Okay so I'm running a hyperthreaded processor, with two 10,000 rpm disks on a hardware controller RAID 1. Now I could have had 7200 rpm disks in my tummy, but Susan said she doesn't consider anything less than 10,000 rpms as “Server quality”. Which reminds me... I was kinda hanging out in the SBS newsgroup the other day and someone wanted to use a beefy desktop to run a SBS 2003 server for an office.
People. When you expect a piece of machinery to 'serve' you, you have to spec out hardware that is ready to serve.
I think I'm up for the task.
I can have all the good software in the world, but if Susan didn't go with reasonable stuff, I wouldn't be able to properly handle these blogs like I'm doing. So let's get down to talking about all this prodding and poking that folks have been doing to me to make sure I'm in good shape.
First off my excellent keeper Vlad Mazek, CTO of OwnWebNow, wonderous being, fabulous guy, extreme all around good guy [I'd say Vlad rocks but it's been seen that “rocks' is too much of a buzz word around here, so I'll refrain for now], noticed that my page file use kept ticking up a bit, which was the same symptom I had before at my old home that kept killing the blogs. So for now until we get up to Community Server we have a routine where the App pool is cycled. So for now we're going to let that handle that issue until we can get updated to the CS engine once the version 2.0 releases.
So remember the other night they were doing testing on me? Here's some of the comments I overheard... now I didn't catch everything...but I jotted down a few things.....
.....This data looks pretty good. smlogsvc is down to half a %....this is goodness...... Sqlservr - No surprise here W3wp - again, no major surprise. From an I/O perspective, I'm quite happy with things. You're averaging <50 IOPS (read+write), and on a 2 spindle 10K RPM RAID1 with a little cache, this is just fine. You definitely have headroom to spare.
Goodness... hey that's sounds pretty good. I got “goodness“. I must be performing okay. So I think I heard them say they are going to take that test on me again next week as with the holidays they didn't think I had a full load hitting me and then again if/when they start changing things. Some of the things I heard them say was .....
I'm interested to see another 6 hour snap after you make any of the following changes:
- SQL upgrade
- .Text -> CS
- gzip enablement
- conditional GET enablement
- TTL changes
- any other change that might affect the load profile of the server
It sounds like they want to do more testing on me. So I'll keep you guys posted if I overhear them say anything else of interest.
Oh yeah...I know...since I moved to Vlad's home town of Orlando I've been losing the accent... as sometimes these days I sound more like Vlad does in case you are wondering...
May the force be with you.....
So here's what they are doing to me...
First they went into Counter logs in Perfmon and set up a new log. Called it “LoadMeasure-QuickSnap” and doing a 5 minute monitor of what I'm doing with a 2 second interval for the following items.....
- ASP.Net application
- Logical Disk
- Physical Disk
- Web Service Cache
- Web Service
They just ran it on me...heh...that tickles.....tomorrow they run a 6 hour monitoring on me that will grab stuff every 30 seconds.
It's Yoda the Blog server again and well it's Sunday night and I'm settling into my new home here on the East Coast of the Country next to DisneyWorld. I have lots of nice cool people looking out after me and making sure I'm doing what I'm supposed to be doing. Right now we're keeping an eye out on me and setting some benchmarks.
That's a view of my Task Manager graph up above and you can see that I bounce around a bit. My Cousin Sam The SBS Server tends to be a bit less bouncy than I am but I think it's because I get to do a lot more fun cool stuff than Sam gets to do. He has to be pretty boring, whereas I get to show off out here on the Internet a bit more. You can see his graph down below.... it's pretty boring as you can see...not like mine at all.
Right now there are some perfmon's being taken of me and we're going to see where my bottlenecks are. I'll let you know what they find out.
Once again, thanks especially to Vlad Mazek of Vladville.com and SBSshow.com and CTO of OwnWebnow.com
P.S. I really have two CPU [hyperthreading you know] I just cropped the picture so you could see the graph a bit better, Sam has 4 CPUs.
Hi there! My name is Yoda. I'm soon to be the server underneath the MSMVPS.com blog site. Right now I'm getting ready along with the database that is migrating over to me.
One thing is for sure.... my owner is soooooo blonde.... and soooooo not a SQL server expert at all. Fortunately she called in some experts and got help.
Here's what he did.
First off he told her that the database was running in “Full mode“. As Nick said “you know that this creates 814Mb of transaction logs“... “Uh..no“.... she said.
Then he looked at the tables in the blog and both the referrers tables and the url tables had 4 million... M I L L I O N entries. Since this blog has had major issues lately we asked the blog community if it was okay to dump those logs... off they went. Man is the blog site peppier with just that.
Then he found out just how blonde my owner was... all this time she's been setting up blogs by hand with the SQL Enterprise manager... I tell ya does she need a SQL course or what? I'll be so glad when we get SQL 2005 on that SBS box she has in the R2 era ... I think the more GUI interface will force her into learning more about the stuff she has and doesn't use like she should be using.
Nick told her that there was no need for a full database, simple model will do for a web server. My owner just has to remember that she can only go to the last good backup [which the ISP does nightly anyway and she can just plan on pulling over a backup.]
We went from a backup file of 3.27 GBs...you read that right... almost 3.5 gigs to a shrunk down file of 323MB.
Pretty sad isn't it? Obviously we need to work on her SQL capabilities... she's got a long way to go.