Teaser Backup slides VNF Benchmark-as-a-Service (VBaaS): A Teaser (draft-rorosz-nfvrg-vbaas-00.txt) Rosa, Raphael V. †‡ Rothenberg, Christian E. ‡ Szabo, Robert † ‡ FEEC/UNICAMP and † Ericsson Research Hungary NFVRG, IETF 94 11/4/2015 1 / 11
Teaser Backup slides VNF Benchmarking Different NFVI PoP/host may perform +---------+ differently ,----. |Customers| ( VNF1 ) +----------+ +-+-------+ VNF development and NFVI upgrades are ‘----’ | VNF | | SAP1 +----+ +----+ ,----. |Developers| | O---|VNF1|---|VNF2|---O independent ( VNF2 )+--------+-+ | +----+ +----+ SAP2 ‘----’ | | <========VNF-FG=======> +-+-------+---+ Orchestration needs to know how much | NFVO / VNFM | resources (e.g., cpu, memory, storage) to | (Service | | Provider) | allocate for given target KPI values (e.g., +-+----+----+-+ / | \ throughput, latency). V V V +-------+ +-------+ +-------+ | VIM1 | | VIM2 | | VIM3 | +-+-----+ +---+---+ +-----+-+ | | | | | | | | | SAP1 +-----+-+ WAN +---+---+ WAN +-+-----+ SAP2 O-| PoP 1 |-----| PoP 2 |-----| PoP 3 |-O +-------+ +-------+ +-------+ PoP1 PoP2 PoP3 Container Enhanced Baremetal OS Hypervisor 2 / 11
Teaser Backup slides VNF Benchmarking Different NFVI PoP/host may perform +---------+ differently ,----. |Customers| ( VNF1 ) +----------+ +-+-------+ VNF development and NFVI upgrades are ‘----’ | VNF | | SAP1 +----+ +----+ ,----. |Developers| | O---|VNF1|---|VNF2|---O independent ( VNF2 )+--------+-+ | +----+ +----+ SAP2 ‘----’ | | <========VNF-FG=======> +-+-------+---+ Orchestration needs to know how much | NFVO / VNFM | resources (e.g., cpu, memory, storage) to | (Service | | Provider) | allocate for given target KPI values (e.g., +-+----+----+-+ / | \ throughput, latency). V V V +-------+ +-------+ +-------+ | VIM1 | | VIM2 | | VIM3 | +-+-----+ +---+---+ +-----+-+ Problem to be solved: | | | | | | Gain information autonomously about | | | SAP1 +-----+-+ WAN +---+---+ WAN +-+-----+ SAP2 VNFs’ benchmark metrics with given O-| PoP 1 |-----| PoP 2 |-----| PoP 3 |-O +-------+ +-------+ +-------+ reserved resources at a “host” (execution PoP1 PoP2 PoP3 Container Enhanced Baremetal environment). OS Hypervisor 2 / 11
Teaser Backup slides Highlights SAP1+----+ +----+ O--|VNF1|--|VNF2|--O {VNF1{ +----+ +----+SAP2 VNF-FG, <========VNF-FG=======> Metrics{ ,----. +---------+ Components {vCPU, mem}-> ( VNF2 ) ,----. |Customers| {BW, Delay}}} ‘----’ ( VNF1 ) +-+-------+ ,-------. ‘----’ | VBaaS service function /VBaaS-IB:\ +----------+ | \ VNF-BPs / | VNF | | VBaaS Information Base for VNF ‘---+---’ +--|Developers| | Benchmark Profiles | | +--------+-+ | _____+___ | | | structural | |<------+ +---+--------++ | VBaaS |<------------->| NFVO / VNFM | functional: manager, monitors and |___ _____| VBaaS | (Service | + API | Provider) | agents | +-+------+----+ ,----+-----. | | \ / NF-IB: \ V V V \VNF-Profiles/ +-----+ +-----+ +-----+ Work-flows ’----------’ |VIM1 | |VIM2 | |VIM3 | {VNF1: {10Mbps,200ms}{ ++----+ +--+--+ +-----+ for ETSI NFVO and VIMs {{2CPU, 8GB}@PoP1} | | | {{8CPU, 16GB}@PoP2} | | | for recursive orchestration {{4CPU, 4GB}@PoP3}}} | | | {20Mbps,300ms}...} +--+--+ WAN ++----+ WAN +-----+ {VNF2:{10Mbps,200ms}{ O-|PoP 1|-----|PoP 2|-----|PoP 3|-O 3 / 11 {{8CPU, 16GB}@PoP1} SAP1 +-----+ +-----+ +-----+ SAP2
Teaser Backup slides Summary VNF Benchmarking as a Service (VBaaS) aims at defining complementary functional components to ETSI NFV and other approaches; defining interfaces to the VBaaS service; defining possible VBaaS work-flows. Acknowledgements This work is partially supported by FP7 UNIFY, a research project partially funded by the European Community under the Seventh Framework Program (grant agreement no. 619609). The views expressed here are those of the authors only. The European Commission is not liable for any use that may be made of the information in this document This work is partially supported by Ericsson Brazil. 4 / 11
Teaser Backup slides Outline Teaser 1 Backup slides 2 5 / 11
Teaser Backup slides How? VNF Benchmarking as a Service (VBaaS) VNF 1 VNF 2 VNF 1 VNF-FG Target PoP1 Customer PoP2 PoP3 VNF-FG VBaaS NFVO VBaaS VBaaS-IB API VNF-BPs Ref Point: NFVO – VIM VNF-FG Monitors VIM Agent 1 VNF 1 Agent 2 PoP1 PoP2 NFVI PoP3 Manager PoP1 — PoP2 — PoP3 6 / 11
Teaser Backup slides How? VNF Benchmarking as a Service (VBaaS) VBaaS objectives VNF 1 VNF 2 VNF 1 “Black box” VNF VNF-FG Target PoP1 benchmarking, with Customer PoP2 PoP3 respect to the VNF-FG NVFO Benchmarking VBaaS NFVO VBaaS VBaaS-IB API VNF-BPs Service Ref Point: NFVO – VIM VNF-FG Monitors VIM Agent 1 VNF 1 Agent 2 PoP1 PoP2 NFVI PoP3 Manager PoP1 — PoP2 — PoP3 6 / 11
Teaser Backup slides How? VNF Benchmarking as a Service (VBaaS) VBaaS objectives VNF 1 VNF 2 VNF 1 “Black box” VNF VNF-FG Target PoP1 benchmarking, with Customer PoP2 PoP3 respect to the VNF-FG NVFO Benchmarking VBaaS NFVO VBaaS VBaaS-IB API VNF-BPs Service VBaaS uses NFVO Ref Point: NFVO – VIM VNF-FG (or VIMs) to deploy Monitors VIM benchmark Agent 1 VNF 1 Agent 2 measurements PoP1 PoP2 NFVI PoP3 e.g., by VNF-FG Manager PoP1 — PoP2 — PoP3 6 / 11
Teaser Backup slides VBaaS Process Walk-through VNF 1 VNF 2 VNF-FG {VNF1: {10Mbps,200ms}{ Customers {{2CPU, 8GB}@PoP1} NF-IB {{4CPU, 4GB}@PoP3}}} 2 VNF Profiles {20Mbps,300ms}...} {VNF2: ...} 1 3 12 11 4 NFVO / VNFM 5 VBaaS 7 {VNF1{ 8 VNF-FG, VBaaS-IB 6 Metrics{ VNF-BPs {VCPU, mem}-> {BW, delay}}} VIMs VNF2...} 10 9 NFVI PoPs Manager Agents, Monitors and SUT VNF 7 / 11
Teaser Backup slides Usage: Benchmarking, Dimensioning and Verification Benchmarking To measure VNF’s throughput, latency, jitter parameters for given cpu, memory, storage reservation at given VIM. 8 / 11
Teaser Backup slides Usage: Benchmarking, Dimensioning and Verification Benchmarking To measure VNF’s throughput, latency, jitter parameters for given cpu, memory, storage reservation at given VIM. Dimensioning To determine cpu, memory, storage reservation parameters for given VNF at given VIM for target throughput, latency, jitter parameters. 8 / 11
Teaser Backup slides Usage: Benchmarking, Dimensioning and Verification Benchmarking To measure VNF’s throughput, latency, jitter parameters for given cpu, memory, storage reservation at given VIM. Dimensioning To determine cpu, memory, storage reservation parameters for given VNF at given VIM for target throughput, latency, jitter parameters. Verification To assess if given throughput, latency, jitter parameters of a VNF is met with given cpu, memory, storage reservation at given VIM. 8 / 11
Teaser Backup slides Usage: Benchmarking, Dimensioning and Verification Benchmarking To measure VNF’s throughput, latency, jitter parameters for given cpu, memory, storage reservation at given VIM. Dimensioning To determine cpu, memory, storage reservation parameters for given VNF at given VIM for target throughput, latency, jitter parameters. Verification To assess if given throughput, latency, jitter parameters of a VNF is met with given cpu, memory, storage reservation at given VIM. Observation Dimensioning and verification boil down to benchmarking operation(s). 8 / 11
Teaser Backup slides Recursive Orchestration with VBaaS UNIFYing carrier network and cloud resources +--------------+ | | Recurrent joint software and networking |Orchestration | Domain 4 | [VBaaS] | +--+-----------+ control API a **********************|****************** * +--------------+ | * |NFVO: | | * |Service | | * |Lifecycle | | * |Management | | * +-------+------+ / * | / <-- Joint API * +-+---------+--+ * | | * |Orchestration | a draft-unify-nfvrg-recursive-programming-02 ******************** * | [VBaaS] | +--------------+ * * +--+---+-------+ Domain 3 |NFVO: | * ********|***|************************* |Service | * / | |Lifecycle | /---------/ | |Management | / * | +---------+----+ | * | | | * | <-- Joint API +--+-------+---+* | | |* | |Orchestration |* | | [VBaaS] |* | | |* | +------+-------+* | | * *********|********** <-- Joint API +------+-------+* * +------+-------+ * |VIM: |* * |VIM: | * |Virtualized 2 |* * |Virtualized 2 | * |Pys mapping |* * |Pys mapping | * +--------------+* * +--------------+ * Domain 1 * * Domain 2 * ************************* * * 9 / 11
Recommend
More recommend
