| 1 |
On 8/21/20 10:15 PM, Caveman Al Toraboran wrote: |
| 2 |
> just to double check i got you right. due to flushing the buffer to |
| 3 |
> disk, this would mean that mail's throughput is limited by disk i/o? |
| 4 |
|
| 5 |
Yes. |
| 6 |
|
| 7 |
This speed limitation is viewed as a necessary limitation for the safety |
| 8 |
of email passing through the system. |
| 9 |
|
| 10 |
Nothing states that it must be a single disk (block device). It's |
| 11 |
entirely possible that a fancy MTA can rotate through many disks (block |
| 12 |
devices), using a different one for each SMTP connection. Thus in |
| 13 |
theory allowing some to operate in close lock step with each other |
| 14 |
without depending on / being blocked by any given disk (block device). |
| 15 |
|
| 16 |
Thank you for the detailed explanation Ashley. |
| 17 |
|
| 18 |
> or did i misunderstand? |
| 19 |
|
| 20 |
You understand correctly. |
| 21 |
|
| 22 |
> i sort of feel it may suffice to only save to disk, and close fd. |
| 23 |
> then let the kernel choose when to actually store it in disk. |
| 24 |
|
| 25 |
As Ashley explained, some MTAs trust the kernel. I've heard of others |
| 26 |
issuing a sync after the write. But that is up to each MTA's |
| 27 |
developers. They have all taken reasonable steps to ensure the safety |
| 28 |
of email. Some have taken more-than-reasonable steps. |
| 29 |
|
| 30 |
|
| 31 |
|
| 32 |
-- |
| 33 |
Grant. . . . |
| 34 |
unix || die |
Archives