| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Official Messages
Forum: We're Moving to Krembil Research Institute!
Thread: Transition Questions and Answers |
| No member browsing this thread |
|
Thread Status: Active Thread Type: Sticky Thread Total posts in this thread: 113
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
May have been commented but reading this forum on my phone and tablet, Android, is a seriois challenge!, but the new logo looks horrible in the Android boinc App.
|
||
|
|
alged
Master Cruncher FRANCE Joined: Jun 12, 2009 Post Count: 2360 Status: Offline Project Badges: 
|
NO MORE QUESTIONS ? NO MORE ANSWERS?
----------------------------------------IS THAT THREAD "FROZEN" OR "LOCKED"??  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Will there be storage replication between the IBM and Kembril data centers or will the move consist of a backup/restore type. Seems like a backup/restore type would require a considerable amount of downtime in order to maintain consistency between the data and the database whereas replication would only require turning off of the source at move time. Logic would suggest days (backup/restore) versus hours (replication).
|
||
|
|
Acibant
Advanced Cruncher USA Joined: Apr 15, 2020 Post Count: 126 Status: Offline Project Badges: 
|
Logic would suggest days (backup/restore) versus hours (replication). They mentioned previously that it would involve a couple days of downtime, so that suggests backup/restore. We should be getting an early warning if we want to temporarily increase our work unit caches. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Logic would suggest days (backup/restore) versus hours (replication). They mentioned previously that it would involve a couple days of downtime, so that suggests backup/restore. We should be getting an early warning if we want to temporarily increase our work unit caches.I think "a couple of days" is optimistic at best. Especially considering terabytes of data. I would plan for a week and hope for the best. Look at how long it takes to do a filesystem check after a storage error and all the data is already on disk. I bet it is a day to backup and compress, a day to transmit to other center, another day to uncompress and restore. Then do filesystem checks, database checks, system check out... |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges: 
|
Logic would suggest days (backup/restore) versus hours (replication). They mentioned previously that it would involve a couple days of downtime, so that suggests backup/restore. We should be getting an early warning if we want to temporarily increase our work unit caches.I think "a couple of days" is optimistic at best. Especially considering terabytes of data. I would plan for a week and hope for the best. Look at how long it takes to do a filesystem check after a storage error and all the data is already on disk. I bet it is a day to backup and compress, a day to transmit to other center, another day to uncompress and restore. Then do filesystem checks, database checks, system check out... We will perform a repeating set of well built rsyncs starting many days ahead of the outage such that we only have 1 days worth of incremental data to copy during the actual outage. That way the data that has to be transferred during the outage consists of just the 1 day incremental sync (which is still a lot) and the databases (which are less than 100GB total). [Edit 1 times, last edit by knreed at Oct 25, 2021 9:31:31 PM] |
||
|
|
Billy Ewell 1931
Cruncher Joined: Mar 1, 2008 Post Count: 23 Status: Offline Project Badges: 
|
During the transition to Krembil is it possible that the 'K' programmers will modify the IBM/WCG software to include Server Status so that we contributors can manage our computers for CPU dedication; particularly when considering the GPU driven Tasks. In some GPU processed tasks/work units it is virtually mandatory that a CPU or Thread be dedicated to the GPU otherwise the speed of GPU task processing can be dramatically impaired and processing time extended
----------------------------------------Today: 18 November 2021.The work load for the Krembil Transition Team must be immense and I personally appreciate their dedication in executing the transfer of WCG to KRI. It would please me so much if a WCG Tech would simply give a YES or NO reply if SERVER STATUS for task/work unit availability will be incorporated into the WCG operating system software. I have modified my posting a number of times and even then I/we have not received an input from an entity that can address the subject of SERVER STATUS and the immediate information it would provide on the availability of both GPU and CPU work units and thereby greatly affect efficient utilization of all our processing equipment. It makes no management common sense to isolate one or more CPUs to support GPU work units for extended periods when in fact most WCG GPU work units are processed in a matter of brief hours or even minutes. Should this posting be relocated to a different forum, then please do so. Bill [Edit 5 times, last edit by Billy Ewell 1931 at Nov 28, 2021 8:07:20 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Is the project planned to be migrated by the end of the year to coincide with the closing of the financial year? Any update would be appreciated.
|
||
|
|
|