Hongmei Zhang Institute of High Energy Physics, Beijing 1 2017-3-10 - - PowerPoint PPT Presentation

Hongmei Zhang Institute of High Energy Physics, Beijing

2017-3-10 1

 Experiments @ IHEP  Computing environment  The operation management platform  The Billing System  The Billing System  Next Plan

2017-3-10 2

2017-3-10 3

 BEPC:Beijing Electron‐Positron Collider  BESIII (Beijing Spectrometer III at BEPCII)  BESIII (Beijing Spectrometer III at BEPCII)

 Produces ~100 TB/year raw data

y

 >3.5PB now  >5PB in five years  7000 Cores for data  7000 Cores for data

process and physics analysis

2017-3-10 4

 Daya Bay Reactor Neutrino Experiment

 To measure the mixing angle θ13  To measure the mixing angle θ13  Produces ~200 TB/year (2011‐2018)

2017-3-10 5

 Jiangmen Underground Neutrino Observatory (under

construction)

 Start to build in 2014, operational in 2019  Estimated to produce 2PB data/year for 10 years

2017-3-10 6

 L

Hi h Altit d Ai Sh Ob t ( d

 Large High Altitude Air Shower Observatory(under

construction)

 located on the border of Sichuan and Yunnan Province  Expected to be operational in 2019

p p

 ~1.2PB data/year * 10 Years

O it t d ti D t ill b filt d d

 On‐site storage and computing resources. Data will be filtered and

compressed before transferring back to IHEP

2017-3-10 7

 Circular Electron Positron Collider(Future Project)  Circular Electron Positron Collider(Future Project)

 Next Generation Accelerator in China after BEPCII  Will complete its mission about 2021  Estimated to produce 200TB/year raw data for Higgs factory  Estimated to produce 200TB/year raw data for Higgs factory

and >100PB/year for Z factory

2017-3-10 8

 Experiments @ IHEP  Computing environment  The operation management platform  The Billing System  The Billing System  Next Plan

2017-3-10 9

17000 CPU ~17000CPUcores 5PBtapelibrary

2017-3-10 10

Powersupply&coolingsystem ~5PBdiskspace

L l l DIRAC b d di ib d

 Local clusters

 ~13,500 CPU cores

 DIRAC‐based distributed

computing system

13,500 CPU cores

 300 GPU cards

computing system

 ~ 2,000 CPU cores

 Grid site (WLCG)  IHEPCloud based on

 1,200 CPU Cores

Openstack

 ~ 720 CPU cores 

720 CPU cores

2017-3-10 11

 Lustre as main disk storage

 Capacity: 5.7 PB storage

 Gluster system

 734TB storage with replica feature

 DPM & dCache  DPM & dCache

 940TB, With SRM interface

 HSM, with modified CASTOR

 Capacity: 5 PB

2017-3-10 12

L ili i

 Low resources utilization rate

 Less than 60% on average

Less than 60% on average

 Computing resources are non‐shared

2017-3-10 13

 Experiments @ IHEP  Computing environment  The operation management platform  The Billing System  The Billing System  Next Plan

2017-3-10 14

 Goal

 Platform between users and resources

& ll ll f d

 Manage & allocate all resources of IHEP data center  Apply for & manage resources

pp y g

 Compute the bill

 Promoting resource sharing

 Share resources  Improve resource utilization

p

2017-3-10 15

Billing Sharing Jobs Files Equipments Users … Reports DSS Al t Monitoring DSS Alerts

2017-3-10 16

Equipments APP

2017-3-10 17

Fault Statistics Cabinets

 Experiments @ IHEP  Computing environment  The operation management platform  The Billing System  The Billing System  Next Plan

2017-3-10 18

 Objective(stage Ⅰ)

 Support 3 computing resources  Support individual user & experimental group  Apply for resources  Apply for resources  Compute the bill

 Features

 Can be configured  Easy to be used  Easy to be used

2017-3-10 19

HTCondor Local IHEP SSO

HTCAccount

IHEP Billing BEPC BES BSRF O t k Cloud g System HXMT ADS JUNO bill Web Openstack

OpenstackAcc t

CSNS bill HR

• unt

ARP ARP （Academia Resource

Pl i ）

etc. Dirac

Di A t

2017-3-10 20

Planning） Planning）

DiracAccount

Decorator Utils

BootStrap Spring MVC

Utils CKEditor CKFinder Message Layer zTree

Spring

zTree Log4j Shiro

MyBatis

Ehcache

Alibaba Druid（Database connection POOL） Alibaba Druid（Database connection POOL） MySQL MySQL

Bill Price Instance Type Application BIL Management Management LL Afs User Afs Group User Distribute User ARP User SSO User User Role Menu monitor Log Area Dictionary Office Base Config monitor Area Dictionary Office e Config

2017/3/22 22

Apply

• Users select resources(instance type ,time)
• Administrators allocate resources

Apply

• Get bill from Account System automatically

Account

• Get bill from Account System automatically

Compute

• Automatic settlement，export bill
• Users can check the bill

2017-3-10 23

System Manager Openstack Manager Users

Real data

Billing System

• BS. db

Novadb

Keystonedb

Gl db Request and information

• Account. db

Glancedb

2017-3-10 24

dataflow

ACCount System

2017-3-10 25

2017/3/22 26

Bill Cloud bill Dirac bill Cl t bill R i bill Cluster bill Runing bill

2017/3/22 27

2017-3-10 28

2017-3-10 29

 Experiments @ IHEP  Computing environment  The operation management platform  The Billing System  The Billing System  Next Plan

2017-3-10 30

 Support more resources

St D t b t

 Storage, Database etc.

 Integrate monitoring system

Integrate monitoring system

 Mobile client

2017-3-10 31

Th k ！ Thanks！

2017-3-10 32