Attachment #23843 for bug #40907

View | Details | Raw Unified | Return to bug 40907
Collapse All | Expand All




.Dq Li "newfs -f 1024 -b 8192 ..." .
.Pp
If a large partition is intended to be used to hold fewer, larger files, such
as database files, you can increase the
.Em bytes/inode
ratio which reduces the number of inodes (maximum number of files and
directories that can be created) for that partition.

aware of.
First, softupdates guarantees filesystem consistency in the
case of a crash but could very easily be several seconds (even a minute!)
behind on pending writes to the physical disk.
If you crash you may lose more work
than otherwise.
Secondly, softupdates delays the freeing of filesystem

options exist that can help you tune the system.
The most obvious and most dangerous one is
.Cm async .
Don't ever use it; it is far too dangerous.
A less dangerous and more
useful
.Xr mount 8

.Va net.inet.tcp.recvspace
sysctls are of particular interest if you are running network intensive
applications.
They control the amount of send and receive buffer space
allowed for any given TCP connection.
The default sending buffer is 32K; the default receiving buffer
is 64K.

You can adjust the buffer size for incoming and outgoing data separately.
For example, if your machine is primarily doing web serving you may want
to decrease the recvspace in order to be able to increase the
sendspace without using too much kernel memory.
Note that the routing table (see
.Xr route 8 )
can be used to introduce route-specific send and receive buffer size

.Va net.inet.tcp.rfc1323
sysctl.
These extensions should be enabled and the TCP buffer size should be set
to a value larger than 65536 in order to obtain good performance from
certain types of network links; specifically, gigabit WAN links and
high-latency satellite links.
RFC1323 support is enabled by default.

You can do a simple calculation to figure out how many you need.
If you have a web server which maxes out at 1000 simultaneous connections,
and each connection eats a 16K receive and 16K send buffer, you need
approximately 32MB worth of network buffers to deal with it.
A good rule of
thumb is to multiply by 2, so 32MBx2 = 64MB/2K = 32768.
So for this case

circumstances.
.Sh KERNEL CONFIG TUNING
There are a number of kernel options that you may have to fiddle with in
a large-scale system.
In order to change these options you need to be
able to compile a new kernel from source.
The

to boot without those options.
If it works, great!
The operating system
will be able to better use higher-end CPU features for MMU, task switching,
timebase, and even device operations.
Additionally, higher-end CPUs support
4MB MMU pages, which the kernel uses to map the kernel itself into memory,
increasing its efficiency under heavy syscall loads.
.Sh IDE WRITE CACHING
.Fx 4.3
flirted with turning off IDE write caching.

IDE drives lie about when a write completes.
With IDE write caching turned
on, IDE hard drives will not only write data to disk out of order, they
will sometimes delay some of the blocks indefinitely under heavy disk
load.
A crash or power failure can result in serious filesystem
corruption.
So our default was changed to be safe.

on by setting the
.Va hw.ata.wc
loader tunable to 1.
More information on tuning the ATA driver system may be found in the
.Xr ata 4
man page.
.Pp
There is a new experimental feature for IDE hard drives called
.Va hw.ata.tags

than 10BaseT, or use 1000BaseT rather then 100BaseT, depending on your needs.
Most bottlenecks occur at the WAN link (e.g.\&
modem, T1, DSL, whatever).
If expanding the link is not an option it may be possible to use the
.Xr dummynet 4
feature to implement peak shaving or other forms of traffic shaping to
prevent the overloaded service (such as web services) from affecting other

Return to bug 40907

Lines 223-229 Link Here

(-)tuning.7.fixed (-16 / +17 lines)
223	.Dq Li "newfs -f 1024 -b 8192 ..." .	223	.Dq Li "newfs -f 1024 -b 8192 ..." .
224	.Pp	224	.Pp
225	If a large partition is intended to be used to hold fewer, larger files, such	225	If a large partition is intended to be used to hold fewer, larger files, such
226	as a database files, you can increase the	226	as database files, you can increase the
227	.Em bytes/inode	227	.Em bytes/inode
228	ratio which reduces the number of inodes (maximum number of files and	228	ratio which reduces the number of inodes (maximum number of files and
229	directories that can be created) for that partition.	229	directories that can be created) for that partition.
Lines 273-279 Link Here
273	aware of.	273	aware of.
274	First, softupdates guarantees filesystem consistency in the	274	First, softupdates guarantees filesystem consistency in the
275	case of a crash but could very easily be several seconds (even a minute!)	275	case of a crash but could very easily be several seconds (even a minute!)
276	behind updating the physical disk.	276	behind on pending writes to the physical disk.
277	If you crash you may lose more work	277	If you crash you may lose more work
278	than otherwise.	278	than otherwise.
279	Secondly, softupdates delays the freeing of filesystem	279	Secondly, softupdates delays the freeing of filesystem
Lines 288-294 Link Here
288	options exist that can help you tune the system.	288	options exist that can help you tune the system.
289	The most obvious and most dangerous one is	289	The most obvious and most dangerous one is
290	.Cm async .	290	.Cm async .
291	Don't ever use it, it is far too dangerous.	291	Don't ever use it; it is far too dangerous.
292	A less dangerous and more	292	A less dangerous and more
293	useful	293	useful
294	.Xr mount 8	294	.Xr mount 8
Lines 425-431 Link Here
425	.Va net.inet.tcp.recvspace	425	.Va net.inet.tcp.recvspace
426	sysctls are of particular interest if you are running network intensive	426	sysctls are of particular interest if you are running network intensive
427	applications.	427	applications.
428	This controls the amount of send and receive buffer space	428	They control the amount of send and receive buffer space
429	allowed for any given TCP connection.	429	allowed for any given TCP connection.
430	The default sending buffer is 32K; the default receiving buffer	430	The default sending buffer is 32K; the default receiving buffer
431	is 64K.	431	is 64K.
Lines 442-448 Link Here
442	You can adjust the buffer size for incoming and outgoing data separately.	442	You can adjust the buffer size for incoming and outgoing data separately.
443	For example, if your machine is primarily doing web serving you may want	443	For example, if your machine is primarily doing web serving you may want
444	to decrease the recvspace in order to be able to increase the	444	to decrease the recvspace in order to be able to increase the
445	sendspace without eating too much kernel memory.	445	sendspace without using too much kernel memory.
446	Note that the routing table (see	446	Note that the routing table (see
447	.Xr route 8 )	447	.Xr route 8 )
448	can be used to introduce route-specific send and receive buffer size	448	can be used to introduce route-specific send and receive buffer size
Lines 469-475 Link Here
469	.Va net.inet.tcp.rfc1323	469	.Va net.inet.tcp.rfc1323
470	sysctl.	470	sysctl.
471	These extensions should be enabled and the TCP buffer size should be set	471	These extensions should be enabled and the TCP buffer size should be set
472	to a value larger than 65536 in order to obtain good performance out of	472	to a value larger than 65536 in order to obtain good performance from
473	certain types of network links; specifically, gigabit WAN links and	473	certain types of network links; specifically, gigabit WAN links and
474	high-latency satellite links.	474	high-latency satellite links.
475	RFC1323 support is enabled by default.	475	RFC1323 support is enabled by default.
Lines 577-583 Link Here
577	You can do a simple calculation to figure out how many you need.	577	You can do a simple calculation to figure out how many you need.
578	If you have a web server which maxes out at 1000 simultaneous connections,	578	If you have a web server which maxes out at 1000 simultaneous connections,
579	and each connection eats a 16K receive and 16K send buffer, you need	579	and each connection eats a 16K receive and 16K send buffer, you need
580	approximate 32MB worth of network buffers to deal with it.	580	approximately 32MB worth of network buffers to deal with it.
581	A good rule of	581	A good rule of
582	thumb is to multiply by 2, so 32MBx2 = 64MB/2K = 32768.	582	thumb is to multiply by 2, so 32MBx2 = 64MB/2K = 32768.
583	So for this case	583	So for this case
Lines 619-625 Link Here
619	circumstances.	619	circumstances.
620	.Sh KERNEL CONFIG TUNING	620	.Sh KERNEL CONFIG TUNING
621	There are a number of kernel options that you may have to fiddle with in	621	There are a number of kernel options that you may have to fiddle with in
622	a large scale system.	622	a large-scale system.
623	In order to change these options you need to be	623	In order to change these options you need to be
624	able to compile a new kernel from source.	624	able to compile a new kernel from source.
625	The	625	The
Lines 661-671 Link Here
661	to boot without those options.	661	to boot without those options.
662	If it works, great!	662	If it works, great!
663	The operating system	663	The operating system
664	will be able to better-use higher-end CPU features for MMU, task switching,	664	will be able to better use higher-end CPU features for MMU, task switching,
665	timebase, and even device operations.	665	timebase, and even device operations.
666	Additionally, higher-end CPUs support	666	Additionally, higher-end CPUs support
667	4MB MMU pages which the kernel uses to map the kernel itself into memory,	667	4MB MMU pages, which the kernel uses to map the kernel itself into memory,
668	which increases its efficiency under heavy syscall loads.	668	increasing its efficiency under heavy syscall loads.
669	.Sh IDE WRITE CACHING	669	.Sh IDE WRITE CACHING
670	.Fx 4.3	670	.Fx 4.3
671	flirted with turning off IDE write caching.	671	flirted with turning off IDE write caching.
Lines 676-683 Link Here
676	IDE drives lie about when a write completes.	676	IDE drives lie about when a write completes.
677	With IDE write caching turned	677	With IDE write caching turned
678	on, IDE hard drives will not only write data to disk out of order, they	678	on, IDE hard drives will not only write data to disk out of order, they
679	will sometimes delay some of the blocks indefinitely when under heavy disk	679	will sometimes delay some of the blocks indefinitely under heavy disk
680	loads.	680	load.
681	A crash or power failure can result in serious filesystem	681	A crash or power failure can result in serious filesystem
682	corruption.	682	corruption.
683	So our default was changed to be safe.	683	So our default was changed to be safe.
Lines 692-699 Link Here
692	on by setting the	692	on by setting the
693	.Va hw.ata.wc	693	.Va hw.ata.wc
694	loader tunable to 1.	694	loader tunable to 1.
695	More information on tuning the ATA driver system may be found in	695	More information on tuning the ATA driver system may be found in the
696	.Xr ata 4.	696	.Xr ata 4
		697	man page.
697	.Pp	698	.Pp
698	There is a new experimental feature for IDE hard drives called	699	There is a new experimental feature for IDE hard drives called
699	.Va hw.ata.tags	700	.Va hw.ata.tags
Lines 747-753 Link Here
747	than 10BaseT, or use 1000BaseT rather then 100BaseT, depending on your needs.	748	than 10BaseT, or use 1000BaseT rather then 100BaseT, depending on your needs.
748	Most bottlenecks occur at the WAN link (e.g.\&	749	Most bottlenecks occur at the WAN link (e.g.\&
749	modem, T1, DSL, whatever).	750	modem, T1, DSL, whatever).
750	If expanding the link is not an option it may be possible to use	751	If expanding the link is not an option it may be possible to use the
751	.Xr dummynet 4	752	.Xr dummynet 4
752	feature to implement peak shaving or other forms of traffic shaping to	753	feature to implement peak shaving or other forms of traffic shaping to
753	prevent the overloaded service (such as web services) from affecting other	754	prevent the overloaded service (such as web services) from affecting other