Is this possible?

[rikona]

Hello Dave,

Monday, October 3, 2016, 3:35:52 PM, Dave wrote:

> Quoting rikona <rikona at sonic.net>:

>> Hello Dave,
>>
>> Monday, October 3, 2016, 11:57:28 AM, Dave wrote:
>>
>>> Quoting rikona <rikona at sonic.net>:
>>
>>>> Hello Dave,
>>>>
>>>> Monday, October 3, 2016, 9:13:33 AM, Dave wrote:
>>>>
>>>>> Quoting Liam Proven <lproven at gmail.com>:
>>>>
>>>>>> On 3 October 2016 at 06:57, rikona <rikona at sonic.net> wrote:
>>>>>>> AMD 4or6 core processor [have a 6 core]
>>>>>>
>>>>>> Why?
>>>>>>
>>>>>> Apart from at the very low end, Intel is faster than AMD and has been
>>>>>> since the Core 2 Duo range.
>>>>
>>>>> Top on my AMD box shows 92% idle, can Intel wait faster?
>>>>
>>>> You bring up an interesting point. My large data runs, with many disk
>>>> accesses, might actually be IO bound, so processor speed may not be
>>>> the most important factor.
>>>>
>>>> If I put the data on SSDs, though, will that very expensive storage
>>>> have a short life because of the VERY intensive data access required?
>>>>
>>>> --
>>>>
>>>>  rikona
>>
>>> I think you'd need to profile your usage. When you are running your
>>> i/o bound processes, what does top show? or sar? iotop?
>>
>> Just ran a small test on my old box - very interesting result.
>> Definitely IO bound. iotop gives me 4-6 M/s, sometimes 99.99%. CPU
>> mostly around 20-30%, with some jumps to 70-80% or so. 8G mem full.
>> Swap about half full [but lots of pgms running]. 2 cores, one core
>> often gets very near/at max, when CPU is high, but usually sits at
>> about 60-70% visually [using htop].
>>
>> I didn't think of this until your note, but looks important. Thanks!

> ok so the swapping says you're running out of RAM so add RAM and see.

Doubled mem to 16G - the max for this box. Didn't change the numbers
much. The data job took about all the mem after a short while. Swap is
different - I didn't have a couple of dozen pgms open as I did before,
and didn't have a history of going between them for many hours, so
swap was not used very much. There are a few different tasks
associated with the data job that run alternately. In this test the
core that is maxed out occasionally changes back and forth between the
2 cores. Didn't see that earlier...

If you have one, I'd like an opinion re SSD life when very heavily
used - this might be a good thing for me if it works without resulting
in short-life disks.

Appreciate your help...

-- 

 rikona