History of RNP measurement activities • WORKING GROUPS (GT’s): • 2002-2007: GT-QoS / GT-QoS2 / GT-Medições / GTMedições2: – Active / Passive measurements – First international interactions: – perfSONAR developers meeting (2006) and a meeting in Brazil (2007) – First developments: – piPEs-BR, nSLA, CLMP, ICE, CACTISonar • EXPERIMENTAL SERVICE: • 2008-2009: MonIPÊ Experimental Service: – Developments: NMWG in PHP, CACTISonar, perfSONAR RNP Reports and PHP-SQL-MA – Deployments: Lookup Service (LS), Network Diagnotic Tester (NDT) and NFSen • PRODUCTION SERVICE: • 2010-2012: MonIPÊ Service: – RNP deployment in all PoPs 2 MonIPÊ Service: reaching out to clients Keeping the service focus on the clients – Lower hardware acquisition costs for service delivery – Simplify infrastructure deployment – Improve the user experience of the service 3 MonIPÊ Hardware: Lowering costs Service infrastructure: – Measurement Points (MPs) in Virtual Machines – Low cost measurement kit » Latency MP: Raspberry Pi » Low cost GPS antenna: Adafruit » Achievable bandwidth MP: CuBox – Development: » MP for measurements up to 10Gbps – New web-based graphical user interface » Execute and display measurements results » Configuration and control of MPs (all hosts: high-end servers, VMs and low-cost boxes) 4 MonIPÊ perfSONAR Measurement Portal Seamless portal navigation On-demand tests Retrieval of on-demand tests Retrieval of archived tests 5 MonIPÊ Service MonIPÊ compatible with: - perfSONAR-PS (Internet2 and ESNet) - perfSONAR MDM (GÉANT) Measurement Scenarios: - International, Backbone and Institutions Scheduling Manners and Mechanisms: - On-demand, Periodic and Permanent - Point-to-Point, Point-to-Multipoint and Multi-to-Multipoint Metrics: - Packet loss, One-way delay, Round trip time, TCP achievable bandwidth and UDP bandwidth Technical Architecture: - International (10G MPs), Backbone (Virtual/10G MP) and Clients (Low cost kit) 6 RIPE Atlas RIPE Atlas (proposed by RIPE NCC) target: “build an Internet measurement network employing a global network of probes that measure Internet connectivity and reachability” Goals: • Provide users active measurements to baseline • Enable on-demand individual measurements • Produce Internet traffic maps and usage of other data by the technical community • Act as a trusted source of data regarding real-life, active measurements 7 RIPE Atlas Main concepts: • User, Host and Sponsor Technical infrastructure: • Anchor and Probe Metrics: • Network configuration information • Probe uptime • Round trip time measurements (1st and 2nd hop) • Ping, traceroute and SSL queries (predetermined destinations) • DNS queries (root DNS servers) 8 MonIPÊ versus RIPE Atlas Focus MonIPÊ Service RIPE Atlas Network Performance Active Measurements Measurements Main Components Information services and Anchors and Probes Measurement Points Service Use Cases International, Backbone and User, Host and Sponsor Institutions Service Sponsor RNP Metrics Packet loss Configuration information One-way delay Uptime Round trip time Round trip time TCP achievable bandwidth Traceroute UDP bandwidth DNS query SSL query On-demand, Periodic and Predefined measurements Permanent User defined measurements Scheduling RIPE NCC Point-to-Point, Point-toMultipoint and Multi-toMultipoint Programmatic Interface perfSONAR NMWG RIPE Atlas REST API 9 Service Infrastructure RNP PoP Client Site • 2 virtual servers: RNP Measurement Portal + perfSONAR Measurement Archive (MA) PoP • 2 virtual servers: MA & MP + Measurement Portal • Existing GPS antennas (serial port to synchronise the clock) • 2 dedicated NICs: delay + throughput Client Site • Cubox: 1 Gbps NIC (500 Mbps throughput) • Raspberry PI: serial port • Adafruit GPS Antenna: pulse per second (PPS) signal 10 MonIPÊ Service Pilot Pilot: • the MonIPÊ service piloted in a few client sites connected to two RNP PoPs Deployment: • PoPs: deployed a PoP MP • Institutions: deployed the low cost measurement kits built and provided by the MonIPÊ project Out of scope: • 10 Gbps capable MP Central Portal: • http://portal.monipe.rnp.br/ • Access: guest / guest 11 Pilot Participants • PILOT PARTICIPANTS: IDS Mamirauá 8 Mbps PoP-MG UFV 10 Gbps 310 Mbps IFC UFSC - Videira PoP-SC - LCM 4 Mbps 10 Mbps Gbps 10 12 Pilot Implementation Timeframe: • October - December 2013 First action: • Kick off meeting Main goals: • Deploy the infrastructure • Schedule permanent tests: – amongst PoPs – between PoP and directly connected client sites • Allow on demand test execution After kick off meeting: • PoPs and client sites have received the equipment necessary to physically install in their premises 13 Pilot Infrastructure • PILOT INFRASTRUCTURE: MonIPÊ Pilot GPS (Adafruit ) GPS (Adafruit ) Delay Bandwidth (Raspberry PI) (Cubox) GPS (Adafruit ) Delay Bandwidth (Raspberry PI) (Cubox) GPS (Adafruit ) Delay Bandwidth (Raspberry PI) (Cubox) Delay Bandwidth (Raspberry PI) (Cubox) On Demand Tests Mamirauá UFV IFC-Videira LCM Customer Site Scenario Customer Site Scenario GPS PoP-MG GPS PoP-SC IPÊ Network Backbone Scenario MP PoP MP PoP 14 Deployment Example: IFC Videira Cubox component installed in IFC Videira Raspberry Pi component installed in IFC Videira Low Cost Measurement Kit (Cubox, SSD Disk and Raspberry Pi) installed in IFC Videira Adafruit GPS Antenna installed in IFC Videira 15 Pilot Evaluation Criteria • Deployment of the low cost measurement kits executed accordingly by the institutions following the guidance of the development team • The developed solution attends the proposed scenarios • The scheduled measurements executed accordingly • Proper clock synchronisation for the delay MPs Due to time constraints, the MPs located in each PoP were not tuned for best performance – time spent on bug fixing 16 Pilot Results • Solution was deployed as planned (exception: LCM-UFSC – equipment for software improvement) – UFSC: installed in a notebook (Measurements between Florianópolis campus and PoP-SC) • Given the short period of the pilot, the results are very preliminary and a longer period is needed to guarantee proper functioning of the low cost measurement kit for the long period • Even though, the results are considered appropriate to measure the connectivity of the client directly connected to the Ipê network • Scenarios validated: backbone and client site (international postponed) 17 Pilot Results Problems: UFV: • Raspberry PI damaged in transport – local team managed to glue the broken component • SSD memory failed – replaced by an USB stick lent by the client site Mamirauá: • Software bugs preventing configuration – corrected by updating the measurement portal IFC Videira: • OS in SD flash card (Raspberry Pi) got corrupted – replaced 18 Pilot Measurement Results Client Site Scenario TCP Achievable Bandwidth from PoP-SC to IFC Videira: TCP Achievable Bandwidth from IFC Videira to PoP-SC: 19 Pilot Measurement Results Client Site Scenario TCP Achievable Bandwidth from PoP-MG to UFV: TCP Achievable Bandwidth from UFV to PoP-MG: 20 Pilot Measurement Results Client Site Scenario TCP Achievable Bandwidth from PoP-MG to Mamirauá: TCP Achievable Bandwidth from Mamirauá to PoP-MG: 21 Pilot Measurement Results Client Site Scenario One-way Delay from PoP-MG to UFV: One-way Delay from UFV to PoP-MG: 22 Pilot Measurement Results Client Site Scenario Round trip Time from PoP-MG to Mamirauá: Round trip Time from Mamirauá to PoP-MG: 23 Pilot Conclusions and Lessons Learnt • Send components in proper packaging to avoid damaging during transport, not attaching components • Data loss must be investigated (development issues or hardware/OS resources) • Monitoring infrastructure required for operations • Pilot Focus: validate measurement functionality and execution, and bug fixing of the components • Customisation of measurement environment required • Results considered remarkable (most issues were problems of real world environments) 24 Users Feedback • All interviewees considered the service highly relevant to their client site • The metric considered most relevant was TCP achievable bandwidth • The measurement scheduling considered most relevant was on demand, mainly because most users are not used to analyse these metrics in a regular fashion • Some new features requested by the users include: – – – – Scheduling of tests amongst client sites Support to traceroute tool Support to link availability measurements Support to SNMP interface counters 25 Users Feedback • Ideas for improvement in the following areas: • • • • Deployment process Service experience Usage of the service by the client sites General comments about the MonIPÊ service 26 Future Work Development Roadmap • Improvements in the measurements portal interface • Support new tools: Traceroute and Network Diagnostics Tester (NDT) • Build a new low cost measurement point to run both delay and achievable bandwidth tests on the same device, using separate network interfaces 27 Future Work Deployment Roadmap • Migration to virtualized MPs on the backbone, as most of the hosts nowadays are still on physical nodes • Deployment of some 10G enabled MPs on selected PoPs and on international connections • Spread a higher number of low cost measurement kits to attend metro and campus networks of research and education institutions throughout Brazil 28 MonIPÊ Team Project Coordination: RNP Research and Development Directorate • Michael Stanton – Director • Iara Machado – Deputy Director • Alex Moura – Manager • Fausto Vetter – Coordinator Development • Edison Melo – Administrative Coordinator • Murilo Vetter – Development Coordinator • Rodrigo Pescador – Hardware / Infrastructure • Guilherme Eliseu Rhoden – Hardware / Infrastructure • Paulo Brandtner – Web Expert / Development • Luis Fernando Cordeiro – Web Expert / Development 29 1
© Copyright 2026 Paperzz