host13's been really sluggish for me today even though my linode isn't experiencing any extra load. the control panel says the host load is medium, but the graph is almost full.

it kind of worries me that other linodes can have such a dramatic effect on my performance. but here's my question:
should i be worried that my performance might degrade further as other linodes become busier?
-or-
does the fact that the graph is almost full indicate that this is almost as bad as it gets?

also, i'm going to sneak in a feature request: is there any way we could access more detailed host load information? the graph is great, but the description of medium seems pretty meaningless.

The host looks fine from my perspective. What type of sluggishness are you seeing? Network or disk?


Neither, really. And to be honest, it can get worse. See the previous "io-nice" posting in Feature requests. This is something I'll be working on.


It's not completely meaninless. The graph shows a loadavg of 0-10. The reason even a loadavg of 6 or 7 is still medium is because that's going to be CPU-bound. No one *ever* notices CPU contention -- there's always enough of that to go around.

When the graph says "high", we're in I/O contention territory, and you will likely notice that.

-Chris

so is the load reported by my linode the load for the entire host? Because I just spotted a spike to a load of 20 when my linode wasn't doing much.

It caused all my sysstats modules to fail for 1 update (they update at most every minute), but no big deal.

No, the load reported by the website is the host loadavg.

That high of a loadavg inside your Lindoe means either you were swapping and the limiter kicked in, or there was a period of high disk contention on the host which caused disk-bound processes inside your linode to block.

-Chris