Download presentation
Presentation is loading. Please wait.
Published byAustin Lewis Modified over 9 years ago
1
Sponsored by the National Science Foundation GENI I&M and Monitoring GENI Engineering Conference 14 Boston, MA Sarah Edwards Chaos Golubitsky Jeanne Ohren July 9, 2012 www.geni.net
2
Sponsored by the National Science Foundation2July 9, 2012 Introduction Useful data lives everywhere in GENI –Relationships: slices, slivers, users, resources –Counters: interface traffic, OpenFlow flowspace rules –Measurements: CPU load, memory, bandwidth, latency –Health status: reachability, API functionality We can use this information to… –Troubleshoot issues –Optimize configurations –Help experimenters understand their slice resources –Help experimenters analyze their experiments How do we help each other bring it all together?
3
Sponsored by the National Science Foundation3July 9, 2012 Agenda Introduction –Sarah Edwards, GPO Guest Speakers: –Kevin Bohan, GMOC GMOC Monitoring Demonstration –Anirban Mandal, RENCI Client Authentication & Authorization for GENI XMPP Messaging Service –Martin Swany, Indiana University GEMINI: Active Network Measurement –Prasad Calyam, OSC Measurements on Layer 2 and OpenFlow Paths Bringing It All Together –Jeanne Ohren, GPO Discussion
4
Sponsored by the National Science Foundation4July 9, 2012 GMOC Monitoring Demonstration –Kevin Bohan, GRNOC
5
Sponsored by the National Science Foundation5July 9, 2012 Client Authentication & Authorization for GENI XMPP Messaging Service –Anirban Mandal, RENCI
6
Sponsored by the National Science Foundation6July 9, 2012 GEMINI: Active Network Measurement –Martin Swany, Indiana University
7
Sponsored by the National Science Foundation7July 9, 2012 Measurements on Layer 2 and OpenFlow Paths –Prasad Calyam, OSC
8
Sponsored by the National Science Foundation8July 9, 2012 Bringing It All Together –Jeanne Ohren, GPO
9
Sponsored by the National Science Foundation9July 9, 2012 Bringing It All Together Useful data lives everywhere in GENI –Relationships: slices, slivers, users, resources –Counters: interface traffic, OpenFlow flowspace rules –Measurements: CPU load, memory, bandwidth, latency –Health status: reachability, API functionality We can use this information to… –Troubleshoot issues –Optimize configurations –Help experimenters understand their slice resources –Help experimenters analyze their experiments How do we help each other bring it all together?
10
Sponsored by the National Science Foundation10July 9, 2012 Bringing It All Together Let’s discuss a couple of examples of issues to consider when working on projects –Data Naming –Data Transport Let’s walk through some of the types of data that are being collected or are planned to be collected soon
11
Sponsored by the National Science Foundation11July 9, 2012 Data Naming Example Scenario 1 Scenario 1 – Consistent naming of resources and devices –Resources on two aggregates are sharing a network link, each referencing an endpoint. –Each aggregate names their endpoint when submitting data about the link. –The names must be consistent in order for the consumer to be able to relate the data from both endpoints. Aggregate A Aggregate B
12
Sponsored by the National Science Foundation12July 9, 2012 Slice: urn:publicid:IDN+pgeni.gpolab.bbn.com+slice+joslice 550e8400-e29b-41d4-a716-446655440000 Slice: urn:publicid:IDN+pgeni.gpolab.bbn.com+slice+joslice 550e8400-e29b-41d4-a716-446655440000 Data Naming Example Scenario 2 Scenario 2 – Globally unique and consistent naming –Two aggregates are reporting data on their active slivers including to which slice the sliver belongs. –Aggregate A reports a sliver on the slice by URN (e.g. urn:publicid:IDN+pgeni.gpolab.bbn.com+slice+joslice) –Aggregate B reports a sliver on the slice by UUID (e.g. 550e8400-e29b-41d4-a716-446655440000 ) –The experimenter who created the slice may report I&M data on that slice by slice name (e.g. joslice). Sliver A Sliver B
13
Sponsored by the National Science Foundation13July 9, 2012 Data Naming Example Scenario 2 (cont’d) Scenario 2 – Globally unique and consistent naming –The consumer of the data may need to determine if these two slivers belong to the same slice. –Without consistent naming and namespaces, the consumer of the data has to figure out if and how the two slivers and the experiment data relate. –This is already being addressed by GENI AM API v3 by using the combination of URN and UUID. Monitoring and some I&M projects are adopting the same slice naming. –URN + UUID provides uniqueness over time and space. –How does this affect other projects? –What are some other examples?
14
Sponsored by the National Science Foundation14July 9, 2012 Data Transport Example Scenario Scenario –As an aggregate, I collect data about the slivers that I manage, the resources assigned to those slivers, the resources that I have available, etc and report that data to GMOC. –As an experimenter, I am interested in what resources are available at each of the aggregates. –As an operator, I am interested in statistics on the slivers that have been created/deleted over a period of time.
15
Sponsored by the National Science Foundation15July 9, 2012 Data Transport Example How data is accessed today How do each of these consumers access this data? –Aggregates (ExoGENI, InstaGENI, MyPLC) Push data to GMOC at regular intervals using the GMOC APIs Currently access control is using non-GENI credentials –GENI Clearinghouse (Future) Will provide an API to pull data on slices, users, and projects. –IMF and others Provides a pub/sub interface to allow interested parties with the appropriate credentials to subscribe to data events –I&M (GEMINI, GIMI, INSTOOLS) Provide the ability for the user to push data to an archive on iRODS with metadata. iRODS account holders can control and track who has access to archived data
16
Sponsored by the National Science Foundation16July 9, 2012 Data Transport Example Access Control and Reliability Access control –How do we ensure that the appropriate people are able to access the data? –How do we ensure that the wrong people do not get to the data? –How do we keep the access control from getting too complicated for the users? Reliability –How do we ensure the data makes it to the other end uncorrupted? –How do we ensure that the data is getting recorded correctly? How can we all walk away from the table with access to good, reliable data?
17
Sponsored by the National Science Foundation17July 9, 2012 Data Sources Relational data collected by GMOC Time-series data collected by GMOC Active network measurement data collected by I&M tools Passive host measurement data collected by I&M tools Measurement Data Object Descriptor Other independent monitoring tools
18
Sponsored by the National Science Foundation18July 9, 2012 Data Sources Relational data collected by GMOC –Physical location of aggregate resources –Points of Contact (POC) for each aggregate –Slice Authority Info type, version, operating organization, etc. –Aggregate Info name, version, type, etc. –Slivers for each aggregate –Sliver data who created them, when they were created, current state, etc. –Data about resources within each aggregate VM servers, routers, etc. –Mapping of resources to slivers –Data about interfaces on resources MAC/IPv4/IPv6 addresses, VLAN tags, netmask, etc. Schema: http://groups.geni.net/geni/attachment/wiki/GENIMetaOps/gmocv3.rng http://groups.geni.net/geni/attachment/wiki/GENIMetaOps/gmocv3.rng
19
Sponsored by the National Science Foundation19July 9, 2012 Data Sources Time-series data collected by GMOC –CPU utilization –Disk Utilization per partition –Number of active VMs for hypervisors –Interface traffic counters TX/RX pps, TX/RX bps –OpenFlow stats (per datapath and per sliver) ports, RO/RW rules, TX/RX messages, breakdown of messages by type –Health checks AM is accessible via AM API Details: http://groups.geni.net/geni/wiki/GENIMetaOps/DraftMonitoringMetrics http://groups.geni.net/geni/wiki/GENIMetaOps/DraftMonitoringMetrics
20
Sponsored by the National Science Foundation20July 9, 2012 Data Sources
21
Sponsored by the National Science Foundation21July 9, 2012 Data Sources GEMINI –Provides tools to collect active network measurements Bandwidth, latency –Provides tools to collect passive network and host measurements CPU utilization, memory usage, network traffic count –Data will be stored in measurement store service (coming soon) Will provide pub/sub interface and support high-rate data transfers –Experiment topology and service data stored in UNIS service Queryable history of topology changes –Data can be pushed to iRODS archive Command line interface with access control Web interface with access control Searchable
22
Sponsored by the National Science Foundation22July 9, 2012 Data Sources
23
Sponsored by the National Science Foundation23July 9, 2012 Data Sources GIMI –Provides tools to collect data from experiment nodes bandwidth, delay jitter, datagram loss data CPU load, memory usage, per-process state, system usage data –Collected on OML server –Data can be pushed to iRODS archive Command line interface with access control Web interface with access control Searchable
24
Sponsored by the National Science Foundation24July 9, 2012 Data Sources
25
Sponsored by the National Science Foundation25July 9, 2012 Data Sources Measurement Data Object Descriptor (MDOD) –Measurement data objects have associated metadata that provides information on the schema and provenance of the data –Would like to extend MDOD to cover all types of objects, i.e., software images –Would like to use MDOD schema to define Event Record schema –Plan to archive measurement data objects in an archive system based on iRODS –Facilitates searching and correlating data –I&M group has completed v1 of MDOD schema Working towards a simpler v2
26
Sponsored by the National Science Foundation26July 9, 2012 Data Sources Other Independent Monitoring Data Sources –PlanetLab Monitoring - CoMon http://comon.cs.princeton.edu Provides monitoring statistics at both a node level and a slice level Only covers regular PLC nodes –ProtoGENI Monitoring Node Control Center: https://www.emulab.net/nodecontrol_list.php3?showtype=pcs https://www.emulab.net/nodecontrol_list.php3?showtype=pcs Shared Pool: https://www.emulab.net/showpool.php https://www.emulab.net/showpool.php Testbed Node Availability Stats: https://www.emulab.net/node_usage/ https://www.emulab.net/node_usage/ Experiment Information Listing: https://www.emulab.net/showexp_list.php3?showtype=all&sortby=name&thumb=1 https://www.emulab.net/showexp_list.php3?showtype=all&sortby=name&thumb=1 Encourage new independent tools that provide monitoring or I&M info –more accessible and usable across all of GENI if people collaborate and use interfaces like those we are reviewing today
27
Sponsored by the National Science Foundation27July 9, 2012 Discussion –Data Naming How have lack of globally unique and consistent naming affected other projects? What are some other data naming examples? –Data Transport What are you using that others might find useful? How can we all walk away from the table with access to good, reliable data? –What other data sharing issues have you encountered? –Data Resources What other data resources should we all know about?
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.