Services To define the complete service softwarehardware the

  • Slides: 5
Download presentation
Services To define the complete service (software+hardware) the abilities of the software need to

Services To define the complete service (software+hardware) the abilities of the software need to be known, because only then a correct hardware set-up can be installed to achieve a high availability and stable service. Software abilities : redundancy error recovery reliability importance (critical service) easiness of problem recovery problem tracing and debugging ease of movement between instances backup ‘state-full’ or ‘state-less’ scalability (# instances) requirements for memory, CPU, disk space data base required (My. SQL, ORACLE) network access requirements ……. . 24. 08. 2005 Bernd Panzer-Steindel, CERN/IT 1

Hardware possibilities : § standard CPU nodes (single disk, single power supply) § large

Hardware possibilities : § standard CPU nodes (single disk, single power supply) § large space disk server (multi power supplies, redundant disks) § mid-range servers (dual power supply, redundant system disk) + expansion probabilities for more memory and few more disks § two mid-range servers + dual FC disk array in a SAN set-up (mainly for data bases) § combinations of the 4 options § something different ( NFS servers, multi-cpu nodes, etc. ) to be considered : costs , increasing the heterogeneous environment, power consumption (critical area, UPS), complexity 24. 08. 2005 Bernd Panzer-Steindel, CERN/IT 2

List of services (certainly incomplete…): BDII, RB, CE, SE, FTS, my. Proxy, VOMS, g.

List of services (certainly incomplete…): BDII, RB, CE, SE, FTS, my. Proxy, VOMS, g. Lite VO, SRM, Grid. FTP, SFT, UI, g. Lite IO, g. Lite RB, R-GMA monitoring and archive, Grid. View, DPM gridpeek, Firemen, RLS, LFC, COOL, VO boxes, Lxgate, castorgrid, WAN monitoring…. service = software + hardware + administration need to go through with the ‘developer’/responsible for the services to understand the status and the deficiencies what is needed for a high-availability and stable service ? ! software deficiencies higher complexity/costs for hardware and administration 24. 08. 2005 Bernd Panzer-Steindel, CERN/IT 3

FTS, File Transfer Service EXPERIMENTS : user application with FTS GD/GM FTS application support

FTS, File Transfer Service EXPERIMENTS : user application with FTS GD/GM FTS application support DES application server expertise: Tomcat ORACLE consultancy ORACLE application tuning (hardware+ app software + exp. software ADC ORACLE FIO system management: quattor, lemon ADC Linux OS FIO hardware mgmt : disk servers, CPU nodes SRM service Grid. FTP service GRID authentication service Castor disk pools correct performance tuning ‘impedance’ between a lot of items WAN service (ADC, FIO, CS, DES, GD, GM) 24. 08. 2005 Bernd Panzer-Steindel, CERN/IT 4

e. g. Resource Broker § ‘state-full’ job information token renewal downtime to be minimized

e. g. Resource Broker § ‘state-full’ job information token renewal downtime to be minimized § some redundancy via DNS balancing § some scalability via ‘VO instances/copies’ § critical service hardware on UPS, special network connectivity § shared file system as a possible solution, SAN storage does not need much storage, but lot’s of small stuff, syncs § crash clean-up not trivial 24. 08. 2005 Bernd Panzer-Steindel, CERN/IT 5