[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Discuss] Whence distributed operating systems?
- Subject: [Discuss] Whence distributed operating systems?
- From: richard.pieri at gmail.com (Rich Pieri)
- Date: Fri, 22 Apr 2016 10:08:36 -0400
- In-reply-to: <20160422011955.GX12314@randomstring.org>
- References: <email@example.com> <firstname.lastname@example.org> <20160422011955.GX12314@randomstring.org>
On 4/21/2016 9:19 PM, Dan Ritter wrote: > (The Debian method is se set up so that /etc/apache2-NAME is a > completely different copy of apache, and will be controlled by > apache2ctl-NAME, handle its own config in /etc/apache2-NAME, > and so forth. The only thing left to sysadmin whim is making > sure that they Listen on different IP/port combos. Say you need 20 thousand separate instances of Apache. That's 20 thousand /etc/apache2-$NAME directories which is going to cause performance and maintenance issues on the file system. > But lots of services are designed to do that sort of hotel > service well, especially if you have a separate IP available for > every instance. If. One of the problems with scaling up very large SSI is that you don't have lots of IP addresses available. You have many nodes and each node has a unique IP but you can't use these addresses because the processes you want probably aren't there. Everything has to point to the SSI host's master interfaces somehow. Not only do you need 20K apache2-$NAME directories you also need 20K virtual IP addresses choked down to the master interfaces for the cluster. You can do it but it's not nearly as easy as you think when you scale it up. -- Rich P.