The one-size nature of the Pharos specs has always been one of the key weaknesses of the OPNFV program IMO. It does not reflect the likely diversity of hardware infra that will be used in different deployment environments, e.g. datacenter, edge, customer premises (enterprise, consumer), including compute, storage, and networking components. Even so, it’s been difficult to nail down how a Pharos POD should be configured (aside from how it *could* be as described in the spec, in loose terms), and that unclarity is I think a key aspect of why it’s so hard to deploy OPNFV installer-based distros in labs. For example, there is nothing AFAICT about how switches should be configured (in detail, as needed), though this is a key aspect of getting the networking to work.
I think that OPNFV should focus on documenting Pharos lab configs that have been proven compatible (with PDF’s provided) with specific installers, so that it’s possible to replicate that. In the process we would likely see the true diversity of POD configs that are needed, being clarified, as lab owners apply their PODs to different deployment use cases. This would be a more useful approach to this IMO. Re 32GB or 128GB, personally for a datacenter-focused lab server I would recommend no less than 256GB or even 512GB (which we use in production). For an end-focused deployment the RAM may be less, and for a CPE deployment maybe 128GB or less, depending upon the type of CPE. A key consideration in the memory footprint will be what is the nature of the control plane, e.g. OpenStack (heavyweight) or cloud-native (lightweight). Thanks, Bryan Sullivan | AT&T From: opnfv-tech-discuss-boun...@lists.opnfv.org [mailto:opnfv-tech-discuss-boun...@lists.opnfv.org] On Behalf Of Jack Morgan Sent: Monday, September 25, 2017 10:39 PM To: opnfv-tech-discuss@lists.opnfv.org Subject: Re: [opnfv-tech-discuss] Discussion Topics for Weekly Technical Discussion this Thursday 09/28 Bin, I think it would be a good conversation to have around hardware requirements in OPNFV and how they evolving. This is particularly interesting when thinking about OPNFV and ONAP integration. For example, the pharos specification only has a minimum of 32Gb of memory on nodes but I've gotten a couple of requests recently for an OPNFV POD with 128Gb of memory as a minimum. I wonder how sustainable this is long term for lab owners. I also wonder what other hardware requirements need to be evaluated within OPNFV (besides memory) I think this would be an interesting discussion. Anyone else? On 09/25/2017 09:56 PM, HU, BIN wrote: Hello community, Please let me know if you have any topics or issues that need to be discussed at our weekly technical discussion this Thursday 09/28. If no issues or topics, we may cancel the discussion this week. -- Jack Morgan OPNFV Pharos Intel Lab
_______________________________________________ opnfv-tech-discuss mailing list opnfv-tech-discuss@lists.opnfv.org https://lists.opnfv.org/mailman/listinfo/opnfv-tech-discuss
