Ayende @ Rahien

Refunds available at head office

Performance implications of method signatures

In my previous post, I asked: What are the performance implications of the two options?

image_thumb

Versus:

image_thumb1

And the answer is quite simple. The chance to optimize how it works.

In the first example, we have to return an unknown amount of information. In the second example, we know how much data we need to return. That means that we can optimize ourselves based on that.

What do I mean by that?

Look at the method signatures, those requires us to scan a secondary index, and get the results back. From there, we need to get back to the actual data. If we knew what the size of the data that we need to return is, we could fetch just the locations from the index, then optimize our disk access pattern to take advantage of sequential reads.

In the first example, we have to assume that every read is the last read. Callers may request one item, or 25 or 713, so we don’t really have a way to optimize things. The moment that we have the amount that the caller wants, things change.

We can scan the index to get just actual position of the document on disk, and then load the documents from the disk based on the optimal access pattern in terms of disk access. It is a very small change, but it allowed us to make a big optimization.

Comments

Lee Atkinson
04/03/2012 10:28 AM by
Lee Atkinson

May I suggest that the original question was not clear - you asked what the implications for the methods, but the answer is about the implications for implementing those methods.

Poul Foged
04/03/2012 11:24 AM by
Poul Foged

IMO. This is why IQueryable was invented; Lets the callee see the intentions of the the caller ...

Bart
04/03/2012 11:27 AM by
Bart

I also do not understand why you aren't going with an IQueryable approach. Can you explain why?

Kristof Claes
04/03/2012 11:30 AM by
Kristof Claes

Poul Foged: As Mark Seemann states in his blogpost (http://blog.ploeh.dk/2012/03/26/IQueryableIsTightCoupling.aspx), this is not always a good idea.

gandjustas
04/03/2012 12:02 PM by
gandjustas

If IQueryable is too heavy solution, then you can inplement your own ResultSet type and Linq extension methods.

Johannes Gustafsson
04/03/2012 03:04 PM by
Johannes Gustafsson

I believe the point is to force the user to decide how much data it wants. IQueryable won't help there.

gandjustas
04/03/2012 05:03 PM by
gandjustas

Why user should decide it? What if I want just "default" number of results?

Moreover this repository-like interface is too restrictive. How can I get documents after eTag starting with some prefix?

Rafal
04/03/2012 09:42 PM by
Rafal

So, did you just prove that having more information is better than having less?

Jonathan Allen
04/03/2012 10:14 PM by
Jonathan Allen

I still don't understand. If you are streaming data, just get the first batch and wait until they have consumed enough the trigger the next batch. You don't need to know how many batches will be fetched in total since you are only reading them on demand.

If you are not streaming data, then you shouldn't be using IEnumerable.

Pop Catalin
04/04/2012 06:50 AM by
Pop Catalin

"In the first example, we have to assume that every read is the last read"

That doesn't sound right to me.

Generally when working with a rotating disks (but not limited to), you generally prefetch and use buffers in order to avoid disk seeks, even if this means fetching more data than necesary.

What's the basis for aforementioned assumtion?

Comments have been closed on this topic.