Re: draft-morton-bmwg-virtual-net-00
ramki Krishnan <[email protected]>
| Newsgroups | gmane.ietf.bmwg |
|---|---|
| Message-ID | <C7634EB63EFD984A978DFB46EA5174F2C00372AF33@HQ1-EXCH01.corp.brocade.com> |
Hi Al, There are some important and interesting differences between energy efficiency in switches/routers vs VNFs (which runs on servers). Energy efficiency in Switches/Routers - Powering off switch/router components for energy efficiency could have a detrimental effect on the applications unless the servers (sourcing the applications) are also turned off - There are no open standards for this Energy efficiency in VNFs - The NFV requirements (high level) for energy efficiency are well defined in the NFV Virtualization Requirements document (http://www.etsi.org/deliver/etsi_gs/NFV/001_099/004/01.01.01_60/gs_NFV004v010101p.pdf) - A number of standard interfaces have been developed to support the need to manage workloads across individual/multiple servers to achieve the desired performance and power consumption. Notable established standards are: * Intelligent Platform Management Interface (IPMI 2.0) * Data Center Management Interface (DCMI 1.5) * DMTF Systems Management Architecture for Server Hardware (SMASH) As you pointed out, we need to take closer look, but I don't see a big problem in coming to consensus on topic in a reasonable amount of time. Thanks, Ramki -----Original Message----- From: MORTON, ALFRED C (AL) [mailto:[email protected]] Sent: Wednesday, April 09, 2014 8:07 AM To: ramki Krishnan; Muhammad Durrani; Bhavani Parise; [email protected] Subject: RE: [bmwg] draft-morton-bmwg-virtual-net-00 Hi Ramki, If you search our past work proposals, draft-manral-... proposed energy efficiency metrics and benchmarking. There is some controversy and complexity on making power measurements, we need to collect these issues and see what can be done. Al ________________________________________ From: bmwg [[email protected]] On Behalf Of ramki Krishnan [[email protected]] Sent: Wednesday, April 09, 2014 10:12 AM To: Muhammad Durrani; Bhavani Parise; [email protected]<mailto:[email protected]> Subject: Re: [bmwg] draft-morton-bmwg-virtual-net-00 Additionally, it would be worthwhile to discuss the common energy efficiency aspects across all VNFs. These would include discussing the tradeoffs between CPU energy efficiency/latency/throughput for various packet processing schemes such as - Fixed polling - Adaptive polling - Changing processor P-states (this could be platform specific) Thanks, Ramki -----Original Message----- From: bmwg [mailto:[email protected]] On Behalf Of Muhammad Durrani Sent: Wednesday, March 19, 2014 12:27 AM To: Bhavani Parise; [email protected]<mailto:[email protected]> Subject: Re: [bmwg] draft-morton-bmwg-virtual-net-00 Hi Bhavani, - do we need to consider or compare performance with the real/physical network vs. the virtual network - in 3.2 do we need Virtual Switch [MD] We do need to provide methodologies to benchmark for three architectures vSwitch/SR-IOV and PCI bypass. Regarding comparing performance only PCI bypass can be compared with physical world. Again we are not benchmarking networks here but the virtual appliance and wherever needed comparison will be laid out for both Physical and virtual appliance. for 3.2 how about load balancers, firewalls - do we need to consider, what kind of DPI - firewalls - stateful vs stateless [MD] Virtual appliance is broader term - it means virtual Load balancer and firewalls (both state full and stateless) We won't be suggesting "kind of DPI" - Standard method will be provided to qualify appliance for any type of DPI Regards, Muhammad -----Original Message----- From: bmwg [mailto:[email protected]] On Behalf Of Bhavani Parise Sent: Wednesday, March 19, 2014 12:03 AM To: [email protected]<mailto:[email protected]<mailto:[email protected]%3cmailto:[email protected]>> Subject: [bmwg] draft-morton-bmwg-virtual-net-00 Al, including the WG list also for my comments/Qs for this draft. These are a few things which we can look into including as considerations for this draft. - also for 3.2 - for VNF itself, do we need location of the VNF - guest VM,separate process, appliance etc - - what kind of considerations we need to note when testing a Virtual Firewall Function - how many Tenants can a VNF support? - for 3.3 - how about VM mobility or VNF mobility itself thanks, Bhavani ________________________________________ _______________________________________________ bmwg mailing list [email protected]<mailto:[email protected]<mailto:[email protected]%3cmailto:[email protected]>> https://www.ietf.org/mailman/listinfo/bmwg _______________________________________________ bmwg mailing list [email protected]<mailto:[email protected]<mailto:[email protected]%3cmailto:[email protected]>> https://www.ietf.org/mailman/listinfo/bmwg _______________________________________________ bmwg mailing list [email protected] https://www.ietf.org/mailman/listinfo/bmwg