If the underlying throughput is limited and you want to write more data
than the disk can handle, the only solution is to write less data /
index less keys. And that's more of a workaround which is why I've been
working on more performant blockchain databases:
https://wiki.unsystem.net/index.php/Libbitcoin/Blockchain

On 05/01/14 20:38, William Swanson wrote:
> On Sat, Jan 4, 2014 at 11:55 AM, mlmikael <mlmikael@???> wrote:
>> Can't write requests block internally somehow in such a way that writes
>> always succeed unless there's a physical error about the storage medium?
>
> An obvious way to do it would be to limit the number of outstanding
> network block requests to the number of free slots in the queue. That
> way, you will never overflow, and will throttle back gracefully as
> things fill up.
>
> -William
>