[Adium-devl] Hardware plans

[Graham Booker]

On Aug 23, 2007, at 10:52 AM, Peter Hosey wrote:

> On Aug 23, 2007, at 05:31:31, Graham Booker wrote:
>> On Aug 23, 2007, at 5:12 AM, Evan Schoenberg wrote:
>>> On Aug 22, 2007, at 11:57 PM, Peter Hosey wrote:
>>>> 2. Would 2 GiB of RAM be enough for each distributed-build node?
>>
>> I can compare 2.5G of RAM to 1G of RAM (two computers with very  
>> similar processor speeds), and there is a major speedup with  
>> successive complies, but even with 2.5G of RAM, it is continually  
>> accessing the disk (reads) during clean builds (which we would  
>> likely want to use).
>
> Sure, but none of the machines have 1 GiB. The question is, while 2  
> GiB is enough for a regular build, does distcc add enough overhead  
> that 2 GiB would not be enough for a distributed-build node?
>

The distributed nodes won't matter all that much, it is the disk  
cache on the node initiating the build that matters (it is a master- 
slave relationship, not a peer to peer).

I seriously think that if we had a collection of minis, the limiting  
factor will not be the processing power, but the contention for the  
single 5400 rpm drive which is inside the computer hosting the build  
process and its cache of the said disk.

Now, if we were talking about the difference between two Xserves, or  
a single Xserve and a bunch of minis, I think the latter would  
perform much better due to the fast disk and architecture of the  
Xserve to host multiple machines.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1601 bytes
Desc: not available
URL: <http://adium.im/pipermail/devel_adium.im/attachments/20070823/ba69f039/attachment.p7s>