On 10/2/07, Orlando Andico <[EMAIL PROTECTED]> wrote: > Well the case with Linux is ALWAYS that... you can get ANYTHING done. > But it will cost you a ton of hairs on your head. >
Yes thats right. and somebody should have built an easy to use system by now. Thats also possible right? Because you can do ANYTHING. Im attracted to linux/*bsd and open source systems because of that. Its the possibility of improving on the cli that im here. not for the love of CLI. > And if it's not a "well-known" problem domain (e.g. base OS, or Web > server, or File-and-Print server, or FTP server....) the Linux or > Open-Source solution will be generally crude and lacking in > functionality. > Yes because somebody did not do something about it yet. > You can do ANYTHING with Linux.. but the solution you glue together > will not be that great, and no self-respecting enterprise will touch > your Rube Goldberg contraption for fear it will break and they will > have no support and indeterminate downtime. > Well thats a closed source mentality. Somebody will eventually pay somebody to fix it for them and then everybody will be using it. We can do anything with linux. And that includes making it easy. I just think that businesses don't want to be paying for technology that will be used by their competitors for free. > I did an extensive review of the available Linux-HA solutions a few > months ago. It's nowhere near the state of the commercial enterprise > offerings. > > For example, LVS only does the Virtual IP thing, it doesn't handle > fencing or robust heartbeat or split-brain resolution. These are very > important functionality that the open-source community simply can't > provide a robust solution for. > > Linux-HA fencing is "shoot the other node" -- using a controllable > UPS. Heartbeat is done over a dedicated ethernet or serial cable. > Switch-over times are in the order of SECONDS. This approach is simply > unacceptable for enterprise scale applications, where a switch-over > time in the MILLISECONDS is often desired. > Now thats an interesting problem... > That said, if you're doing simple stateless web load balancing, LVS > will work fine. But just get into session replication (which is a very > common use case) and you're in for a world of hurt. Anything more > complicated... well, good luck! > Well people with cash can buy commercial clustering solutions. Those who cant well have to fix it themselves. Well if business A spent a few millions to build their cluster and business B built one with the same capability for only a few hundred thousand, who is better off? If a government used tax money to pay a developer to fix it that would be even better. -- Lay low and nourish in obscurity _________________________________________________ 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
