Change Details

Labs Project Tested: ores Site/Location: EQIAD + CODFW Number of systems: 4 per DC, 8 in total Service: ORES Networking Requirements: internal Processor Requirements: >= 8 cores per box, 32 total. CPU nu Memory: 64GB per box, that 256 per DC Disks: >50GB, in each box performance not important, so SATA is an option NIC(s): 1x1Gb ethernet Partitioning Scheme: RAID+LVM Other Requirements: So this request is for both datacenters. Based on some simple spreadsheet calculations on T142046 we 'd need around 30 cores per DC and 210GB. I 've rounded up to 32 cores and 256GB respectively. Memory seems to be the issue here so that's why I am playing it a bit more safe. Some other simple calculations show that the best budget ratio probably comes from 4 boxes per DC, hence the request. That being said if we can get the same numbers in total with 2 or say 6 boxes per DC that's fine as well.

Labs Project Tested: ores Site/Location: EQIAD + CODFW Number of systems: 9 per DC, 18 in total Service: ORES Networking Requirements: internal Processor Requirements: 16 cores total Memory: 64 GB Disks: 1 TB capacity as RAID1 (software RAID is ok) NIC(s): 1gbe network card So this request is for both datacenters. Note that the requirements listed here are the same as T161700 (in turn the same as T141624). There is a reason for that. It is expected that ORES will migrate eventually to kubernetes and it would be great if then we would be able to fold the ores service cluster back into the kubernetes cluster as we move traffic to it without much hassle as to hardware specs and differences.

Labs Project Tested: ores Site/Location: EQIAD + CODFW Number of systems: 49 per DC, 18 in total Service: ORES Networking Requirements: internal Processor Requirements: >= 816 cores per box, 32 total. CPU nu Memory: 64GB per box, that 256 per DC GB Disks: >50GB, in each box performance not important, so SATA is an option1 TB capacity as RAID1 (software RAID is ok) NIC(s): 1x1Gb ethernet Partitioning Scheme: RAID+LVM Other Requirements: gbe network card So this request is for both datacenters. Based on some simple spreadsheet calculations on T142046 we 'd need around 30 cores per DC and 210GB. I 've rounded up to 32 cores and 256GB respectively. Memory seems to be the issue here so that's why I am playing it a bit more safe. Some other simple calculations show that the best budget ratio probably comes from 4 boxes per DC,Note that the requirements listed here are the same as T161700 (in turn the same as T141624). hence the requestThere is a reason for that. That being said if we can get the same numbers in total with 2 or say 6 boxes per DC that's fine as wellIt is expected that ORES will migrate eventually to kubernetes and it would be great if then we would be able to fold the ores service cluster back into the kubernetes cluster as we move traffic to it without much hassle as to hardware specs and differences.