non-linearity in the insert rate means you have indexes on some columns.
depending on your situation, mysql can be more efficient if drop those
indexes, do bulk inserts, and then add the indexes again.
On 1/23/10 5:02 AM, "Krishna Chandra Prajapati" <prajapatikc@stripped>
> Hi shawn,
> As the data grows to 20 millions the insert rate will become very slow. In
> such case i am getting 2000 insert/seconds only.
> Therefore my objective is not achieved.
> I cannot slow up the insert rate of 10,000/second. I am getting data
> (inserted by users at this rate)
> Is there any other way to do so. (distributed servers)
> On Thu, Jan 21, 2010 at 7:39 PM, Shawn Green <Shawn.Green@stripped> wrote:
>> Krishna Chandra Prajapati wrote:
>>> Hi list,
>>> I want to insert 10000 records/sec into table. There can be n number of
>>> tables with unique data in each. What are the possible ways to do ?
>>> The manual is your friend. It doesn't hurt to consult it.
>> INSERT ...
>> LOAD DATA INFILE ...
>> And we even have both general and specific suggestions on improving INSERT
>> performance in our Optimization chapter:
>> Warmest regards,
>> Shawn Green, MySQL Senior Support Engineer
>> Sun Microsystems, Inc.
>> Office: Blountville, TN