the juice is worth the squeeze.
24" vs 27"? Is there any advantage to one or the other when editing photos?
IPS vs PLS? I realize they are similar but are there differences worth noting?
Glossy vs Matte Anti-Glare? seems to be a lot of comments regarding the anti-glare coating, mostly poor. Yet I can see issues using a glossy screen in my study with a window to my back.
One manufacturer vs another? I realize Eizo, NEC and LaCie are at the top of the heap. But with my budget, after upgrading my pc and camera equipment, I can't make that work now. So I need to choose from the next group down (Dell, HP, Samsung, Asus...)
One or two monitors? It looks like many (mid-grade) wide gamut monitors do a lousy job of displaying anything but color managed sites. Is that necessarily true of all the mid-grades? Or can some be used for graphics but as well for routine net surfing, MS Office, etc...? Or am I better off getting two monitors, one for graphics and one for the rest? That would pretty much limit me to 24" or less given my budget (used to using a Dell 21" TN monitor that oddly crapped out just as my new pc was done).
The more I read reviews the more confusing it gets. There seems to be a difference of opinion even among pros on whether to go wide gamut or stick with an easier sRGB. Realizing a standard gamut monitor would be cheaper, I do want to make the right decision up front, given my budget.
The one thing I have found astounding is that there is nowhere to actually see many of the monitors I am considering. We live in Nashville TN but my wife is from Atlanta Ga so we drove there a few weeks ago to visit family and for me to visit monitor shops. Even the largest ones there (Fry's and Microcenter) had minimal IPS monitors, a few Dells and HP's. The knowledge of their sales folks was so poor I finally gave up. Felt bad about this until I posted this on another board and got a reply from a guy in LA (second largest city in the US) that he wanted to see a particular monitor and there was no place even there to do so.
Anyhow, here is what I have considered:
24" Wide Gamut: Dell U2410 and Asus PA246Q. Dells appear to be good IF you get a good one. The Asus appears to be a clone of the Dell that gets a lot of good press.
27" Wide Gamut: Dell U2711 that also gets a ton of good reviews
24: Standard Gamut: Dell U2412 and HP ZR2440.
27" Standard Gamut: Samsung S27A850D and Apple Cinema- The Samsung uses PLS technology versus IPS while the Apple is a glossy screen that will work with a pc.
Sorry for the long post. Any comments are greatly appreciated.

dkg62 wrote:
I realize the whole chain must be 10 bit
Not trying to talk you out of setting up a 10 bit pipeline, but it's still not very mature, and it really isn't a necessity to get a good editing experience.
Personally I find advantage in using two 4:3 ratio monitors for Photoshop work. All my panels are on the right monitor, while pretty much the entire left one shows the Photoshop main window and the working canvas space. My desktop is 3200 x 1200 pixels overall, and I find having the panels remain visible all the time is important.
Regarding whether a wide gamut is important... Will you be printing to devices that deliver a wide gamut? What other things will you be doing with your system?
It's not a no-brainer whether a wider gamut monitor is always "better" for everything, since it can accentuate the differences between the output from color-managed and non-color-managed applications, and it's definitely true that not everything is color-managed. With a monitor that's close to sRGB, for example, you might find Internet Explorer output acceptable, while using a wide gamut monitor will result in garishly oversaturated IE displays. On the other hand, FireFox (with a settings tweak) seems to get color management right, so there is an alternative.
I think, as John has implied above, you should work to get your head completely around how color-management works, soup to nuts. If you don't, there will always be things that are a mystery or which surprise you at the wrong times. Being able to order a print and have it come back with the expected color can be very important, as you might imagine.
-Noel

SNMP on Standard Edition RAC, do I need to pay for a license

Hi,
I want to enable SNMP on Standard Edition RAC.
Is OEM free to install on Standard Edition ?
Does OEM need a specific Management Pack for SNMP ?
Thanks
john

What I do know is that you need at least one Management Pack to be allowed to use e-mail notification.
I guess this is the same when talking about SNMP notification.
To be sure contact your Oracle Sales Rep.
Regards
Rob
http://oemgc.wordpress.com

Help/advice Needed! Paid for 2 years with no servi...

I have been with BT for 2 years since we moved into our first home, paying approx £40 a month for phone line and broadband! When we initially signed up we were asked various questions to see what product would be best for us, we told them we wouldn't be using phone line as we use our mobiles and would only use the wifi for childrens iPad! We were told to download BT app and log into wifi using email and password which we have done for 2 years! The wifi has always been very slow Nd hit and miss but we live in middle of no where so thought it was our area! After 2 years we have decided to start using home phone for childrens school purposes! So we bought phone, plugged it in and no dialling tone! Called BT who sent out engineer within a week! When engineer arrived he asked where is our hub? My response being whats a hub? After explaining how we usually connect to wifi he explained that that's not our wifi and a hotspot! So basically been paying for over 2 years never used land line or Internet due no fault of our own! We were never told about a hub or received one? The engineer told me to call and complain so which I did over 3 weeks ago! The advisor says he would fill in a form and pass on to his manager, but sent is a hub out which is charged us for? The hub has came and I still can't get any internet after spending another 2 hours on phone doing various tests they can't get it working and are sending out another engineer and was told could possibly be charged £129 for! As you can imagine I am furious with? I still haven't heard back from anyone and have sent a complaints form in also! Can anyone give me any advice on where I stand as I'm not very good with technology? Thanks

Hi Emmaburke87,
Welcome to the community forum. I'll be able to take a look at the details of your complaint. Please can you send me in your details using the "Contact The Mods" link found in my profile.
Thanks
PaddyB
BTCare Community Mod
If we have asked you to email us with your details, please make sure you are logged in to the forum, otherwise you will not be able to see our ‘Contact Us’ link within our profiles.
We are sorry but we are unable to deal with service/account queries via the private message(PM) function so please don't PM your account info, we need to deal with this via our email account :-)

URGENT ADVICE NEEDED!- outputting for iTunes

I have edited some promo videos for a small record company. They have asked me to supply them edited masters at highest quality to submit to iTunes for sale. They were shot in HDV. What should I give them? Full quality quick time movies? Compress for iPod within FCP? Welcome your advice!
Thanks

I think you best give them both. The smallest file size to post on the i-tunes store and the high quality uncompressed to provide to a client via ftp-server. Thats the way i do it.

7.1 SLD-LMBD. Advice needed: duplicated entries for SAP systems

Hi Experts,
I've noticed few duplicate entries that appear in SOLMAN 7.1 during the content synchronization from SLD to LMDB and migration from SMSY to LMDB.
For instance:
I already had CRM syetem in Solution Manager: CRD
After the SLD-SMSY-LMDB sync & mirgation new CRM system was created in SOLMAN: CRD00001. and it has absolutely the same setting and parameters as an old one: CRD.
This happens for other systems like: BW, XI, etc..
Any advice please how to avoid such duplications in the SOLMAN systems' landscape?
Thanks & regards,
123

Hi,
This normally happenes if there is some differences in the data in SMSY and SLD.
It may be any header data..Database host name ot any small entry.
Please check the below note:
1507548 - Duplicate SID in Transaction SMSY.
The solution can be, cleanup the duplicate entries manually and update the data properly. check if any thing missing in the SMSY definition.
Hope this helps.
Regards,
Sunny

Diff hardwares for 2 node rac cluster

Dear All,
Can i implement RAC 2 node cluster on 2 nodes with diff hardwares like diff server models?
Regards

hungry_dba wrote:
Dear All,
Can i implement RAC 2 node cluster on 2 nodes with diff hardwares like diff server models?
Hi,
Also you can read tech note below:
*RAC: Frequently Asked Questions [ID 220970.1]*
Can I have different servers in my Oracle RAC? Can they be from different vendors? Can they be different sizes?
Oracle RAC does support a cluster with nodes that have different hardware configurations. An example is a cluster with 3 nodes with 4 CPUs and another node with 6 CPUs. This can easily occur when adding a new node after the cluster has been in production for a while. For this type of configuration, customers must consider some additional features to get the optimal cluster performance. The servers used in the cluster can be from different vendors; this is fully supported as long as they run the same binaries. Since many customers implement Oracle RAC for high availability, you must make sure that your hardware vendor will support the configuration. If you have a failure, will you get support for the hardware configuration?
Regards,
Levi Pereira

Do I need 3 machines for Sun Two-Node Cluster HA for Apache?

The vendor is trying to set up a Sun Cluster HA for Apache and said it requires a dedicated machine
for monitoring purpose(The Sun Cluster HA for Apache probe??). Is this monitoring machine mandatory and have to be dedicated for that purpose? And can I use 1 machine to monitor more than 1 cluster? Because the vendor requires 2 extra machine to monitor 2 set of 2-node cluster..and the document I read from Sun only mentions the 3rd machine is required for admin console.
Also, if I only have 2 node cluster, can I configure Apache as a scalable + failover service at the same time or I have to choose either one? Thanks!!

Incorrect. You can set up a Solaris Cluster on a single node if you wish although two nodes give you redundancy. The probe for Apache runs from the node on which Apache is executing.
If you have a two node cluster you could have multiple implementations of Apache: fail-over and scalable. You can mix them until you run out of resources: CPU, networking bandwidth, IP addresses, etc.
Tim
---

Multiple databases/instances on 4-node RAC Cluster including Physical Stand

OS: Windows 2003 Server R2 X64
DB: 10.2.0.4
Virtualization: NONE
Node Configuration: x64 architecture - 4-Socket Quad-Core (16 CPUs)
Node Memory: 128GB RAM
We are planning the following on the above-mentioned 4-node RAC cluster:
Node 1: DB1 with instanceDB11 (Active-Active: Load-balancing & Failover)
Node 2: DB1 with instanceDB12 (Active-Active: Load-balancing & Failover)
Node 3: DB1 with instanceDB13 (Active-Passive: Failover only) + DB2 with instanceDB21 (Active-Active: Load-balancing & Failover) + DB3 with instanceDB31 (Active-Active: Load-balancing & Failover) + DB4 with instance41 (Active-Active: Load-balancing & Failover)
Node 4: DB1 with instanceDB14 (Active-Passive: Failover only) + DB2 with instanceDB22 (Active-Active: Load-balancing & Failover) + DB3 with instanceDB32 (Active-Active: Load-balancing & Failover) + DB4 with instance42 (Active-Active: Load-balancing & Failover)
Note: DB1 will be the physical primary PROD OLTP database and will be open in READ-WRITE mode 24x7x365.
Note: DB2 will be a Physical Standby of DB1 and will be open in Read-Only mode for reporting purposes during the day-time, except for 3 hours at night when it will apply the logs.
Note: DB3 will be a Physical Standby of a remote database DB4 (not part of this cluster) and will be mounted in Managed Recovery mode for automatic failover/switchover purposes.
Note: DB4 will be the physical primary Data Warehouse DB.
Note: Going to 11g is NOT an option.
Note: Data Guard broker will be used across the board.
Please answer/advise of the following:
1. Is the above configuration supported and why so? If not, what are the alternatives?
2. Is the above configuration recommended and why so? If not, what are the recommended alternatives?

Hi,
As far as i understand, there's nothing wrong in configuration except you need to consider below points while implementing final design.
1. No of CPU on each servers
2. Memory on each servers
3. If you've RAC physical standby then apply(MRP0) will run on only one instance.
4. Since you are configuring physical standby for on 3rd and 4th nodes of DB1 4 node cluster where DB13 and DB14 instances are used only for failver, if you've a disaster at data center or power failure in entire data center, you are losing both primary and secondary with an assumption that your primary and physical standby reside in same data center so it may not be highly available architecture. If you are going to use extended RAC for this configuration then it makes sense where Node 1 and Node 2 will reside in Datacenter A and Node 3 ,4 will reside in Datacenter B.
Thanks,
Keyur

[b]Migrating the DB-Tier (DB and CM) to Two node non RAC cluster[/b]

Hi,
The current set-up of our E-business suite is a two node install:-
The DB Tier (Database and Concurrent Manager) on one node
The Apps Tier (Forms /Web Server) on another node.
For the HA solution (NON ORACLE RAC) we are planning to:-
Move the DB Tier (Database and Concurrent Manager) to a three node hardware sun cluster managed by veritas cluster manager (NOT ORACLE RAC). We need to know will the Database Tier (Database and Concurrent Manager) work on Hardware cluster node and will it support COLD FAILOVER from one Node to another. We know the database on its own would be fine with a cold failover because we have tested the database cold failover on the three node cluster for non E-business suite database. But here we have the added thing of the Concurrent manger sitting on the node along with the Database on the DB Tier.
The Apps Tier (Forms / Web server) will be put on a separate set of server using Load balances etc.
Has anybody implemented similar HA set-up and will this planned set-up work or are there any issues with this.
Any help / info would be appreciated.
Thanks

Hi,
Yes, you can do the cold failover the database and all the 11i services also.
1. In the concurrent manager service
== when you failover the concurrent manager do the following things before failover.
create the listener/tnsnames files which includes the new hostname and keep it with the veritas failover service. i mean, when you failover, these files should replace the existing files, before the existing files should backedup. and create a script to change the hostname,logfile_hostname,outfile_hostname in fnd_concurent_processes table.
add the nodes in the install->nodes navigation
do the failover manually and check the listener files are properly pinging using tnsping. and start the concurrent manager.
so tatally, you have to prepare two sql scripts
one for change the hostname from node b to node a
one for node a to node b.
and 2 listener + 2 tnsnames files which contains the seperate hostname accordingly.
use the adcmctl.sh top stop and start
finally, create a shell script to kill the sysmgr for the current managers when the manager takes long time to shutdown. before run the kill script, wait for 5 mts atleast.
I done the same scenario many times, with veritas failover service with 2 sun v880 servers.
regards,
Pandian

Theoretical question on having "stand by" instance on 4 node RAC cluster.

We have been having discussions on how we can improve the overall stability and performance of our RAC cluster. Currently we are running multiple databases on a 2 node cluster. The idea is that we will add two more nodes to the cluster and then spread the instances across the 4 nodes but only have two active and 1 as a standby in the event of node failure.
In other words:
Node1, Node2, and Node3 would run DB1
Node 4 Node3 and Node2 would run DB2
Additionally DB1_Instance3 would be shutdown on Node3 and only started if Node1 or 2 failed.
This would be the same for DB2 where DB2_Instance3 on Node2 would only be started if Node3 or Node4 failed.
Underneath them all would be the same ASM database shared for all 4 nodes.
Has anyone seen such a configuration? Is anyone aware of a white paper that discusses it.

the parameter active_instance_count is a 9i parameter and should only be set if you want to run RAC with a 2 node cluster - one instance active (all users connected here) and the second instance up but no one actively using it. IF instance one goes away, all work is move to instance 2. This is hot failover .
If you are adding a new instance to a rac database, you are recommended to use DBCA or EM to add the instance. If you do not want to have users connected to the instance all the time, you can either leave it running but do not enable any services on it until you need it or stop the instance with srvctl stop instance. The cluster will not try to restart the instance automatically.
The services option, when you create a service, you define which instances you wish it to run on. You can dynamically change where it runs. For the end user, they just connect to the service, the listener directs the connection to the least loaded instance providing the service at connect time. See the workload management with RAC 10g white paper on otn (otn.oracle.com/rac) or Chapter 6 of admin & deploy guide

SOA with Oracle 2 node RAC cluster

Hi All,
Just a simple doubt, I have successfully installed and configured SOA suite 11.1.1.3 & BAM in one wls Domain 10.3.3 in a linux box and could access all the application like BAM console, BPEL console etc .... also could see all my data-sources deployed in the Data Sources with a single node database.
1. Now I have to RE-configure this whole SOA suite with RAC (2 node database cluster) what changes or configuration needed to implement SOA suite with RAC database?
2. Do I need to create a "Multi Data Source" to configure RAC with SOA suite..?
Thanks
Sam

DB wrote:
This is regarding Oracle RAC..so if there is a specific category..please let me know..
I have installed OEL linux 5.6 as guest OS (using virtualbox) in two laptops.
I want to install 2 node oracle 10gR2 RAC with the OEL linux as OS and each laptop as one node.
Read docs and understood that there must be shared storage for oracle clusterware and oracle ASM for oracle RAC to work.
Please let me know the steps to create shared storage for oracle clusterware and oracle ASM (considering virtualbox OEL) and to configure public,private and virtual IPs.
I already have document to create 2node oracle RAC using virtualbox with two nodes on the same laptop.so please dont suggest that doc.
Thanks,
DBMay be my step by step RAC installation guide can help you somehow?
http://kamranagayev.wordpress.com/2011/04/05/step-by-step-installing-oracle-10g-rac-on-vmware/

JVM patch required for DST on 10.2.0.2 RAC cluster

I have looked all over the internet and Metalink for information regarding the JVM patching on a RAC cluster and haven't found anything useful, so I apologize if this question has already been asked multiple times. Also if there is a forum dedicated to DST issues, please point me in that direction.
I have a 10.2.0.2 RAC cluster so I know I have to do the JVM patching required because of the DST changes. The README for 5075470 says to follow post-implementation steps in the fix5075470README.txt file. Step 3 of those instructions say to bounce the database, and then not allow the use of java until step 4 is complete (which is to run the fix5075470b.sql script).
Here's my question: since this is a RAC database, does that mean I have to shutdown both instances, start them back up, run the script, and then let users log back in? IN OTHER WORDS, AN OUTAGE IS REQUIRED?
Is there a way around having to take an outage? Can I bounce each instance separately (in a rolling fashion) so there's no outage, and then run the script even though users are logged on if I think java isn't being used by the application? Is there a way to confirm whether or not it's being used? If I confirm the application isn't using java, is it ok to run the script while users are logged on?
Any insight would be greatly appreciated.
Thanks,
Susan

According to Note: 414309.1 USA 2007 DST Changes: Frequently Asked Questions an Problem for Oracle JVM Patches, question 4 Does the database need to be down before the OVJM patch is applied, the bounce is necessary. That says nothing about a rolling upgrade in RAC.
You might file an SR asking if a rolling upgrade is possible.

Automatic restart of services on a 1 node rac cluster with Clusterware

How do we enable a service to automaticly start-up when the db starts up?
Thanks,
Dave

srvctl enable service -d DBThanks for your reply M. Nauman. I researched that command and found we do have it enabled and that it only works if the database instance was previously taken down. Since the database does not go down on an Archiver Hung error as we are using FRA with an alt location, this never kicks in and brings up the service. What we are looking for something that will trigger off of when the archive logs error and switch from FRA(Flash Recovery Area) to our Alternate disk location. Or more presicely, when it goes back to a Valid status(on the FRA - after we've run an archive log backup to clear it).
I found out from our 2 senior dba's that our other 2 node rac environment does not suffer from this problem, only the newly created 1 node rac cluster environment. The problem is we don't know what that is(a parameter on the db or cluster or what) and how do we set it?
Anyone know?
Thanks,
Gib
Message was edited by:
Gib2008
Message was edited by:
Gib2008

Cluster Apex MS's across multiple nodes for load balancing?

Can Cluster Apex MS's across multiple nodes for load balancing? If Yes/No, any supportive doc's will be much appreciated.
Here is what I did ..........
I tried multiple MS's on different nodes for Apex Domain (clustered)
Deployed apex.war, i.war with targets pointing to cluster.
Apex listener techstack is shared filesystem $HOME/apex, where apex-config.xml and bdb are available
Anyone of the MS's whichever started up first is locking up bdb and 500 server shows up when internal workspace is being launched with host specific url, where as it works fine for the host which actually locked up bdb (first MS that was started)
2nd MS shows errors as below in log files ..
Caused By: com.sleepycat.je.EnvironmentLockedException: (JE 4.0.103)$HOME/apex/bdb The environment cannot be locked for single writer access. ENV_LOCKED: The je.lck file could not be locked. Environment is invalid and must be closed.
Its expected behavior until here .....
++++ I tried something to get it working ++++
Whilst both MS's are up and running, delete $HOME/apex/bdb but let apex-config,xml be as it is...now the internal workspace urls for both hosts work fine. I logged into host specific urls and navigated, created/edited users.,etc and it is all looking good but it did not create new bdb directory under $HOME/apex ...!!
Having said its working fine, the same issue repeats after the MS's bounce. So I did put a command line to delete bdb dir once the MS comes up to avoid the issue but really not sure if this is the right way at all?
Geek's ...awaiting your advice!!

Hey there,
I managed to fulfill my requirement..
If its a cluster on same machine or across machines, this should work
1. Login to machine, cd $DOMAIN_HOME
2. mkdir -p Apex_lsn_config/AdminServer Apex_lsn_config/<MS1> Apex_lsn_config/<MS2> # MS1 and MS2 are the Managed Server names as appropriate
#If you are planning for cluster spawning MS's across machines, make sure you create the dir's on step 2 for each machine respectively. (in my case $DOMAIN_HOME is not shared)
3. Copy apex-config.xml from the /tmp/apex or whatever location you have it currently to Apex_lsn_config/<MS1> Apex_lsn_config/<MS2>
4. cd $DOMAIN_HOME/bin; cp -p SetDomainEnv.sh SetDomainEnv.sh.orig #Backup the file
5. Append -Djava.io.tmpdir in SetDomainEnv.sh as below for JAVA_OPTIONS # Do it on both machine if you are not sharing DOMAIN_HOME and planning cluster across machines
-Djava.io.tmpdir=$DOMAIN_HOME/APEX_CONFIG/${SERVER_NAME}
Hint: Search for "iterativeDev" and append the same line with -Djava.jo.tmpdir
6. Modify "java.io.tmpdir" from the web.xml file of apex.war as below and re-deploy the war
<context-param>
<param-name>config.dir</param-name>
<param-value>${java.io.tmpdir}</param-value>
</context-param>
7. Bounce Weblogic Admin and Manged Servers. Make sure to tail the Managed Server log to see apex-config.xml is picked from the new location.
8. Brew a Coffee for yourself :)
- You find the instructions on creating a cluster from weblogic documentation, the steps mentioned above are only to overcome the bdb locking issue whilst creating a cluster.
Did it help?
Edited by: Oratime on Mar 25, 2013 2:44 AM

Large RAC cluster advice needed - 16 nodes for 120,000 TPS?

Similar Messages

Maybe you are looking for