Frequency of garbage collection


#1

Dreamhost’s ‘unlimited’ resource allocations are attractive but they fail to give guidance about how much is ‘reasonable’.

For example, consider disk usage. In the normal course of website management, garbage accumulates on disk, and the question arises, how often to do garbage collection.

Every day? Every month? Every year? Obviously, the answer depends on how much garbage there is. But in terms of a precise numerical guideline, the answer is not obvious.

Automatic garbage-collection at too-frequent intervals could waste more resources in processor time and disk-head erosion than it would gain in disk-space.

The closest equivalent host to Dreamhost that I’ve been able to find (similar price, similar technical orientation) not offering ‘unlimited’ resource usage allows 100GB disk usage on its basic shared hosting plan.

This seems to suggest that it is reasonable to allow total disk usage on a Dreamhost account to accumulate to up to about 100GB before worrying about how to garbage-collect it.

Comments welcome.


#2

What would constitute “too much” garbage data is kind of subjective. As a rule of thumb, though, I’d say that if you have more than perhaps 1 GB of cache or other “garbage” data, or if it’s more than 10% of the size of the rest of your data, that’s probably too much, and you should clean it up. 100 GB would definitely be too much.

Note that this isn’t an official position here. It’s just my personal recommendation.


#3

Thank you for the feedback, Andrew.

But, I wasn’t suggesting that 100GB of garbage could (in general) be reasonable. It certainly feels like much too much (again, in general). The problem is that many people, myself included, don’t necessarily have a clear idea of how much of their disk utilization is garbage. The real question is: at what point should one put effort into working out how much of it is garbage?

So the suggestion, which I put forward for discussion, is that as the possibly exemplary host which I mentioned seems to feel comfortable with their basic customers using 100GB, such customers might well feel comfortable with not thinking about garbage collection until their total usage approaches that amount.