Dear Craig and All,
craig.ben...@oracle.com said: > Just curious as to what benefit you'd get over just LAN based and a > manual DHCP config when the network wire itself is private? Well, the simply answer is because I have always done it that way :-) The longer answer (or at least my thoughts) go like this... We run a set of around half a dozen class C networks. Our end users with PCs, Apples and other devices connect to one of our networks. Most of our server class machines have mutliple interfaces and I tend to make them all live. Thus, most of our servers have multiple IP addresses. Thus, when users connect to a server (for whatever reason, computation, file access, web access or whatever) they reach (most) of the servers via a switched network path and their traffic does not cross any routers. On the same vein of thought, I have tried to run our SunRays on "their own network" with just SunRays, one interface from each SunRay Server and switches involved. Thus, SunRay traffic does not make its way on to otyher networks and other traffic does not make its way into the SunRay private network. As this was all very easy to create using utadm -a nxge1 (or whatever on other architectures) I have always done it that way. I am trying to tackle two problems at the moment. One being users (students) complaining about the "old fashioned" interface of Solaris 10 and the other being that we have been seeing quite slow "interactive response" when running many applications displaying on the SunRays. [Things like netbeans, firefox and others stagger somewhat when moving between windows etc., plus the other sorts of problems other have noted.] On the poor "interactive response" front, I have done all the things people have suggested, setting hires_tick=1 in /etc/system, engineering the network so that the connections to our current two servers and the sunrays all make landfall directly on the same switch, dropping the switch port to the server back down to 100mbps to try to avoid packet loss anywhere and so on. Nothing makes much difference. I have tracked new versions of SunRay server software as they were realeased (one that came out last autumn made a big improvement I should add). Whenever I try to collect statistics on packet loss (or whatever) by using utcapture or looking at switch statistics I never seen to locate any "bad numbers" at all... Anyway, I have digressed. I do not expect swapping to Solaris 11 on the servers will improve performance of the SunRay service (but it would be nice if it did) but at least I may get the users believing they have a "more modern user interface". Anyway, I will think more about what you say. I will need DHCP services for the SunRays, but if I read correctly, running utadm -A 192.168.199.0 would start a DHCP server one the sunray server and thus server the SunRays anyway - yes? And, providing I have this all rigged as a closed network (at the swicth/ethernet level) then I should not get spurious traffic in our out of the netwoerk unless the machine(s) running the SunRay server software starting forwarding IP packets between othjer interfaces and the SunRay serving interfaces. I guess I can control IP forwarding and route announcements using other tools (I guess ipadm and routeadm in Solaris 11). Anyway, thanks for info, thoughts and advice. I will have a go shorlty and post things again later. Thanks, Dave Price _______________________________________________ SunRay-Users mailing list SunRay-Users@filibeto.org http://www.filibeto.org/mailman/listinfo/sunray-users
