Smalltalk › Squeak › Squeak VM

[Fwd: Network performance on Mac and Linux]

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

3 messages Options

Andreas.Raab

[Fwd: Network performance on Mac and Linux]

-------- Original Message --------
Subject: Network performance on Mac and Linux
Date: Mon, 29 Aug 2005 15:15:41 -0400
From: David P. Reed <[hidden email]>
To: David A. Smith <[hidden email]>, Andreas Raab
<[hidden email]>

I'm sure you'd be as excited as I am to hear that I finally found the
last piece of the network performance issue I've been chasing on the Mac
and Linux.

Over the weekend, I built a Mac VM with the same Socket improvements
that I had done on Windows and Linux. It had the same
flaws as the Linux code (not surprisingly, since the SocketPlugin source
code is identical).

The pieces of the puzzle that were Mac/Unix specific were:

1) the rate of polling for I/O events was once per 200 milliseconds,
hard coded into checkForInterrupts part of the interpreter. (Windows
Socket I/O didn't depend on this since it uses high-priority threads to
wait for I/O).
I changed it to once per 1 millisecond. This reduced the delay from
well over 200 milliseconds to about 32 msec., averaged over a number of
packets.

2) The unix driver didn't set TCP_NODELAY as a socket option by default,
but Windows did. I changed the plugin to do this. This cut the measured
end-to-end delay on packets from 32 msec. to < 1 msec.
(Gaah - I had just assumed that this was done compatibly, and hadn't
checked for it).

So I will soon (tonight) have 3 VMs that have higher performance TCP
network code. They should support the VP2PPN layer cross-platform.

I'm wondering whether TCP_NODELAY should be default on or off for
Squeak. The default should be the same on ALL platforms, in any case,
and documented. For "real-time" stuff like Croquet and VoIP, TCP_NODELAY
should be on (as it was for Windows). For batch file transfer,
TCP_NODELAY should be off, because it helps by packing packets full of data.

Hans-Martin Mosner

Re: [Fwd: Network performance on Mac and Linux]

Andreas Raab wrote:

> I'm wondering whether TCP_NODELAY should be default on or off for
> Squeak. The default should be the same on ALL platforms, in any case,
> and documented. For "real-time" stuff like Croquet and VoIP,
> TCP_NODELAY
> should be on (as it was for Windows). For batch file transfer,
> TCP_NODELAY should be off, because it helps by packing packets full of
> data.

Since the option can be set/reset for each socket, the default setting
should be the one which satisfies most naive socket uses, and IMHO this
would be the batch file transfer kind of stuff. TCP_NODELAY only makes
sense for realtime applications where interactive response is critical,
such as Croquet, and these applications should know what they're doing
and should set TCP_NODELAY...

Cheers,
Hans-Martin

johnmci

Re: [Fwd: Network performance on Mac and Linux]

In reply to this post by Andreas.Raab

Don't forget we added the feature to alter socket options via a
primitive call.
Applications could set TCP_NODELAY as needed.

FYI the unix ones supported are below (from the source code).

static socketOption socketOptions[]= {
{ "SO_DEBUG", SOL_SOCKET, SO_DEBUG },
{ "SO_REUSEADDR", SOL_SOCKET, SO_REUSEADDR },
{ "SO_DONTROUTE", SOL_SOCKET, SO_DONTROUTE },
{ "SO_BROADCAST", SOL_SOCKET, SO_BROADCAST },
{ "SO_SNDBUF", SOL_SOCKET, SO_SNDBUF },
{ "SO_RCVBUF", SOL_SOCKET, SO_RCVBUF },
{ "SO_KEEPALIVE", SOL_SOCKET, SO_KEEPALIVE },
{ "SO_OOBINLINE", SOL_SOCKET, SO_OOBINLINE },
{ "SO_LINGER", SOL_SOCKET, SO_LINGER },
{ "IP_TTL", SOL_IP, IP_TTL },
{ "IP_HDRINCL", SOL_IP, IP_HDRINCL },
{ "IP_MULTICAST_IF", SOL_IP, IP_MULTICAST_IF },
{ "IP_MULTICAST_TTL", SOL_IP, IP_MULTICAST_TTL },
{ "IP_MULTICAST_LOOP", SOL_IP, IP_MULTICAST_LOOP },
{ "TCP_MAXSEG", SOL_TCP, TCP_MAXSEG },
{ "TCP_NODELAY", SOL_TCP, TCP_NODELAY },

On 29-Aug-05, at 12:24 PM, Andreas Raab wrote:

>
>
> -------- Original Message --------
> Subject: Network performance on Mac and Linux
> Date: Mon, 29 Aug 2005 15:15:41 -0400
> From: David P. Reed <[hidden email]>
> To: David A. Smith <[hidden email]>, Andreas Raab
> <[hidden email]>
>
> I'm sure you'd be as excited as I am to hear that I finally found the
> last piece of the network performance issue I've been chasing on
> the Mac
> and Linux.
>
> Over the weekend, I built a Mac VM with the same Socket improvements
> that I had done on Windows and Linux. It had the same
> flaws as the Linux code (not surprisingly, since the SocketPlugin
> source
> code is identical).
>
> The pieces of the puzzle that were Mac/Unix specific were:
>
> 1) the rate of polling for I/O events was once per 200 milliseconds,
> hard coded into checkForInterrupts part of the interpreter. (Windows
> Socket I/O didn't depend on this since it uses high-priority
> threads to
> wait for I/O).
> I changed it to once per 1 millisecond. This reduced the delay from
> well over 200 milliseconds to about 32 msec., averaged over a
> number of
> packets.
>
> 2) The unix driver didn't set TCP_NODELAY as a socket option by
> default,
> but Windows did. I changed the plugin to do this. This cut the
> measured
> end-to-end delay on packets from 32 msec. to < 1 msec.
> (Gaah - I had just assumed that this was done compatibly, and hadn't
> checked for it).
>
> So I will soon (tonight) have 3 VMs that have higher performance TCP
> network code. They should support the VP2PPN layer cross-platform.
>
> I'm wondering whether TCP_NODELAY should be default on or off for
> Squeak. The default should be the same on ALL platforms, in any
> case,
> and documented. For "real-time" stuff like Croquet and VoIP,
> TCP_NODELAY
> should be on (as it was for Windows). For batch file transfer,
> TCP_NODELAY should be off, because it helps by packing packets full
> of data.
>
>
>
>
>
>
>
>
>
>
>
>

--
========================================================================
===
John M. McIntosh <[hidden email]> 1-800-477-2659
Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com
========================================================================
===