On 10/3/07, Rogelio Serrano <[EMAIL PROTECTED]> wrote: .. > So whats the commercially accepted baseline? and what test can you suggest?
Like I said - Maintain application state (forget about web servers, think DB or app servers) - Should be able to rollback in-flight transactions and re-apply them on the surviving node - 50 millisecond or less failover time - Minimal application changes, at most the application should receive an exception, re-connect, and keep going on its merry way > > So if you're just worried about failing over stateless web servers, > > so whats so bad about stateless protocols? i prefer them actually. > specially over the internet. you can do distributed transactions over > over them and stay sane. Because stateless protocols don't solve the vast majority of the world's problems! EVERY software application is stateful. Even HTTP tries (badly) to maintain session state by using cookies. If you are failed over to another web server, how does your session state (cookies and associated data) follow you? This is a huge problem domain that load-balancing alone doesn't even begin to touch. > So enlighten me. What are the big boys capable of? How does oracle do > it? A custom network stack? custom machines? non volatile transaction > memory? election algorithms? Please see above. Some of the other offerings in the market offer true zero-hands failover and true application-transparent failover. And yes, some of them use a custom network stack and custom election algorithms. As I'm just a lowly consultant, I don't have access to the details of the implementations. One of the downsides of closed-source. You'd be amazed at how far and how much closed-source can do. There's nothing that I know of like it in the Free Software world. > cant we at least tell him to try it and that YMMV? > > my main problem with your reply is that you are saying trash linux-ha > and dig into your packet. > > i find it totally unacceptable as a linux advocate to give up and tell > people to go somewhere else. if its not good enough i would like to > find out why and try to fix it. or at least find someone who can > volunteer to fix it and that means sending emails to the linux-ha > guys. Err, yeah. You know after I evaluated Linux-HA I considered making contributions to make it better. But the amount of work involved to get Linux-HA up to "bet the Fortune-50 company" levels is huge. Sure for playing around, YMMV. I don't work in the playing around space. _________________________________________________ Philippine Linux Users' Group (PLUG) Mailing List [email protected] (#PLUG @ irc.free.net.ph) Read the Guidelines: http://linux.org.ph/lists Searchable Archives: http://archives.free.net.ph
