sim@p1 how to profit of the atlas hlt farm during the ls1 after n.
Download
Skip this Video
Loading SlideShow in 5 Seconds..
Sim@P1: how to profit of the ATLAS HLT farm during the LS1 & after PowerPoint Presentation
Download Presentation
Sim@P1: how to profit of the ATLAS HLT farm during the LS1 & after

Loading in 2 Seconds...

play fullscreen
1 / 7

Sim@P1: how to profit of the ATLAS HLT farm during the LS1 & after - PowerPoint PPT Presentation


  • 67 Views
  • Uploaded on

Sim@P1: how to profit of the ATLAS HLT farm during the LS1 & after. Sergio Ballestrero & Alessandro Di Girolamo. ATLAS infrastructure @ P1. Hardware: mainly 3 bunches of pc 340 Dell PE 1950: 4 cores, 16GB RAM, 80GB disk 128 Dell PE 6100: 8 cores, 24GB RAM, 250GB disk

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about 'Sim@P1: how to profit of the ATLAS HLT farm during the LS1 & after' - yorick


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
sim@p1 how to profit of the atlas hlt farm during the ls1 after
Sim@P1:how to profit of the ATLAS HLT farm during the LS1 & after

Sergio Ballestrero & Alessandro Di Girolamo

15 January

atlas infrastructure @ p1
ATLAS infrastructure @ P1
  • Hardware: mainly 3 bunches of pc
    • 340 Dell PE 1950: 4 cores, 16GB RAM, 80GB disk
    • 128 Dell PE 6100: 8 cores, 24GB RAM, 250GB disk
    • 990 Dell PE 6100: 12 cores, 24GB RAM, 250GB disk
  • Networking infrastructure
    • P1 GPN “bypass” 2 Gbps
    • P1 SFO-Castor links 20 Gbps

15 January

security considerations
Security considerations
  • ATCN network is a more delicate and critical environment than even a Tier0 site
    • Avoid any interference with internal traffic to DCS systems
    • Avoid any risk, even remote, for security
    • Avoid any risk of saturating the GPN link
  • Isolate the network traffic to a VLAN on the Data Net
    • dedicated VLAN, over physical network dedicated to dataflow (DN), with separate subnets that will not be routable to the rest of the control network
  • Isolate the workload on Virtual Machines
    • VMs will only see the VLAN interface, while the control network of the hosts (XPUs) will remain in the ACTN
  • Route the VLAN through the SFO-Castor link
    • dedicated physical links from the DN edge switch to the SFO-Castor switch
    • dedicated fibers from the SFO-Castor switch to B513

15 January

openstack at p1
OpenStack at P1
  • Decouple HLT and Grid production
    • Cloud as an overlay infrastructure
      • Provides necessary management of VM resources
      • Support & control of physical hosts can remain with TDAQ
      • Delegate Grid support to dedicated team
    • Easy to quickly switch from HLT <-> Grid
      • during LS1: monthly full-scale test of TDAQ sw upgrade 
      • maybe in the future also during short LHC stop
  • CERN IT, BNL and CMS on OpenStack:
    • Similar use cases:
      • support if needed,
      • sharing experiences
    • BNL has already part of its resources cloudified: ATLAS is successfully using them

15 January

testbed
Testbed
  • Setup a testbed
    • Few machines of same kind of the one in HLT farm
    • Similar network and VM/host configuration
    • Ideally SLC6 netbooted using the same image needed by the HLT sw
  • Setup OpenStack

15 January

integrating with adc
Integrating with ADC
  • Re-use the ADC experience on Cloud Resources:
    • E.g. VMs contextualization
  • New Panda Resource within CERN site
    • Remote I/O with EOS
    • Different ResourceType in monitoring
      • New attribute to be added in Dashboard to cope with the possible need of monitoring only resources of type e.g. Cloud
  • Pilots through condorg:
    • Condor master at CERN already for other Cloud resources
    • To be evaluated the benefits of having Condor master in the Cloud
  • Setup a Squid for Frontier and CVMFS

15 January

summary
Summary
  • Sim@P1
    • Plan has been agreed
    • BNL has offered its help!
  • Next 2 months to
    • setup the testbed
    • integrate it into Panda,
    • … be ready to use the HLT Farm during LS1

15 January