 |
 |
Ten points to those that noticed my spelling mistake. MariaDB not MarinaDB, the slightly less successfully Vauxhall MySQL Implementation :-) On 26 Jun 2013, at 10:00, Mark Mitchell wrote: Hi all, sorry for missing the meeting yesterday but moving office kinda got in the way. The update from Paris is relatively straight forward. The LHCONE network environment is expanding but slowly. As this is effectively a Private Internet the setup requirements are probably more complex than most sites are presently used to. In our case for the UK we would require two devices to terminate the current connection to JANET and one to terminate the network connectivity into LHCONE. ESNET should be supplying a series of architectural guidelines for connectivity into the environment in the coming months but what I have described above is the recommended design for the service. LHCONE will effectively be an MPLS based network with a rather complex BGP configuration as the networks aren't held with one AS. The support for QOS and COS tagging of packets is still under discussion and will be expanded upon at the next meeting in the US. Presently, there is no support for IPv6 on the LHCONE but representatives from DESY, CERN, CALTECH and GRIDPP (Me) discussed this as a future option. Other areas of interest was work conducted at CALTECH on Software Defined Networking using Openflow and networking monitoring using either SFLow or Netflow. Netflow is presently the solution of choice but its limitations are becoming evident as it doesn't supply the granularity that is probably required for monitoring routing updates. I will clarify that statement. It doesn't report on any routing protocol state changes at present, which is going to be key for optimal path selection for data transfer. SFlow contains this functionality but isn't designed for Wide Are Deployments. Further discussion is required around this area. The GridPP rep restated the fact that SNMP will supply virtually all the data required and is a standard that is adopted by virtually all network and OS vendors. Sflow and Netflow are vendor specific in their implementations and are still under development. therefore, they are unsuitable for this Type of Wide Area Monitoring. Again further discussion is required in this area. An interesting presentation was given by Michael Ernst of BNL on the ATLAS plans for networking and the utilisation of OpenStack specifically for ATLAS. This presentation is worth looking at as it also gives figures for using Amazon and issues with spot pricing, etc. At the last workshop at CERN I decided to have a look at building a Content Delivery Network for HEP. There was some discussion with ATLAS and CMS around this area at the meeting and BNL are going to look at the technology I suggested to see if it is scaleable. From a UK perspective the LHCONE presently offers us few advantages over the current JANET network. However, Prof Pete Clarke and myself are going to speak to JANET in July about supplying an LHCONE connection to sites in the UK on a trial basis so that the UK is more actively involved in the programme and so that we can evaluate the technologies involved in delivery of services. IPv6 Update Currently, the Glasgow test cluster is being rebuilt using Puppet 3 and Puppet 2.6. We have deployed SL5 and SL6 in the environment and are in discussion with Milan about building a dual stack CE for the HEPIX testbed. There is an IPv4 CE in place presently.The DPM service is running at Glasgow and is ready for testing. We have solved the MySQL issue reported with SL5 and SL6 by installing a MarinaDB back end so that the MySQL can support IPv6. This issue was caused by neither SL deployments supporting an up-to-date version of MySQL. The minimum specification for IPv6 support is 5.5. Additionally, we have been in discussion with Extreme Networks about having an Openstack enabled test Cluster. The test equipment includes a Big Switch software control unit for VM deployment. We will report on this and its integration into Openflow and the Grid Middleware hopefully at the next GridPP. The CERN IPv6 test node is back within the test framework. Tony Wildish from CMS is continuing his load tests on the system. Also, there is a new wiki page for the IPv6 environment. I will add this to the GridPP IPv6 Page. regards, Mark -------------------------------------------- Mark Mitchell, ScotGrid Technical Co-ordinator, Rm 481, Kelvin Building, School of Physics and Astronomy, University of Glasgow, G12 8QQ, UK Telephone: +44-141-330 6439 E Mail: [log in to unmask]<mailto:[log in to unmask]> -------------------------------------------- Mark Mitchell, ScotGrid Technical Co-ordinator, Rm 481, Kelvin Building, School of Physics and Astronomy, University of Glasgow, G12 8QQ, UK Telephone: +44-141-330 6439 E Mail: [log in to unmask]<mailto:[log in to unmask]>