Re: Frustration with SCSI system
Warner Losh wrote: In message [EMAIL PROTECTED] Douglas Swarin writes: : Ideally, I would use one of the IDE flash-based drives on the market. One : brand is SanDisk, and they take a standard IDE connector and fit into a : 3.5" drive bay. You can get them very reasonably priced up to 128MB or : so, which is just fine for a boot partition. Since flash drives have no : moving parts, mechanical failure is not an issue, and since the root : partition is not written to much, the flash will not wear out for a : long time (flash cells wear out after about 100,000 writes; the flash : drives do load balancing and stuff to ensure that the (many) cells in : the drive are written to evenly). We use these devices heavily at Timing Solutions. Or rather we use a IDE - CF adapter and haven't had any devices wear out. And some of these devices have had rather heavy use. I think that it is closer to 1 million writes per cell, but I don't have my spec sheets handy. The newer devices do 1 million writes per cell. When I left Xylan earlier this year, some of our early (late '94 or early '95) flash devices were just beginning to fail. These were development machines that saw a lot of write cycles, and their home-grown flash filesystem does a pessimal job of rewriting the same cells over and over again. Are you sure that they do write balancing? The indications I have from the base chip technology is that they don't. I could have missed that in the data sheets. It has been a little while since I looked at them, so I might be misremembering. I can't seem to find the data sheets I looked at before. SanDisk does, in the controller chip. Good technology. -- "Where am I, and what am I doing in this handbasket?" Wes Peters Softweyr LLC [EMAIL PROTECTED] http://softweyr.com/ To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: Converting Sun Automounter maps
hi, there! On Fri, 22 Sep 2000, Daniel Eischen wrote: On Fri, 22 Sep 2000, Stephen Hocking wrote: I've stumbling into the wonderful world of auto-mounting, and trying to convert some maps from a Sun box to the FreeBSD format. I have amd.conf set up OK as per the man page, but keep on getting errors when changing into the relevant directorys (like amd can't find an appropriate match). Has anyone been down this path? Somewhat related to this... I've tried to get FreeBSD to behave with Sun automount home directories. When all the (Sun) home directories are on one system, I could get by by exporting the home directory on the Sun and automounting this as /home on the FreeBSD box. But we're now expanding so that home directories will exist on 2 different Sun boxen, and this will no longer work. I've resorted to disable automounting of users home directories on the Suns (where /home/user could be automounted to any Sun box/filesystem) and am now auotmounting /home/machinename instead. So home directories are /home/box1/user23 or /home/box2/user134. FreeBSD can now deal with this by automounting box1 and box2 under /home. amd-utils doesn't seem to support the autofs mapping needed to allow FreeBSD to exist in such an environment where each users home directory is automounted. I remember searching the newsgroups regarding this issue, and there did seem to be some initial support for autofs though. we do automounting user homes this way: --- cut here (amd.home amd map) --- /defaults type:=nfs;opts:=rw,vers=3,proto=tcp,intr,soft,nodevs,nosuid,rsize=8192,wsize=8192; # # master users ros host==master;type:=link;fs:=/usr/home/${key} || rhost:=master;rfs:=/usr/home;sublink:=${key} oleghost==master;type:=link;fs:=/usr/home/${key} || rhost:=master;rfs:=/usr/home;sublink:=${key} nik host==master;type:=link;fs:=/usr/home/${key} || rhost:=master;rfs:=/usr/home;sublink:=${key} [...] sol host==master;type:=link;fs:=/usr/home/${key} || rhost:=master;rfs:=/usr/home;sublink:=${key} ilnar host==master;type:=link;fs:=/usr/home/${key} || rhost:=master;rfs:=/usr/home;sublink:=${key} eug host==master;type:=link;fs:=/usr/home/${key} || rhost:=master;rfs:=/usr/home;sublink:=${key} paulhost==master;type:=link;fs:=/usr/home/${key} || rhost:=master;rfs:=/usr/home;sublink:=${key} # # biboca users al host==biboca;type:=link;fs:=/usr/home/${key} || rhost:=biboca;rfs:=/usr;sublink:=home/${key} # # lark users fjoehost==lark;type:=link;fs:=/usr/home/${key} || rhost:=lark;rfs:=/usr;sublink:=home/${key} # # tiamat users osyahost==tiamat;type:=link;fs:=/usr/home/${key} || rhost:=tiamat;rfs:=/usr;sublink:=home/${key} # # lizard users cs host==lizard;type:=link;fs:=/usr/home/${key} || rhost:=lizard;rfs:=/usr;sublink:=home/${key} # # xaa users xaa host==xaa;type:=link;fs:=/usr/home/${key} || rhost:=xaa;rfs:=/usr;sublink:=home/${key} # # drugon users als host==drugon;type:=link;fs:=/usr/home/${key} || rhost:=drugon;rfs:=/usr;sublink:=home/${key} blayhost==drugon;type:=link;fs:=/usr/home/${key} || rhost:=drugon;rfs:=/usr/home;sublink:=${key} plesk host==drugon;type:=link;fs:=/usr/home/${key} || rhost:=drugon;rfs:=/usr;sublink:=home/${key} --- cut here --- /fjoe To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Limits of TCP in FreeBSD kernel?
Hi All:) I'm looking at building an eight or twelve port fast ethernet router, and I got to wondering how well a FreeBSD box could handle that much traffic. Assume, for the moment, that hardware is not an issue. Assume that I have a gigahertz processor, 4 way interleaved memory, 4 separate fast/wide PCI busses. (Thanks to the guys on -hardware for helping me locate it) I tried ping -f localhost on an 800 MHz athlon, and netstat -w 1 -I lo0 indicated about 80,000 pps. The system was 100% busy doing this, about 85% system usage. I'm thinking this is probably spent largely switching in and out of kernel mode to a) have ping send the packet and b) respond to it. If this is correct, the number of packets it could handle while staying within kernel mode would be considerably larger. Or I could be sniffing glue and the cost of copying packets in and out would exceed that of context switching. How many packets per second could I expect to get under ideal circumstances? Thanks in advance:) --- David To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: Limits of TCP in FreeBSD kernel?
In message [EMAIL PROTECTED] David Miller writes: : I tried ping -f localhost on an 800 MHz athlon, and netstat -w 1 -I lo0 ping -f doesn't use TCP at all. : indicated about 80,000 pps. The system was 100% busy doing this, about : 85% system usage. Sounds about right. You asked ping to do as much work as it can and it is doing it. : I'm thinking this is probably spent largely switching : in and out of kernel mode to a) have ping send the packet and b) respond : to it. If this is correct, the number of packets it could handle while : staying within kernel mode would be considerably larger. Or I could be : sniffing glue and the cost of copying packets in and out would exceed that : of context switching. The number of packets is limited by the bandwidth of the pipe coming into the machine, even for gigabit ethernet. Also, larger packets generally do better than smaller ones due to decreated interrupt overhead. ping -f isn't a good measure of throughput because it uses ICMP packets and isn't coded for maximum speed. ftp may be a better way to estimate this, but you may need have a really large file in order to get good measurements. Since you are doing 5MB/s with ping's 64 byte packets, you should likely get 50-100MB/s with TCP/IP and ftp, plus or minus. Make sure that the file you pick is in the buffer cache so that you don't take the hit of reading it into RAM. Also, you might want to consider using the ttcp testing program to get numbers. Since you are using the loopback device, you don't need to worry about a quality ethernet card, but might if you want to go external to the box. : How many packets per second could I expect to get under ideal : circumstances? 80,000 pps is a 5MB/s or 41Mb/s. For a TCP streaming connection, you can expect much higher data rates. 100Mb/s ethernet easily can be saturated, while 1Gb/s ethernet can be saturated with a few tweaks and using jumbo-frames. Zero copy operations are important for gigabit ethernet because the data rates are so high. Ken Merry [EMAIL PROTECTED] and others have done some work in this area. Warner To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Denying ISDN using Radius
hi, I don't know if someone did this before, I want to deny ISDN connections to my PRI's, unless the user is in the ISDN group. can I make radius allow/deny ISDN based on GID cheers To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
FreeBSD, BSDi differences
I'm still new to FreeBSD and like it so far, but I'm thinking of ways I can use it at work. I hear about BSDi and that it is used in many network appliances like f5's BigIP load balancers and other similar products. What I've read BSDi is used because of it excellent TCP stack and other networking. So how different is FreeBSD TCP stack and networking from BSDi's? Steve B. To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: FreeBSD, BSDi differences
On Sat, 23 Sep 2000, steveb99 wrote: I'm still new to FreeBSD and like it so far, but I'm thinking of ways I can use it at work. I hear about BSDi and that it is used in many network appliances like f5's BigIP load balancers and other similar products. What I've read BSDi is used because of it excellent TCP stack and other networking. So how different is FreeBSD TCP stack and networking from BSDi's? Very little. Both FreeBSD and BSDi are derived from the same code. There are no doubt a few differences, but based on what I've seen in other parts of the BSD/OS code I'd be surprised if theirs had received as much work as our network code has. Kris -- In God we Trust -- all others must submit an X.509 certificate. -- Charles Forsythe [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
putting FreeBSD in an extended partition
I am wondering whether there is a good reason for not putting FreeBSD in a DOS extended partition. I have installed four O.S.es on my laptop and know that could be a limitation if FreeBSD can not be put into a DOS extended partition. I do not see any limitations in the boot loader or kernel that prevent this. Thanks for any enlightment. -Zhihui To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message