?

Log in

No account? Create an account

Grazr, Memcached, MySQL

« previous entry | next entry »
Aug. 13th, 2008 | 11:50 pm

Patrick, the maintainer of DBD::mysql and one of our original slash crew, is doing a webinar on Grazr:
http://www.mysql.com/news-and-events/web-seminars/

Why is this of interest?

Grazr does write through caching, unlike the 90% percent of Memcached sites which are read through caches.

He also will be talking about Sphinx and I believe he has a slide or two talking about our work with Gearman.

Link | Leave a comment |

Comments {9}

Lisa Phillips

(no subject)

from: lisa
date: Aug. 14th, 2008 06:59 am (UTC)
Link

Thanks for the heads up! I'm actually just breaking sphinx tonight.........

Reply | Thread

Brian "Krow" Aker

(no subject)

from: krow
date: Aug. 14th, 2008 07:18 am (UTC)
Link

In recent weeks I have run into about a dozen sites which are either looking at it or actively deploying it.

Reply | Parent | Thread

Lisa Phillips

(no subject)

from: lisa
date: Aug. 14th, 2008 07:39 am (UTC)
Link

I'd love to talk to anybody who is actively deploying it. We are about to launch a sizable site with it, and have some clients using sphinx too. Load tests tonight didn't show a blip in sphinx performance...our db on the other hand :(

Reply | Parent | Thread

Robbat2

(no subject)

from: robbat2
date: Aug. 14th, 2008 08:10 am (UTC)
Link

We first considered it for isoHunt 2 years ago (went I went to MySQL UC and met the Sphinx author), but ultimately went with Lucene - specifically for issues around indexing of frequently updated items and having the entire dataset changing rapidly.

I spoke to Andrew (sphinx author) against this year at MySQL UC, and he said while they haven't solved the issue I had with it yet, they are actively working on it still. I'm not holding my breath - Lucene works fine for now, on a redundant pair of machines - but partitioning the index better wouldn't hurt (something that Sphinx does already natively support).

Reply | Parent | Thread

Patrick Galbraith

(no subject)

from: capttofu
date: Aug. 14th, 2008 01:02 pm (UTC)
Link

Slash screw? I suppose I was a rusty screw at Slash ;)

Sphinx: you can have "updated" indexes using distributed indexes and stagger the delta index updates a minute across all boxes, giving you something as close to update-able as possible.

Sphinx is fast, and I really am happy with it. There are some things to learn about it that are causing a lot of grief right now at my job, but I'll figure them out.

I'm going to put up a HOWTO on my blog any day now. Really ;)

Reply | Thread

awfief

(no subject)

from: awfief
date: Aug. 14th, 2008 06:25 pm (UTC)
Link

Actually, I think the bigger win seems to be the database plugin. From what I understand from your webinar (hai!!!), the sphinx db plugin (which you don't use to be fair) it acts like a text lookup table, with the fields (primary key, text)...so it seems to be less disruptive and perhaps only queries, not application code, need to be changed.

is that right?

Reply | Parent | Thread

Patrick Galbraith

(no subject)

from: capttofu
date: Aug. 14th, 2008 07:56 pm (UTC)
Link

Yes, it does allow no app code changes - and is quite cool! I would have used it if it were not for some internal resistance, mainly to having to deal with others having to patch 5.0 and compile it in for their dev boxes

Here's my post where I show usage of the storage engine:

http://capttofu.livejournal.com/9822.html

Reply | Parent | Thread

Jamie McCarthy

the Slash screw

from: jamiemccarthy
date: Aug. 14th, 2008 01:03 pm (UTC)
Link

"one of our slash screw"?

Maybe things were very different back in '99 but... we don't do that here anymore.

Reply | Thread

Brian "Krow" Aker

Re: the Slash screw

from: krow
date: Aug. 14th, 2008 06:38 pm (UTC)
Link

Heh... my ability to spell out the wrong word is amazingly awesome.

Reply | Parent | Thread