Changing performance balance between CPU & non-volatile storage
An enlightening article (if a bit long). I wonder how non-volatile storage being faster than a CPU might affect Pharo systems. Perhaps the importance of migrating between live and backup Images gains in importance to backups via files. https://queue.acm.org/detail.cfm?id=2874238 cheers -ben
On 07-01-16 20:07, Ben Coman wrote:
An enlightening article (if a bit long). I wonder how non-volatile storage being faster than a CPU might affect Pharo systems. Perhaps the importance of migrating between live and backup Images gains in importance to backups via files.
I'm not sure what to think about it. I find it an rather vague and badly written article. 100K iops is not close to saturating what a CPU core can process. The problem is the communication and the article does a really good job at confusing me about that. Further, commercial pricing is value based, not cost. Stephan
On Fri, Jan 8, 2016 at 9:17 PM, Stephan Eggermont <stephan@stack.nl> wrote:
On 07-01-16 20:07, Ben Coman wrote:
An enlightening article (if a bit long). I wonder how non-volatile storage being faster than a CPU might affect Pharo systems. Perhaps the importance of migrating between live and backup Images gains in importance to backups via files.
I'm not sure what to think about it. I find it an rather vague and badly written article. 100K iops is not close to saturating what a CPU core can process. The problem is the communication and the article does a really good job at confusing me about that. Further, commercial pricing is value based, not cost.
Stephan
Yeah, I think I'll have to retract that about being enlightening, rather just a façade of enlightenment. Many of the follow up comments on slashdot hit criticised it along the same lines and that its point of view is that of vendors selling fast-nvram. I probably shouldn't have posted it. Sorry for the noise. cheers -ben
participants (2)
-
Ben Coman -
Stephan Eggermont