Re: Number of servers required to process half TB of data
- From: "SGCSNA" <sean.gorman@xxxxxxxxxxxxxxxxx>
- Date: 2 Aug 2006 10:49:28 -0700
I guess it just depends on the data, the indexes, the schema, the
hardware..... LOL.
Seriously though,
one thing you might want to consider is using a high-end array
subsystem that would have a large read/write cache. Naturally,
everything will need to be tuned to support this one-second query you
are referring to. Most of the higher end systems have blazing fast I/O
when configured properly. Using one very powerful machine hooked into
one of these arrays with everything configured properly could
potentially get you satisfactory results.
jatin.gro...@xxxxxxxxx wrote:
Hi all,
For the development of an idea, I am trying to find out how many
servers will be required to process half a terabyte of data, in one
second? To clarify the problem, we have 13 weeks' daily POS data that
runs into half a terabyte. I think the only way to have the result of a
select query on this amount of data would be to have it all in RAM. So,
half terabyte of RAM, spread across perhaps 32 computers, working over
a fiber channel, or something else blazing fast. Would such a setup be
able to give me a response to a select query with a where clause, in a
second, or maybe a couple of seconds? Can I get by using less
computers?
Need help in figuring this out.
Thanks,
Jatin
.
- Follow-Ups:
- Re: Number of servers required to process half TB of data
- From: jatin . grover
- Re: Number of servers required to process half TB of data
- References:
- Number of servers required to process half TB of data
- From: jatin . grover
- Number of servers required to process half TB of data
- Prev by Date: Number of servers required to process half TB of data
- Next by Date: Re: Number of servers required to process half TB of data
- Previous by thread: Number of servers required to process half TB of data
- Next by thread: Re: Number of servers required to process half TB of data
- Index(es):
Relevant Pages
|