Kerrighed?
12 Replies
Worth noting is that the last release was nearly three years ago, and there has been no development activity since October 2011, so I believe it may be a dead project.
With modern, relatively inexpensive, systems able to have 64 physical cores and a TB of RAM in 2U (I can get a 64-core Dell R815 w/ 128GB of RAM from their web site right now for ~$5K, 256GB of RAM for less than $7K, 512GB for ~$10K) there isn't really much of a point.
(EDIT: Or, my personal favorite of the moment, the R720 w/ 2x E5-2690, 256GB of RAM, and 10Gbe for ~$8500)
Most of the time people go looking for SSI it's because they think it will be easier than running a cluster of machines behind a load balancer or easier than rewriting their application for Hadoop or similar. It's almost certainly not.
You've been vague about what you're actually trying to do so I can't say more than that.
@jasonlitka:
Yeah, but what I'm asking is, what possible use could this group have of more resources than is available on a single box like I described?
As the number of members grows, we should expect the number of simultaneous users to grow as well. Clearly, (say) 500 simultaneous instances of bash (and whatever task the user happens to be running at that time) over ssh require more resources than 50.
> You've been vague about what you're actually trying to do so I can't say more than that.
It's a multiuser system. People log in, and do stuff with the shell. Just like you might do on your home PC linux system, except it's multiple people at once rather than just you–in other words, the sort of stuff Unix (and other timesharing systems) were built for.
@BerenErchamion:
It's a multiuser system. People log in, and do stuff with the shell. Just like you might do on your home PC linux system, except it's multiple people at once rather than just you–in other words, the sort of stuff Unix (and other timesharing systems) were built for.
Yea, you dont need SSI for that.
A loose cluster of hosts interconnected via NFS, central host serving as the authoritative source of configs and user accounts, and a load balancer that directs your SSH connection to the host that is least loaded at the moment.
Unless you're running something really massive, you don't really nead "realtime" balancing of existing sessions.
And well, when you have multiple hosts, users that "feel slow" will probably move to other host on their own. Especially if you'll provide a website and/or CLI command to list all hosts' number of users and load averages.
@rsk:
Look up how sdf.org is doing it?
A loose cluster of hosts interconnected via NFS, central host serving as the authoritative source of configs and user accounts, and a load balancer that directs your SSH connection to the host that is least loaded at the moment.
Unless you're running something really massive, you don't really nead "realtime" balancing of existing sessions.
And well, when you have multiple hosts, users that "feel slow" will probably move to other host on their own. Especially if you'll provide a website and/or CLI command to list all hosts' number of users and load averages.
How will this work with users that, e.g., want to run persistent screen sessions?