blah wrote:
When I asked a few months ago, I was told that I need to be consistently hitting the 50Mb/s limit in order to have it raised. Is that no longer the case?
I can't speak to official policy, but I think that's still basically right, though "consistently" may be open to a little interpretation.
I, also a few months ago, was asked to demonstrate that I could utilize a higher limit, with yes, the Linode Manager graph being an acceptable indicator. In my case, I was going to use the higher bandwidth for occasional transfers, and had much lower regular traffic (so perhaps not that different from your scenario), so I had to initiate a transfer to exhibit saturation on the Linode Manager graph. I ran it long enough to show a clear marker in the graph, so burned a few extra GB of bandwidth allotment that month.
Quote:
It is fairly easy to hit the limit, but not consistently and not for an extended period of time (it would have to be hit for 5 minutes straight to show as such in the bandwidth graph).
If you don't regularly, or can't on demand, show such saturation (I'm not sure I'd consider 5 minutes "extended"), then it seems to me that actually raising the limit would have modest benefit, while should it be utilized, it would be, sort of by definition, due to atypical patterns that might be able to cause more harm with the higher thresholds.
Even shorter transfers can create significant spikes in the graph, but you're right that to show saturation you'll have to hold a transfer for a bit longer than the averaging interval. I think that's a relatively modest threshold to use, but can see where that's debatable.
If you really do have usage patterns where transfers are always below 5 minutes, yet could fully utilize a higher limit, I'd probably try taking it up with support again with appropriate supporting evidence to use instead of the Linode Manager graph. When I was dealing with them, they seemed to just want a clear indication that I could utilize the higher threshold, and I suspect they're smart enough to appreciate different forms of "utilize". The answer might still be no, but it couldn't hurt to try, particularly if you do your homework to make a strong case with your own data and type of utilization.
-- David