Hello ,
We are planning and approaching on Migration project (IIS 8.1 to IIS 9.1) for our client.
Looking for expertise help in setting up the Grid Environment for IIS. Source OS is Suse Linux and Target OS Redhat Linux. We are in analysis phase of certain topics under which Grid Setup is the area we are facing big challenge. It Would be of great help if anyone can help us out in guiding this task?
Details of Existing setup
Prod - Two Head Nodes ( HA Enabled) - 6 Compute nodes for HN1 and 3 Compute Nodes for HN2
Dev,Test and stage are other environments
Thanks in advance
Implementing Grid Setup for IIS
Moderators: chulett, rschirm, roy
Implementing Grid Setup for IIS
Bharathiraja S
Is your system SAN-based or NAS-based configuration grid environment?
Do compute nodes have multiple NIC cards and using 10 gbs switches?
A lot of shops set up their grids, but didn't do nodes translations correctly, so the jobs can still run, but with very poor performance. They didn't realize that routing of job's activity occurred on the public network rather than the desired private network (with 10 gbs switches). So that leave the poor developers still trying to figure out why sometimes jobs run OK and sometimes jobs run so slow. Your prod env. is HA enabled; do you use NetApp storage server?
What Resource Manager software do you use? PBS Pro or IBM LoadLeveler or Sun? Since I have to collect users' usage info. to do the cost allocation and the charge-back, I have found that PBS Pro provides the best feathers for doing those tasks.
Do compute nodes have multiple NIC cards and using 10 gbs switches?
A lot of shops set up their grids, but didn't do nodes translations correctly, so the jobs can still run, but with very poor performance. They didn't realize that routing of job's activity occurred on the public network rather than the desired private network (with 10 gbs switches). So that leave the poor developers still trying to figure out why sometimes jobs run OK and sometimes jobs run so slow. Your prod env. is HA enabled; do you use NetApp storage server?
What Resource Manager software do you use? PBS Pro or IBM LoadLeveler or Sun? Since I have to collect users' usage info. to do the cost allocation and the charge-back, I have found that PBS Pro provides the best feathers for doing those tasks.
-
ray.wurlod
- Participant
- Posts: 54607
- Joined: Wed Oct 23, 2002 10:52 pm
- Location: Sydney, Australia
- Contact:
You also have to think about what you can buy and support for your ongoing compute node servers. These days you can't even buy anything under 12 cores. How big do you want your grid to be? I would not recommend chopping up servers into 2 core systems (some folks think visualization will cure everything). The grid manager will handle if the server can't take any more work, so I recommend not chopping up those servers with an extra layer of delay between you and the hardware.
Platform LSF is good, some big shops are using it.
The math on how you calculate if a server is busy and how you initially dispatch jobs to a serve will be what you really need to work on. That is true for any grid flavor you set up.
If you are going with LSF, look into the ELIM aspect of that product. Very handy way to tag servers with resource availability. Mounts, licensed products, special purpose designated boxes, etc...
You might want to also pitch the fact that GRID is more than just datastage. If the box ain't busy... it can service the company in other number crunching aspects. Use it if you can. You'll be a hero if you can prove that the boxes are underutilized and able to be used by others in the company, but still favor DataStage jobs when the need arises.
Platform LSF is good, some big shops are using it.
The math on how you calculate if a server is busy and how you initially dispatch jobs to a serve will be what you really need to work on. That is true for any grid flavor you set up.
If you are going with LSF, look into the ELIM aspect of that product. Very handy way to tag servers with resource availability. Mounts, licensed products, special purpose designated boxes, etc...
You might want to also pitch the fact that GRID is more than just datastage. If the box ain't busy... it can service the company in other number crunching aspects. Use it if you can. You'll be a hero if you can prove that the boxes are underutilized and able to be used by others in the company, but still favor DataStage jobs when the need arises.