Must of really been worn out post UC because my experiments and blogging have dipped to all time lows… but rest assured I am on my way back. I will be following up posting some posts on my Sun/Intel SSD tests I did for my presentation at the UC. This will give those who did not go to the UC a chance to hear me babble on about SSD’s some more ( OH NO, not more babble about SSD’s! ).
Plus of course their is Waffle. Right now I am perplexed as ever by our mysterious index issue… and its been bothering me for weeks! I think I will post some details and look for external opinions very soon.
Pre-UC I put out a teaser on some dbt2 scores in the 50K range. I mentioned and showed the graphs during my SSD session, but I thought I would show them here for those who skipped the UC or did not attend my session. Basically what most people consider to be a classic “CPU Bound” workload where all of your data easily fits into memory can also see benefits from moving to SSD’s. Remember just because everything fits into memory doesn’t mean your not going to be doing some operations to disk ( logging, flushes, etc ). Take a look:
|Regular Disk BBU (5.1.33)
|SSD WO/Drive Cache (5.1.33)
|SSD W/Drive Cache (5.1.33)
|SSD WO/Drive Cache (5.4)
|SSD W/Drive Cache (5.4)
The other thing to note here is is the speed up in 5.4, that’s a pretty big performance improvement. Now of course with the drive cache enabled on the ssd’s you get a 10%+ boost in performance, but keep in mind its not 100% safe for your data unless you use some other mechanism to mitigate data loss on a power outage ( i.e. ups shuts down server ).
Because someone will ask:
Tests where done on a Sun 4450, 4 disks Intel Enterprise SSD, Raid 10, 16 core 2.93 GHZ cpu’s, adaptec controller with 256MB W BBU, 10K SAS drives. Optimal configuration for the SSD’s software raid, 64K Stripe size, Software raid, Disabled controller read & write cache.
Had a fun time being on Percona’s performance panel late last night. For those who were there, I was probably the only guy you did not recognize:) The panel of Ronald Bradford, Mark Callaghan, Cary Millsap, Jay Pipes, Michael “Monty” Widenius, and myself spent the evening answering questions from the audience and of course drinking Monty’s famous black vodka. While a lot of the questions drifted away from performance I think everyone had a good time and found the session to be informative. Now the question i have is : does anyone have any good photo’s of the event?
Well I just finished my SSD session. I was concerned by the amount of slides I had so I kept trimming them back. What happened? I finished early. Why? The problem with storage is it’s not really that sexy. I mean nobody ( sane anyways ) drools over drive specs ( I do consider myself insane by the way , and do drool over some drives ). CPU’s are sexy… memory… sexy … graphics cards sexy… drives… are not. We only had a small crowd turn out (50-60 people maybe), but they were vocal and interactive. I got some great feedback from others who love IO performance as much as me. In fact one antendee stopped me in the hall and thanked me , telling me I helped him make up his mind to purcahse some SSD drives, which made it worthwhile.
People always have conversations at the UC that spark ideas for new projects or features, this is my random collection of features or items I need to look into. Most people can ignore it… but if your interested in what goes on in my head here ya go:
- I started looking into this before, but I was reminded of it again today… it would be useful to setup separate innodb buffer pools like the key caches in MyISAM.
- maybe an easy hack to put in a tiered LRU based on file per table. What I mean by this is internally the internal BP in inno actually uses the space id ( file per table will have a new ID per table )… this should make it feasible to ( maybe easy ) to move certain spaceid’s up or down the internal LRU list This maybe an interesting feature, not exactly like pinning a table into the BP but close. Basically, if a table is accessed and its higher priority it will stay in the BP longer.
- Need to port waffle to mysql 5.4
- Was asked when Waffle will make it into mainstream Mysql/Innodb… ummm have never thought about it, need to ask around
- Maybe able to direct waffle to “cache” only certain spaceid’s fairly easily. This may be usful to eliminate pushing single or low use tables to memcached
- Fix how we are trying to use NoReply
- Tough one… waffle storing rows? Could be accessed outside of mysql?
- Persistent memcached? i.e. memcached and block device
- Enhance JuiceDB, especially looking at Drizzle
- Dbt2 & drizzle…
- Look at a better memcached benchmark
- Look at testing Virident appliances, What can they mean to waffle?
- Look at testing Schooner Appliance’s, What can they mean to waffle?
Sitting in Yves presentation on NBD performance tuning at the MySQL Camp. Your all missing a good one for those who want to know more about NDB Performance tuning.
Seems like its been a long day already. I started off talking to the MySQL Camp discussing some of the the silly things customer do and hire me to help fix. We started off a little rough ( projector and power issues ) but I think it went well after those were worked out. Sheeri should be posting the presentation soon.
Our Waffle Grid Presentation seemed to go pretty well, lots of people who seemed generally interested in the project. Had a number of people stop by and ask when we can get this into an official release:)
Fun stuff so far.
We have 3 presentations today and you may not have been thinking of attending! What’s wrong with you.
Oh… did you think I was going to say something else?
Stop by say Hi, seek us out after our sessions and introduce yourself. I would love for you to stop by give us your idea’s on Waffle, Bigdbahead, MySQL, benchmarks, Life, Oracle, Baseball, or anything else on your mind.
Remember the what’s printed on the front of the Guide… “Don’t Panic”.
First we all get hit hard with News like this… So we go through lots of stages. Panic, dear, acceptance, etc. They all come and everyone is left asking now what.
First… look around at and get reassurance from others…
Sheeri has a decent post on her take on the news here.
As I mentioned earlier, it’s way to early to speculate on what’s going to happen. Integration can not begin until all the regulatory stuff get handle, which could take months and months. So their is going to be a really long quiet period of time. For us internal folks this will be a scarey time of course. So anything you hear now is going to be pure Rumor more then likely.
Absolute worse case for everyone is that since MySQL is open source you get your MySQL fix from another company. This other company could take the form of a current company or fork (OurDelta, Percona, etc) , another open source project ( i.e. drizzle ), community support, the main distro’s ,or even a new company formed from others. Remember the code is available, anyone can fork it, anyone can patch it… so no matter what MySQL will live on.
So just sit back try and relax and wait. I know it’s hard, but this will all work out, and in the end whatever is left will be stronger for it.
Everyone who was up @ 7:30am seems to feel tired and subdued. Everyone I heard from is worried about what this means for MySQL. Some people have taken to extreme joking running around with cracks about the evil empire and “hail Larry”. It’s a very tense mood all in all. The feeling seems to be that the announcement was timed to have maximum effect, others seem to think it was timed so they can all bring us over to the assimilation machine and get our implants embedded in our brains. I am sure you are going to be hearing a lot of fear & uncertaintity, rumors and the like over the week but their is still a long way to go, and nothing will be worked out officially for months.
That being said it does feel like the ” The end of the world as we know it” … welcome to black monday.
Ummmmmfff…. wind out of my sails today. I believe their was an analyst who predicted this a year ago when Sun bought MySQL…. I wish I could find the article. I am going to head down to the UC floor and get a feel for what people are thinking… will post more as I get time.