Resource reqs differences between PCF 1.8 & 1.9, according to the PCF sizer

[derekj21]

Greetings:

I'm working with some other people at Pivotal to commercialize a hosted PCF service and we are trying to spec out the physical infrastructure we need to run the first X number of customers. I've been using the PCF Sizer for some time to model some hypothetical customers, and I have noticed that since you added PCF 1.9, the estimated resource utilization, particularly RAM needs, is down considerably (20-25%), and the object storage requirements have doubled.

I just want to confirm that what I'm seeing is accurate, particularly in how much more efficiently (I assume) PCF is using RAM in the 1.9 version compared to 1.7 and 1.8. We are building a vSphere-based service and that is what I am focusing on, but I've checked the other sizer IaaS options and the differences seem to hold true for the AWS, AWS Dedicated, and GCP versions. Curiously, Azure specifications are almost completely unchanged between PCF 1.8 and 1.9.

I've included some of the Sizer results that I am seeing. Are these 1.9 numbers legit?

Two more questions (thanks!)

• do you know why it appears that Azure requires substantially less resources to run a given PCF set up than any of the other clouds shown on the Sizer? that is curious.
• is the "NHC vSphere" choice for EMC's Native Hybrid Cloud on-prem device?

Thanks,
Derek

## Input 1:
Select IaaS: vSphere
IaaS Tuning: 4:1
PCF Version: 1.8
Installation Size: Large

• AI: 200
• AI Avg RAM: 2 GB
• AI Avg Storage: 3 GB
• Cell Instance Type: 2xLarge
  - Cell Mem Size: 32 GB
  - Cell Storage Size: 64 GB
• Availability Zones: 3

Services: OpsMetrics, MySQL (config left at defaults)

Results:

• Total AZs: 3
• **Foundation Memory: 908 GB
• Foundation Storage: 3820
• Foundation vCPU: 307**
• Physical Cores: 77
• Total IP Addresses: 140
• Total VMs: 108
  - Total Cell RAM: 704
  - Usable RAM per Cell: 29 GB
  - Usable Disk per Cell: 28 GB
  - Available Cell RAM: 238 GB
  - Available Cell Disk: 16 B
  - Number of Cells: 22
• Cells per AZ: 8

## Input 2:
Select IaaS: vSphere
IaaS Tuning: 4:1
PCF Version: 1.9
Installation Size: Large

• AI: 200
• AI Avg RAM: 2 GB
• AI Avg Storage: 3 GB
• Cell Instance Type: xlarge.disk
• Cell Mem Size: 16 GB
• Cell Storage Size: 128 GB
• Availability Zones: 3

Services: OpsMetrics, MySQL (config left at defaults)

Results:

• Total AZs: 3
  - Foundation Memory: 709 GB
  - Foundation Storage: 6157 GB
  - Foundation vCPU: 270
• Physical Cores: 77
• Total IP Addresses: 140
• Total VMs: 108
  - Total Cell RAM: 496
  - Usable RAM per Cell: 13 GB
  - Usable Disk per Cell: 108 GB
  - Available Cell RAM: 3 GB
  - Available Cell Disk: 2748 GB
  - Number of Cells: 31
• Cells per AZ: 11

[bbyers-pivotal]

Hi @derekj21.

The main reason for resource changes between 1.8 and 1.9 was choosing instance sizes more appropriate for the job they perform. Storage is up because of the Diego cells we choose by default. They now each have 128GB ephemeral disk attached instead of 32GB or 64GB by default. That's where you're seeing the bump in disk. Granted, each Diego cell would never actually use up all 128GB of disk, but we still include the full amount in the total.

For 1.9, yes, those numbers area correct. Disk can be dialed back a little, 128GB per Diego cell happens to be the size we chose, but each Diego cell shouldn't be using anywhere close to that amount.

As for the Azure question, can you give me an example where Azure is using less resources? When I looked, it was within a few VMs of the other IaaS. I would be interested to see what you are noticing.

Yes - NHC vSphere is for Dell/EMCs Native Hybrid Cloud solution. It's the same as vSphere, but gives an option for another Service to be selected which has the NHC required services included

I hope that answers your questions. Sorry it took a bit to reply.

Thanks!
Brian