From mboxrd@z Thu Jan 1 00:00:00 1970 From: erik quanstrom Date: Fri, 2 Oct 2009 12:47:21 -0400 To: 9fans@9fans.net Message-ID: <9a5d8bdf7decf5d608c0e914798f9df5@brasstown.quanstro.net> In-Reply-To: <<20954.1254491102@lunacy.ugrad.cs.cmu.edu>> References: <<20954.1254491102@lunacy.ugrad.cs.cmu.edu>> MIME-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit Subject: Re: [9fans] plan9 on vmware esx Topicbox-Message-UUID: 7c77d92c-ead5-11e9-9d60-3106f5b1d025 > Somehow my laptop gets by using 32-bit IP addresses regardless of > whether it's on my home Ethernet, a dialup connection (rare these > days in the U.S., still sometimes useful overseas), or on a > wireless LAN. > > My landline phone and my cellular phone both use base-10 digits, > and even the same number of digits! > > > Hey, I know, let's name all disks by Ethernet MAC addresses :-) if we did that, then we would never be able to have more than one connection to the disk. there needs to be a distinction between the disks network address and the disk address if one wishes to support multipathing for redundency or increased bandwidth. the sas wwn (also used in ata these days, btw) uniquely identifies a target. it's 8 bytes. and looks for all the world like a mac address with a 2-byte prefix (see contrib/quanstro oui). one could identify all scsi (modern) scsi drives by wwn and bury the topology inside the scsi drivers. that's typically what sas drives do. this is all a bit convoluted. aoe otoh has the advantage that a lun's address is shelf.slot. and the driver discovers mac addresses on which the target is visable. modulo driver bugs, one can replug the network while doing io to a target and it just works. one naturally gets multipathing by spreading out requests among the mac addresses associated with a target. - erik