Re: Linux betas NT in TPC testing, running Oracle8

From: nik <ndsimpso_at_ingr.com>
Date: 1999/05/17
Message-ID: <lJADOzGo#GA.202_at_pet.hiwaay.net>#1/1


jedi_at_dementia.mishnet wrote in message ...
>On Fri, 14 May 1999 08:26:30 -0500, nik <ndsimpso_at_ingr.com> wrote:
>>
>>jedi_at_dementia.mishnet wrote in message ...
>>>On 13 May 1999 10:23:20 -0600, Bob Hauck <bobh_at_wasatch.com> wrote:
>>>>cbbrowne_at_news.hex.net (Christopher Browne) writes:
>>>>
>>>>> On Wed, 12 May 1999 13:52:10 -0500, nik <ndsimpso_at_ingr.com> wrote:
 

>>>>> > I'd like to see you run a full feed (35K+groups, 2GB & 1.5-2million
>>>>> > articles inbound/day) on a 486 with any OS. If you seriously think
>>>
>>> My introduction to Usenet was via a Waffle based BBS
>>> that was running QNX on a 386. The problem with Usenet
>>> full feed is the storage requirement, no the CPU.
>>>
>>I setup my first USENET server in 1988 using a MODEM for the connection
 and
>>UUCP for the feed with BNEWS software, I agree that CPU is not the main
>>factor, but requirements for a full feed today do exceed what can be done
 on
>>a 486 based systems for a variety of reasons:
>>
>>1. Very few 486 systems have PCI, and on those that do the implementation
>>sucks. SO you don't have enough I/O bandwidth available.
>
> You don't need all of PCI's bandwidth to keep up with SCSI-3.

But for a full feed, you'll need more than one SCSI controller unless you plan to expire every 12 hours to keep the binaries heirarchy in check. If you are going to expire that quickly then you'll also be thrashing the history and active files. On the type of I/O that a news server generates, a single SCSI-3 controller could handle 4-5 drives before it is saturated in terms of the I/Os being generated.

>
>>
>>2. The memory capacity of 486 systems is usually too small and the
 interface
>>to slow to handle the memory bandwidth requirements.
>
> My last 486 had 32M and room for more actually.

Wow, a whole 32MB I'm impressed. Typically 486 systems have a 32 bit memory interface vs. 64 bit bit for Pentium and later which means they are severely restricted in terms of bandwidth. They also tend to have BIOS limiations that mean putting more than 64MB of memory iun them means that L2 cache no longer functions.

Anyway, if anyone can point me at a full feed USENET server running on a 486 with 32MB of memory and a single SCSI-3 I/F I'll be willing to withdraw my statement. Note, full feed means at least 35K groups and high performance servers feeding it, for example the server I'm talking about is on the end of a T3 and takes full peered feeds from MCI and from one of the major private news houses at UofOregon (IIRC) So just having a 35K group active file and pulling a feed over a 28K modem doesn't cut it!

--
Nik Simpson
Received on Mon May 17 1999 - 00:00:00 CEST

Original text of this message