Below is the topology we have prepared as per individual product licensing we have,
License: DS 4 Core, QS 4 Core, IA 2 Core, IGC 2 Core
VM1 (2 Core):
Metadata Repository Tier
Services Tier (DS, QS, IA, IGC)
Engine Tier (IA)
VM2 (4 Core):
Engine Tier (DS)
VM3 (4 Core):
Engine Tier (QS)
Installation Part 1 on VM1 >> New Installation
Tiers selected: MetaData Repository + Services
Products selected: DS + QS + IA + IGC
Installation Part2 on VM1 >> Add Tiers
Tiers selected: Engine
Products to be selected: IA (Issue: But here we were not prompted for product selection... Installer program assumes adding Engine Tier for all products available on Services Tier, where we would like to install only IA Engine on this machine.)
Then we tried other way,
Installation Part2 on VM1 >> Add Product
Products to be selected: IA (Issue: But here, IA was NOT shown under products list, installer program assumes as its already installed as part of Services, where we would like to Engine Tier for IA on this machine.)
I don't think we can't install Engine for only IA where is already services been installed for all products (DS, QS, IA, IGC). Can you please explore and guide me to install as desired or please let me know if there is any issue with topology we have selected?
Thanks in advance.
Installation Issue: Installing IA/Engine on Services Tier
Moderators: chulett, rschirm, roy
Not sure I understand how you want to spread the jobs. Are you talking about an install of datastage and IA ONCE and then share mounting the engine binaries to the other clustered hosts VM2 + 3.
You could then handleing the job distribution via APT files.
Or are you talking about three seperate installs with a shared metadata and webspere layer (avoid this IMHO)?
I recommend that you avoid that shared Websphere layer because you will encounter patching issues where your engines will be out of sync. Not to mentione the headaches involved configuring 3 instances of websphere on one box.
I think 1 install of datastage on a clustered environment and spreading your work around via APT driven redirection would suite you best.
But... not sure how many jobs you plan on running concurnetly. I would still try to maximize the use of each box and not limit yourself. Have one box favor IA jobs, but spread DS and QS jobs across both of the others.
You could then handleing the job distribution via APT files.
Or are you talking about three seperate installs with a shared metadata and webspere layer (avoid this IMHO)?
I recommend that you avoid that shared Websphere layer because you will encounter patching issues where your engines will be out of sync. Not to mentione the headaches involved configuring 3 instances of websphere on one box.
I think 1 install of datastage on a clustered environment and spreading your work around via APT driven redirection would suite you best.
But... not sure how many jobs you plan on running concurnetly. I would still try to maximize the use of each box and not limit yourself. Have one box favor IA jobs, but spread DS and QS jobs across both of the others.
-
ray.wurlod
- Participant
- Posts: 54607
- Joined: Wed Oct 23, 2002 10:52 pm
- Location: Sydney, Australia
- Contact:
Hi! Paul,
Thanks for your inputs.
We will be using same version (11.3) across installation n machines, first we are installing Services with all products (IA, IGC, DS, QS) on one machine, then Engine for two products (IA, IGC) on the same machine. And extending Engine tier in other two machines for different products (DS), (QS).
Note: All 3 machines are virtual machines, no clustering, no file system sharing will be used. Treat them as separate machines.
Does it give any harm? I don't hope.
Please look at my topology and advise. Thanks in advance.
VM1 (2 Core):
Metadata Repository Tier
Services Tier (DS, QS, IA, IGC)
Engine Tier (IA, IGC)
VM2 (4 Core):
Engine Tier (DS)
VM3 (4 Core):
Engine Tier (QS)
Thanks for your inputs.
We will be using same version (11.3) across installation n machines, first we are installing Services with all products (IA, IGC, DS, QS) on one machine, then Engine for two products (IA, IGC) on the same machine. And extending Engine tier in other two machines for different products (DS), (QS).
Note: All 3 machines are virtual machines, no clustering, no file system sharing will be used. Treat them as separate machines.
Does it give any harm? I don't hope.
Please look at my topology and advise. Thanks in advance.
VM1 (2 Core):
Metadata Repository Tier
Services Tier (DS, QS, IA, IGC)
Engine Tier (IA, IGC)
VM2 (4 Core):
Engine Tier (DS)
VM3 (4 Core):
Engine Tier (QS)
Are you splitting up DS and QS because you have $$$ concerns for the licenses?
How many concurrent jobs total do you anticipate from all 3 environments?
2 core box for VM1... a little low I would say.
2+4+4 = really small setup
You're not going to get much IA work done on a 2 core box that also services your xmeta and WAS.
IMHO:
VM1 = 4 core (xmeta/iadb/dsodb)
VM2 = 10 core (WAS + Engine for DS/QS/IA)
VM1 PVU costs are not factored in to the $$$ for licenses since Information Server comes with a DB2 license and it is not configured to run a DS job (engine tier).
VM2 will run all of your jobs for IA/DS/QS.
You still have 10 cores of PVU license.
You have less disk costs, can maximize memory usage and scratch disk.
You have less moving parts to go wrong.
Patching will be a one stop shop.
Shared metadata repository for IGC to report on any operational metadata created that you want to expose.
Expansion can be done by adding cores to one host (VM2) or by adding a Compute Node VM3 with a clustered file system (shared drives with VM2). You need to plan that ahead of time for the file system you chose to use on your engine and project mounts.
project mounts != engine binaries mount.
How many concurrent jobs total do you anticipate from all 3 environments?
2 core box for VM1... a little low I would say.
2+4+4 = really small setup
You're not going to get much IA work done on a 2 core box that also services your xmeta and WAS.
IMHO:
VM1 = 4 core (xmeta/iadb/dsodb)
VM2 = 10 core (WAS + Engine for DS/QS/IA)
VM1 PVU costs are not factored in to the $$$ for licenses since Information Server comes with a DB2 license and it is not configured to run a DS job (engine tier).
VM2 will run all of your jobs for IA/DS/QS.
You still have 10 cores of PVU license.
You have less disk costs, can maximize memory usage and scratch disk.
You have less moving parts to go wrong.
Patching will be a one stop shop.
Shared metadata repository for IGC to report on any operational metadata created that you want to expose.
Expansion can be done by adding cores to one host (VM2) or by adding a Compute Node VM3 with a clustered file system (shared drives with VM2). You need to plan that ahead of time for the file system you chose to use on your engine and project mounts.
project mounts != engine binaries mount.
Hi! Paul,
Really Thanks for your insightful response.
There would be roughly 500 Jobs overall and at peak time concurrent jobs would not cross not 20 on all machines.
With the licensing in hand as mentioned in beginning of this post (DS 4Core, QS 4Core, IA 2Core, IGC 2Core).
I could see below license issues in deploying topology you recommended, however, agreed yours is one of best topology we could consider.
1- I hope installing Engine tier with DS, QS, IA on one 10 Core machine may violate license. With DS 4 Core license we can't go for installing Engine for DS on 10 Core machine, right?
2- IGC also would consume Engine tier as it's combination of Metadata Workbenh & Business Glossary. Metadata Workbench requires engine tier while Business Glossary doesn't.
3- By installing Services on 10 Core, IGC 2 Core license will be violated, right? Since maximum cores used on Services or Engine would be counted for IGC.
In addition, below points are in our consideration,
1- Keep out Metadata repository databases completely out of VMs and Non-DB2 will be used on separate machine, so that only Services & Engine for IA, IGC will stay on VM3.
2- For Services we are going to use Liberty Profile, which is light weight and faster as IBM claims.
Please let me know if anything wrong in my understanding about licensing. Thanks in advance.
Really Thanks for your insightful response.
There would be roughly 500 Jobs overall and at peak time concurrent jobs would not cross not 20 on all machines.
With the licensing in hand as mentioned in beginning of this post (DS 4Core, QS 4Core, IA 2Core, IGC 2Core).
I could see below license issues in deploying topology you recommended, however, agreed yours is one of best topology we could consider.
1- I hope installing Engine tier with DS, QS, IA on one 10 Core machine may violate license. With DS 4 Core license we can't go for installing Engine for DS on 10 Core machine, right?
2- IGC also would consume Engine tier as it's combination of Metadata Workbenh & Business Glossary. Metadata Workbench requires engine tier while Business Glossary doesn't.
3- By installing Services on 10 Core, IGC 2 Core license will be violated, right? Since maximum cores used on Services or Engine would be counted for IGC.
In addition, below points are in our consideration,
1- Keep out Metadata repository databases completely out of VMs and Non-DB2 will be used on separate machine, so that only Services & Engine for IA, IGC will stay on VM3.
2- For Services we are going to use Liberty Profile, which is light weight and faster as IBM claims.
Please let me know if anything wrong in my understanding about licensing. Thanks in advance.