by means of running VMs as a grid jobs in the MolDynGrid Virtual Lab - - PowerPoint PPT Presentation

▶
by means of running vms as a grid
SMART_READER_LITE
LIVE PREVIEW

by means of running VMs as a grid jobs in the MolDynGrid Virtual Lab - - PowerPoint PPT Presentation

Oct 16, 2022 337 likes •527 views

Bioinformatics simulations by means of running VMs as a grid jobs in the MolDynGrid Virtual Lab Andrii Salnikov National Taras Shevchenko University of Kyiv Institute of Molecular Biology and Genetics, NAS of Ukraine www.egi.eu EGI-Engage is

slide-1
SLIDE 1

www.egi.eu

EGI-Engage is co-funded by the Horizon 2020 Framework Programme

  • f the European Union under grant number 654142

National Taras Shevchenko University of Kyiv Institute of Molecular Biology and Genetics, NAS of Ukraine

Bioinformatics simulations by means of running VMs as a grid jobs in the MolDynGrid Virtual Lab

Andrii Salnikov

slide-2
SLIDE 2

2 11/10/2015

VM as a Grid job - Goals

  • Use powerful distributed hardware resources (with

accelerated virtualization support) of the already deployed Grid infrastructure

  • Use this resources as PaaS (interactive access)

– Ukraine’s Medgrid VO case: interactive medical data (stored on grid storages) analyses using software for OS Windows

  • Bring your own software environment with VM

(computational jobs)

– MolDynGrid VO case: bioinformatics simulations with different builds of software

EGI Community Forum 2015, Andrii Salnikov

slide-3
SLIDE 3

3 11/10/2015

VM as a Grid job - Implementation

  • Client side: ordinary job submission

– Let’s start with VM submission live demo!

EGI Community Forum 2015, Andrii Salnikov & (jobName="Rainbow Cloud/VM Test") (runTimeEnvironment="CLOUD/VM") (executable="/bin/true") (rsl_substitution=("CF_LOCATION" "lfc://lfc.grid.org.ua/grid/testbed.univ.kiev.ua/cloud-test") ("SYSTEM_IMAGE" "vm.img") ("SYSTEM_IMAGE_DESCRIPTION" $(SYSTEM_IMAGE).description) ) (inputFiles=($(SYSTEM_IMAGE) $(CF_LOCATION)/$(SYSTEM_IMAGE) "cache=copy" ) ($(SYSTEM_IMAGE_DESCRIPTION) $(CF_LOCATION)/$(SYSTEM_IMAGE_DESCRIPTION) ) ("nfstest_1MB.log" $(CF_LOCATION)/nfstest_1MB.log) ) (wallTime="720") (memory="2048") (count="2") (disk="4096") (environment=("CLOUD_VM_SYSIMAGE" $(SYSTEM_IMAGE)) ("CLOUD_VM_DESCRIPTION" $(SYSTEM_IMAGE_DESCRIPTION)) ("CLOUD_DATA_METHOD" "DISK") ("RUNTIME_PASSWORD" "dynamic") ("NOTIFY_EMAIL" "user@grid.org.ua") )

Demo: http://youtu.be/-OgeQkUI2LQ

slide-4
SLIDE 4

4 11/10/2015

VM as a Grid job - Highlights

  • Server side: developed Rainbow (“ARC in the

Cloud”) framework to support VM lifecycle

– Project origins is interactive access to platform – Runs over Nordugrid ARC in Ukraine (wrappers to run

  • n CREAM CE developed)

– Extended to transparently support computational jobs – Pilot mode added for more flexibility

EGI Community Forum 2015, Andrii Salnikov

Rainbow at a Glance: http://goo.gl/ifC6cl

slide-5
SLIDE 5

5 11/10/2015

VM as a Grid job – How it works

EGI Community Forum 2015, Andrii Salnikov

slide-6
SLIDE 6

6 11/10/2015

Rainbow framework components

  • Pass job info to WN
  • VM life cycle handling on WN

RunTime Environment

  • Helpers for networking interfaces creation
  • Helpers for disk image operations
  • More helpers for notifications/pilots/staging

Rainbow WN Components

  • MAC-addresses leasing for VMs and further

MAC-IP bindings

  • Network connectivity testing
  • Port forwarding and Mailing functions

Rainbow Gateway Helper

EGI Community Forum 2015, Andrii Salnikov

slide-7
SLIDE 7

7 11/10/2015

Rainbow framework install for EL5/EL6/EL7

All required packages are available from Grid Blackjack repo

EGI Community Forum 2015, Andrii Salnikov

ARC CE

yum install rainbow-RTE-CLOUD-VM

EMI WN

yum install rainbow-WN-essentials +qemu-kvm +sudo

Gateway

yum install rainbow-GW-helper +dhcp +iptables +xinetd/systemd Wrappers for CREAM CE yum install rainbow-RTE-CLOUD-VM-CREAM yum install rainbow-WN-CLOUD-VM-CREAM

slide-8
SLIDE 8

8 11/10/2015

Rainbow framework configuration

EGI Community Forum 2015, Andrii Salnikov

No default configuration for CE except default paths. GW Helper: /etc/rainbow-helper.conf:

DHCP_LEASES_FILE=/var/lib/dhcpd/dhcpd.leases DHCP_CONFIG_FILE=/etc/dhcp/dhcpd.conf RAINBOW_STATEDIR=/var/spool/rainbow-helper RAINBOW_LIBEXEC=/usr/libexec/rainbow RAINBOW_MAC_BASE=52:54:00:c1:0d:00 FWD_START_PORT=30000 FWD_END_PORT=40000 FWD_WAN_IP=194.44.249.94 RAINBOW_MAILFROM=rainbow@grid.org.ua

WN: /etc/rainbow.conf:

# NETHELPER NETHELPER_SOCKET="10.75.15.254 10108" # WN NETWORK CONFIG VM_NET_TYPE="direct" VM_NET_INT=“eth0" VM_NET_CHECK_INTERVAL=15 VM_NET_CHECK_COUNT=20 # HYPERVISOR SETUP HV_METHOD="qemu“ +add grid-users allowed to run VMs to “rainbow” group

Simple configuration of cluster infrastructure addresses

slide-9
SLIDE 9

9 11/10/2015

Ordinary Job

  • Specify executable
  • Request software

environment

  • Specify data-staging

Rainbow Job Wrap

  • Automatically pass job

environment to VM

  • Pass data to/from VM
  • rainbow-VM-jobwrap

service to start job execution

Rainbow JobWrap Mode

EGI Community Forum 2015, Andrii Salnikov

slide-10
SLIDE 10

10 11/10/2015

Rainbow JobWrap Data Staging

EGI Community Forum 2015, Andrii Salnikov

WN Disk Image Extract Files Virtual Machine ARC

  • 1. Job Description
  • 2. Stage-in
  • 3. Session Dir
  • 4. Build image with Files
  • 5. Second disk
  • 6. Get results from image
  • 7. ARC Stage-out

ISO DISK VVFAT One-by-one TAR all

slide-11
SLIDE 11

11 11/10/2015

MolDynGrid Virtual Lab Infrastructure

  • Own grid-site (UA-IMBG) for

computations and development

– Production ARC CE – CREAM CE for development

  • 80TB(own)+20TB grid-storage

– MD trajectories database – Jobs data stage-in and stage-out

  • MolDynSub CLI for jobs submission
  • Web Portal for jobs submission and

monitoring

EGI Community Forum 2015, Andrii Salnikov

MolDynSub CLI Grid Storage Web Portal

slide-12
SLIDE 12

12 11/10/2015

MolDynGrid Virtual Lab Researches

EGI Community Forum 2015, Andrii Salnikov

More info on poster #7 from Oleksandr Savytskyi

slide-13
SLIDE 13

13 11/10/2015

Version <= 2.1

  • Legacy ARC submission
  • nly
  • GROMACS MD oriented
  • Old portal backend

Version 3.x

  • New modular design
  • Different software modules

for whatsoever (including GPGPU support)

  • Different infrastructures

support (ARC, CREAM, Rainbow, Gloria)

  • New portal backend

MolDynSub CLI

EGI Community Forum 2015, Andrii Salnikov

MolDynSub CLI submission using Rainbow JobWrap mode live demo!

slide-14
SLIDE 14

14 11/10/2015

Old portal

  • Submission via old CLI
  • Job status pull from ARC

Infosys

  • Trajectories Database
  • UI https://moldyngrid.org

New portal

  • Separate backend from UI
  • API for jobs submission
  • Jobs send push status

updates from every infrastructure

  • Gloria integration
  • UI is under development

MolDynSub Web Portal

EGI Community Forum 2015, Andrii Salnikov

slide-15
SLIDE 15

15 11/10/2015

Gloria Pilots Orchestrator

Proof of concept (Under development)

EGI Community Forum 2015, Andrii Salnikov

Job Factory

JSON RESTful API

Gloria Pilot Service Rainbow Pilot Mode MolDynSub CLI submission Rainbow VM ARC CE CREAM CE FedCloud HTCondor

Management Web-UI DB

slide-16
SLIDE 16

16 11/10/2015

More features and Future plans

  • More Rainbow features:

– Libguestfs for more image processing options – Notify user by e-mail that VM is ready and how to access it – Dynamically change VM’s access password (even for windows) – Different Stage-in/Stage-out handling methods – Gateway helper optional authentication – Compressed images support

  • Future Plans

– GPU PCI pass-through support – LXC/Docker support

EGI Community Forum 2015, Andrii Salnikov

slide-17
SLIDE 17

17 11/10/2015

Acknowledgements

  • This work is partially supported by NAS of Ukraine
  • Acknowledgements to EGI and Dr. Tiziana Ferrari,
  • Dr. Sara Coelho personally for travel grant to EGI

CF 2015

EGI Community Forum 2015, Andrii Salnikov

slide-18
SLIDE 18

www.egi.eu

Thank you for your attention.

Questions?

This work by Parties of the EGI-Engage Consortium is licensed under a Creative Commons Attribution 4.0 International License.