PCs as computing farms first experiences Frdric Hemmer

PC's as computing farms: first experiences Frédéric Hemmer http: //wwwinfo. cern. ch/pdp/pc/PCRD February 10, 1998 PC's as computing farms - NCF Task Force 1

Introduction n n PC “R&D” started last autumn Goals • Establish whether PC’s can be used – for data intensive computing – in the context of NA 45/NA 48/Compass CDR – as a alternative to large SMP’s, CS/2’s, etc. – what are the limits • Linux and Windows/NT • Good demonstrator for LHC February 10, 1998 PC's as computing farms - NCF Task Force 2

Overall Status n Hardware (reusable) • • n 7 Dual PII’s @ 266 MHz 1 Alpha PWS @ 433 MHz 1 Sun Ultra 5 @ 270 MHz with Hi. PPI 100 GB disk Initial conclusions needed before NA 48 data taking February 10, 1998 PC's as computing farms - NCF Task Force 3

Background n n Catherine Charbonnier, Fabien Collin, Soren Olofsson, Bernd Panzer, Cristoph Von Praun, Jes Sorenesen Performance (aimed at CDR for Compass) • Network performance • Disk performance • Combination February 10, 1998 PC's as computing farms - NCF Task Force 4

Network performance: Results PC’s interconnected through 100 Base. T 3 Com 3000 switch n Repeated with other H/W n Half duplex behavior n Block size does not matter n Linux uses less CPU than NT Good unidirectional performance Disappointing CPU consumption on NT Disappointing bi-directional performance n February 10, 1998 PC's as computing farms - NCF Task Force 5

PC to PC Network performance February 10, 1998 PC's as computing farms - NCF Task Force 6

Network performance: issues n n n Unexplained 0. 5 MB/s observed with some eepro 100 versions on PCRD hardware, but OK on PCSF Recent DEC E'net boards with chipset > 21140 give poor performance on Linux Surprising results PC/Alpha February 10, 1998 PC's as computing farms - NCF Task Force 7

PC/Alpha Network performance February 10, 1998 PC's as computing farms - NCF Task Force 8

Disk performance n n n PC’s connected to SEAGATE ST 19171 W using two Adaptec 2940 UW NT needs a lot of tuning (default behavior is to swap data out!) Block size, BIOS settings, EDO/FPM does not matter Poor performance Windows NT even worse Memory bandwidth is suspected February 10, 1998 PC's as computing farms - NCF Task Force 9

Disk performance • Striping has no effect • 1 stream 2 stripes : 21 MB/s (22 max) • 1 stream 3 stripes : 21 MB/s (33 max) February 10, 1998 PC's as computing farms - NCF Task Force 10

Disk performance: issues n n Memory bandwidth suspected Need to test with LX/SDRAM RISC PCI does not support variety of boards Combined disk/network performance even worse : 5 -6 MB/s on Linux February 10, 1998 PC's as computing farms - NCF Task Force 11

Memory bandwidth (lmbench) February 10, 1998 PC's as computing farms - NCF Task Force 12

Initial conclusions (1) n n PC’s can be used for low I/O tasks Confirmed with • Simulation on PCSF • NOMAD reconstruction on Linux • NA 45 reconstruction on NT n PC’s are not adequate now as disk servers February 10, 1998 PC's as computing farms - NCF Task Force 13

Initial conclusions (2) n PC world diversity • Large choice, manufacturer independent • Difficulties – EEPRO 100/Base hardware problems – DEC Tulip driver problems – IDE Master/Slave problems on some HW – VGA drivers for S 3 Virge – Lilo for disk > 1024 “cylinders” n PC world changes very fast, what is true today may not be tomorrow February 10, 1998 PC's as computing farms - NCF Task Force 14

PC Technology evolution in 97 n Pentium Pro Pentium II • 50 % raw performance increase • but 50 % cache performance reduction n n SEC new motherboards 440 FX 440 LX (SDRAM, AGP) Recent MB’s embedded SCSI, E’net, VGA 100 Mbit E’net switches standard, 1000 Mbit arriving February 10, 1998 PC's as computing farms - NCF Task Force 15

Next steps n Understand I/O bandwidth problem • Measure LX/SDRAM based PC’s n Measure low-end RISC performance on • Disk • Network (Hi. PPI) n n Experiment with PC/Hi. PPI Management aspect has not been investigated • Wake on LAN • Wf. M February 10, 1998 PC's as computing farms - NCF Task Force 16

Questions n n n Why PC’s, what benefits & drawbacks What are the parameters Should all components be PC’s February 10, 1998 PC's as computing farms - NCF Task Force 17