BNL's dCache

At the Brookhaven National Laboratory we are currently using dCache 1.9.0-9 on the core servers and on the pools.

Configuration

production

Name	# cores	RAM	Description
dcadmin01	4	8G	dCache admin console.
dcdcap02	2	4G	dcap door.
dcdcap03	2	4G	dcap door.
dcdoor10	8	18G	GridFTP? door.
dcdoor11	8	18G	GridFTP? door.
dcdoor12	8	18G	GridFTP? door.
dcdoor13	8	18G	GridFTP? door.
dcdoor14	8	18G	GridFTP? door.
dcinfo	4	5G	Information provider.
dcmaint01	4	8G	dCache services monitoring.
dcsrm	4	8G	SRM server.
dcsrmdb01	4	8G	SRM database.
pnfs01	8	48G	PnfsManager? + pnfs 32-bit + PostgreSQL? 64-bit.

dcadmin02	4	8G	admin console replica.
dcsrm02	4	8G	SRM server (backup).
srmdb02	4	8G	SRM database replica.
pnfs02	4	8G	pnfs database backup.
pnfstest	8	18G	pnfs server replica.

#	Type	Range	Setup	Space per machine	Total
27	Thumper	dc001 to dc028	1 pool of 3.75 TB + 1 write pool of 350 GB + 1 read pool of 3.8 TB (DATAPE) + 1 write pool of 200 GB + 1 read pool of 7.5 TB (MCTAPE)	15.59 TB	~421 TB
31	Thumper	dc030 to dc060	2 pools of 7.75 TB each + 1 write pool of 350 GB + 1 read pool of 3.8 TB (DATAPE) + 1 write pool of 200 GB + 1 read pool of 7.5 TB (MCTAPE)	~31 TB	961 TB
10	Thor	dc062 to dc071	8 pools of 7.75 Tb each + 1 pool of 200 G (HOTDISK)	~62.2 TB	622 TB
20	Thor	dc072 to dc091	12 pools of 7.75 Tb each + 1 pool of 200 G (HOTDISK)	~93.2 TB	1864 TB
					~3.77 PB (3868 TB) out of 3.92 PB (4014 TB)
1	Thumper	dc002	Thumper image.	32 TB
1	Thor	dc061	Thor image	32 TB

Type	Name	# cores	RAM	Description
Hypervisor	gridsrv???	-	-	Hypervisor.
VM	prioritystager	1	2G	US ATLAS priority staging web service.

pre-production (pps)

Type	Name	# cores	RAM	Description
Hypervisor	gridsrv005	8	22G	Hypervisor.
VM	dctest11	2	4G	PnfsManager? + pnfs database.
VM	dctest13	2	4G	SRM server+database.
VM	dctest12	1	1G	admin console.
VM	dctest14	2	2G	GridFTP? door and pools.
VM	dctest15	1	1G	dcap door.
VM	dctest16	1	1G	build machine and SRMWatch.
Stand-alone	dc061			8 read/write pools.

use "ssh -l admin -i /opt/d-cache/identity -c blowfish -p 22223 localhost" to get in admin console

development (dev)

The development environment is composed by a set of virtual machines for the dCache core servers and pools.

Type	Name	# cores	RAM	Description
Hypervisor	gridsrv006	8	22G	Hypervisor.
VM	dctest05	2	4G	PnfsManager? + pnfs database.
VM	dctest07	2	4G	SRM client + SRM server + database.
VM	dctest06	1	1G	admin console.
VM	dctest08	2	2G	GridFTP? doors.
VM	dctest09	1	1G	dcap door and pools.
VM	dctest10	1	1G	build machine and SRMWatch.

use " ssh -l admin -i /opt/d-cache/identity -c blowfish -p 22223 localhost" to get in admin console

Documentation

presentations

• 20090407_08 dCache hands-on workshop (Horsall)
• 20090407 pnfs to Chimera migration @ NDGF (Horsall)
• 20090407 pnfs to Chimera (Horsall)
• 20090407 pnfs to Chimera migration tutorial (Horsall)
• 20090114 PostgreSQL tips for dCache II (FZK T1 Workshop)
• 20090114 PostgreSQL tips for dCache I (FZK T1 Workshop)

Interventions

• 17 Jun'09: dCache pool upgrade to 1.9.1-8

Monitoring

• dCache service monitoring
• ATLAS ARDA DDM Dashboard
• Panda Mover monitoring for BNL-OSG2_MCDISK
• SRM Watch
• RSV probe

• eLog

• BNL ATLAS Computing Facility GCE Servers Grid (Ganglia)

Procedures

	Description
BNL-SM-USATLAS-001	Procedure to fulfill US ATLAS' requests to add disk to a space token.

Space tokens

Information regarding US ATLAS space tokens according to information given by Armen & Kaushik.

Space token	Description
DATADISK	D1T0? , disk location for data you need frequent access - ESD, DPD, something else on request.
DATATAPE	D0T1? , buffer area for data production and transfer going to tape.
GROUPDISK	space token shared and managed by physics groups, to store datasets of common interest within each physics group community.
LOCALGROUPDISK	more permanent storage area for users, to store their favorite data. Usage policy defined, as well as managed by local site. Need some type of quota enforcement.
MCDISK	D1T0? , location for MC data you need frequent access - ESD, DPD, something else on request.
MCTAPE	D0T1? , buffer area for MC production and transfer going to tape.
USERDISK	scratch area, to be regularly cleaned, to store user analysis output (and used for trans-cloud transfers?).
SCRATCHDISK	scratch area, to be regularly cleaned, to store user analysis output (a replacement to USERDISK).

---

PedroSalgado - 27 May 2009