Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
History in Computer Maintenance
26 views
Skip to first unread message
christi...@googlemail.com
May 14, 2013, 3:26:31 AM5/14/13
to
Hey guys,
I am an computer enthusiastic working in HPC for some years and diggin' deeper and deeper into how to monitor and run this bricks properly.
Currently I am with the HPC R&D group at Bull, meditating about the NextGen Monitoring/System Analysis of HPC systems.
I am going to give a talk at the 'International Supercomputing Conference' with the title 'Does Supercomputing #MonitoringSucks?' and I would like to point out where we are coming from in terms of monitoring/maintaining starting in the early days of supercomputing.
With my 30+E years I could only imaging what maintenance of a CDC6600, Cray1,...
Let's say I am in the picture of how to monitor a cluster or grid with of-the-shelf commodity hardware. In my environment there were guys responsible for AIX machines and I am under the impression that this points into the direction. I assume that previous to distributed, commodity hardware systems the term 'Never touch a running system' was kind of what was the politic.
One would handle a kind of monolithic system without much deterministic behaviour that you can write down and give to a greenhorn like me.
It was more 'a know my stuff for years' and if the system heats up at the top left corner someone said "Oh, I know what's happening. It xyz that forces the system to overheat".
OK, I know - it's a little bit story telling, but you might get my point. :)
I would love to hear some examples of what it was like to keep this kind of systems running. To entertain, educate myself and to include it at my talk at ISC13:
http://www.isc-events.com/isc13_ap/presentationdetails.php?t=contribution&o=2101&a=select&ra=eventdetails
Kind regards
Christian
I am an computer enthusiastic working in HPC for some years and diggin' deeper and deeper into how to monitor and run this bricks properly.
Currently I am with the HPC R&D group at Bull, meditating about the NextGen Monitoring/System Analysis of HPC systems.
I am going to give a talk at the 'International Supercomputing Conference' with the title 'Does Supercomputing #MonitoringSucks?' and I would like to point out where we are coming from in terms of monitoring/maintaining starting in the early days of supercomputing.
With my 30+E years I could only imaging what maintenance of a CDC6600, Cray1,...
Let's say I am in the picture of how to monitor a cluster or grid with of-the-shelf commodity hardware. In my environment there were guys responsible for AIX machines and I am under the impression that this points into the direction. I assume that previous to distributed, commodity hardware systems the term 'Never touch a running system' was kind of what was the politic.
One would handle a kind of monolithic system without much deterministic behaviour that you can write down and give to a greenhorn like me.
It was more 'a know my stuff for years' and if the system heats up at the top left corner someone said "Oh, I know what's happening. It xyz that forces the system to overheat".
OK, I know - it's a little bit story telling, but you might get my point. :)
I would love to hear some examples of what it was like to keep this kind of systems running. To entertain, educate myself and to include it at my talk at ISC13:
http://www.isc-events.com/isc13_ap/presentationdetails.php?t=contribution&o=2101&a=select&ra=eventdetails
Kind regards
Christian
christi...@googlemail.com
May 21, 2013, 8:38:20 AM5/21/13
to
0 new messages