Ramki, I agree with these points. We can throw in more constraints such a VNF license terms/conditions, connectivity diversity constraints etc. We can discuss these further. The key here is to express the constraints as policies, which automatically or semi-automatically create meaningful optimization problems to be solved effectively using an algorithm library, rendering significant reusability and scalability across many use cases.
With respect to use cases, as we discussed in the 5G RAN meeting today, we could try various approaches to achieve these goals Thanks much. Sarat From: Ramki Krishnan [mailto:[email protected]] Sent: Tuesday, August 15, 2017 8:45 PM To: PUTHENPURA, SARAT (SARAT) <[email protected]>; [email protected]; PATEL, ANKITKUMAR N (ANKITKUMAR N) <[email protected]>; SHANKARANARAYANAN, N K (N K) <[email protected]> Subject: Optimization Framework - use cases for ONAP R2 Hi Sarat, All, Following up on our meeting today, was thinking a bit deeper on distributed placement of analytics engines which could be ONAP components such as DCAE and external components such as vendor-specific EMS or analytics engines. The placement domain can encompass operator private cloud with VIMs from different vendors (OpenStack, Wind river, VMware Integrated OpenStack etc.) or public cloud (Microsoft Azure etc.). Declarative placement policy examples - Minimize latency for closed loop fault detection/remedy for VOLTE user plane traffic o Desire to place real-time analytics engines near the user plane VNF o Desire to choose the real-time analytics engine with minimal response latency - Minimize cost for historical analytics o Desire to place historical analytics engines in public cloud Exemplary constraints of interest - Network interconnect bandwidth, Xth percentile latency (X could be 99.9) - Closed loop response time - Cloud compute, network, storage pricing - Yth percentile latency of individual components (real-time analytics engine etc.) used for closed loop automation (Y could be 99) o This would include details about vendor-specific components such as EMS - Application metric - end-to-end Zth percentile latency (Z could be 99) This would be of immediate value for R1/R2 use cases such as - VOLTE use case focused on 3GPP CUPS -- https://wiki.onap.org/pages/viewpage.action?pageId=6593603<https://urldefense.proofpoint.com/v2/url?u=https-3A__wiki.onap.org_pages_viewpage.action-3FpageId-3D6593603&d=DwQFAg&c=LFYZ-o9_HUMeMTSQicvjIg&r=fEZLWz1dGC0hCK3H3CPRqqfvm4_nMAI8OVaDfcQityY&m=tcQEZZPXyUPQhBXsP42RRnRtI6xhOe3lVrSpBsSWq1k&s=Fz9XshxIImp8o7NuibTOhb-l9Z0NGKgP7f5Qbj1_tRM&e=> - 5G RAN -- https://wiki.onap.org/display/DW/Use+case%3A+5G-+RAN+deployment%2C+E2E+Slicing%2C+SON<https://urldefense.proofpoint.com/v2/url?u=https-3A__wiki.onap.org_display_DW_Use-2Bcase-253A-2B5G-2D-2BRAN-2Bdeployment-252C-2BE2E-2BSlicing-252C-2BSON&d=DwQFAg&c=LFYZ-o9_HUMeMTSQicvjIg&r=fEZLWz1dGC0hCK3H3CPRqqfvm4_nMAI8OVaDfcQityY&m=tcQEZZPXyUPQhBXsP42RRnRtI6xhOe3lVrSpBsSWq1k&s=BUpJQbyA10zpnSkm8jG0hAUj0WvbMBVerXlN0GLTHTA&e=> Comments and thoughts welcome. Thanks, Ramki
_______________________________________________ onap-discuss mailing list [email protected] https://lists.onap.org/mailman/listinfo/onap-discuss
