Operations/Platform Deployment Survey
|Main||EGI.eu operations services||Support||Documentation||Tools||Activities||Performance||Technology||Catch-all Services||Resource Allocation||Security|
Please reply to both Question 1 and Question 2.
DEADLINE for reply: 20 March 2012
Back to the operations survey main page.
Question 1: Deployment of platforms
Grid software is being provided for multiple OS platforms. For example, the number of supported platforms will increase with EMI 2.0 (Spring 2012): in EMI 2.0 it is expected that all products will be released for sl6, and a subset also for Debian.
EGI operations need to review the list of early adopter sites, to make sure that staged rollout resources are allocated for testing of software on the platforms of interest.
All Resource infrastructure Providers are requested to consult with their site managers to define their priorities. Please define your deployment priorities of sl6 and Debian for 2012 for products released by EMI (ARC, dCache, gLite, UNICORE) and IGE (Globus) by replying to questions in the following table.
Are your sites insterested in installing some EGI and/or IGE components on sl6-based distributions? If so, which components? in how many sites?
Are you insterested in installing some EGI and/or IGE components on Debian-based distributions? If so, which components? in how many sites?
| Other distributions
Specify if you are interested in installation of EGI and/or IGE components on other platforms and if so, which components in how many sites
|Example: NGI_ZZ||dCache, DPM, and EMI WN (4 sites)||EMI UI (10 sites)||no|
|Belarus||We (4 sites) are interested in SL6 versions of all components.||We (4 sites) are interested in a Debian/Ubuntu based WN/UI||UIIP have developed UNICORE distribution for Windows and we are ready to maintain this and further develop.|
|CERN||We intend to use components coming from gLite under SL6.||We do not intend to use Debian or any other distribution.|
|Croatia||No plans for upgrade to SL6 in 2012.||No plans to deploy services on Debian.||We use CentOS distribution on all sites. The top priority for us is to have official support for CentOS.|
|Cyprus||No plans for upgrade to SL6 in 2012.||No plans||No plans|
|Czech Rep.||Yes, all major services are running on SL* machines, SL6 is interested.||Yes, there is interest in a Debian based WN. Server support for Debian would be welcomed, however, without better information we don't have concrete plan.||No plans|
|Finland||All 10 clusters but one are already SL6 based, ARC is currently installed from the NorduGrid repo. DPM and SAM Nagios are on SL5 until they can be migrated.||No plans for Debian, except for clients.||No plans|
|Georgia||No plans||No plans||No plans|
|Germany||we are interested in all EMI components for SL6. As we already have most components running at SL.||no plans||EMI (UNICORE, CREAM, dCache, BDII_site) components for SLES 11 and its successors|
|Hungary||Migration from SL5 to SL6 for all components planned when stable releases arrive (1 site)||ARC InfoSys (1 site)||ARC CE, ARC clients (2 sites), Redhat Linux 5.6, Suse Linux 11|
|Ibergrid (Portugal, Spain)|| All sites (the ones that replied, ~2/3 of all the Iberian sites) requested SL6 support of ALL local gLite middleware components (CREAM, SEs (DPM, STORM, DCACHE), APEL, SITE-BDII, UI, WNs, GLEXEC, ARGUS and LRMS_UTILS). There is some special emphasis on clients (UIs and WNs) so that is where the support work should start.
The sites which manage core gLite services also requested SL6 support with special emphasis on WMS, TopBDII, VOMS and LFC
|Only one site show interest in having middleware components supported in Debian/Ubuntu||CentOS seems better supported than SL (reacts faster to new kernel releases, security issues). Some sites have some services deployed in CentOS5. Mark this as "good to have".|
|Ireland||Plan to run some WNs on SL6 in 2012 at one site. When installing new services / releases of other services we will consider SL6.||No plans||No plans|
|Italy||There is high interest in gLite components on Sl6, WN and UI with high priority (survey sent to all Italian sites (54), 22 responses received.)||there is low interest in gLite components on debian (four site), starting from WN and UI (survey sent to all Italian sites (54), 22 responses received.)||there is low interest in gLite components on other platform (one site): the platform is SUSE (survey sent to all Italian sites (54), 22 responses received.)|
|Latvia||No plans for upgrade to SL6 in 2012.||All major sites are interested in Debian/Ubuntu ARC-CE, ARC client, A-REX server and DPM.||No plans.|
|FYR of Macedonia||No plans for upgrade to SL6 in 2012||No plans||No plans|
|Moldova||We are interested in upgrade to SL6||no||no|
|Montenegro||No plans for upgrade to SL6 in 2012||No plans||No plans|
|Netherlands||We are interested in SL6 versions of all components||There is interest in a Debian/Ubuntu based WN/UI||no|
|Poland||All biggest sites (5) are interested in SL6 versions of gLite components. For UNICORE components all sites express interest in SL6 version (seems nothing SL-specific is needed than just CentOS 6).||No plans||No plans|
|ROC Latin America|
|Serbia||We are interested in SL6 versions of all components (all sites).||There is an interest for Debian/Ubuntu based UI (large number of Grid users).||No plans.|
|Slovakia||We are interested in SL6 versions of all components (4 sites).||No plans.||No plans.|
|Sweden||Upgrade from SL5 to SL6 likely on one site||Ubuntu in use at one site||4 sites currently use Centos5|
|Switzerland||CREAM-CE, DCACHE, WN, GLEXEC, DPM servers/clients, site-bdii, glite-UI, ARC CE/ARC client. With lower priority ARGUS, VOBOX, NAGIOS, APEL, CLUSTER.||YES. Components needed: ARC client and A-REX server. GridFTP server.||DPM disk servers on Solaris. We have some, but the support is not assured.|
|Turkey||All major services and sites are configured on SL and we are interested with the components on SL6: CreamCE, DPM server/disk,WN, site-bdii, gLite-UI. The other services like VOMS, WMS, LFC_mysql, Nagios, top-BDII with lower priority.||No plans||No plans|
|United Kingdom|| All sites that responded (11) said yes to deployment of SL6 based deployment of EGI middleware. The probable uptake will be DPM disk servers first followed by the service nodes and finally the Worker Nodes as and when the user community is ready.
Specific components would be: StoRM (inc gridftp with checksumming) CreamCE; bdii; apel; Argus; glexec; worker node (to be transitioned depending on experiment requirements); Squid.
|Only one site interested in Debian for services generally. One other for UI.||Some interest in: UI for Ubuntu; WNs being more portable in general; UI for MacOS & iOS; CentOS5|
|Ukraine||We are interested in SL6 versions of all components of EMI (ARC A-REX, dCache, CREAM CE, gLExec, WMS, Nagios, WN, UI, DPM, StoRM).||Some sites are insterested in Debian/Ubuntu based WN/UI.||CentOS 5 & 6 (~10 sites), Ubuntu Server (1 site) and Fedora (3 sites).|
Question 2: Deployment of virtual images
EGI SA2 is deploying a new Market Place based on StratusLab. The new Market Place will include different Virtual Machines with different middleware services ready to be used, verified and tested by SA2 team. Is your NGI willing to use the new service to download and use pre-installed VMs from the marketplace?
If you replied yes, in which service images are you interested in? eg. User Interfaces (emi-UI), Storage, Job Submission (WMS), Compute Elements (Cream/unicore/arc), etc.
|Example: NGI_ZZ||YES||EMI UI|
|Belarus||YES||For the services that are usually run as Virtual Machines, ex. LFC, VOMS, APEL, siteBDII, certWMS/LB we are ready to use images in production. Ready to deploy images will reduce the time of administration and/or creation of new sites. For the services that are not used as Virtual Machines, ex. WN, CE, etc. we are interested to deploy images in TestBeds. This will make experiments with new software configurations less time consuming.|
|CERN||NO. We do not request any services by EGI-SA2. We may wish to consider sharing virtual machine images across sites in line with the policy developed within the HEPiX Virtualisation Working group.|
|Finland||NO||Not at this stage, but we will investigate it further.|
|Germany||Yes||UI, site-bdii, cream-ce|
|Hungary||Yes||CREAM CE, WN, UI, DPM|
|Ibergrid (Portugal, Spain)||The majority of the sites use VM for services but were reluctant wrt the scenario where the site downloads a service image from a central repository. The main reason is that local images are very customized. Indeed, the big effort is spent in the customization of the images and not on the VM deployment. For that reason, they prefer to deploy an image they've built and control. Nevertheless, if images are available, there is some willingness to give it a try. So, the answers is NES.||Images of the most basic services could be tested: site-bdii, UI, WNs or glite-Cluster.|
|Ireland||Yes||(Note: Yes, assuming that service VMs can be integrated into our fabric management system). WN, CREAM CE, WMS, UI, DPM, (possibly more...)|
|Italy||there is interest/curiosity on VM but...||the provided images should be locally managed by a different cloud infrastructure (i.e. we would use the marketplace but not install the StratusLab stack).|
|FYR of Macedonia||Yes||EMI UI|
|Montenegro||No||Not at this stage, maybe later|
|ROC Latin America|
|Sweden||No||Not at this time, however, we are investigating this.|
|Switzerland||Yes||ARC CE/ARC clients, DPM servers/clients, site-bdii, glite-UI|
|Turkey||Yes||CreamCE, site-bdii, gLite-UI, WN, LFC_mysql, WMS, VOMS, MyPROXY|
|United Kingdom||Significant minority interest in images for some service types.||UI, WN, CE, Storage (including GridFTP server)|
|Ukraine||Yes||UI, WMS, Nagios, Top BDII, ARC CE, CREAM CE|