Brad Wardell's views about technology, politics, religion, world affairs, and all sorts of politically incorrect topics.
JU getting better
Published on March 14, 2006 By Draginol In Blogging

Our achilles heel.  Stardock's server farm has a lot of redundancy. But it has a bottle neck that until the last few months wasn't really a bottleneck.

It's a server called Marvin.  Marvin is our Stardock.net account server.  It traditionally hasn't been that busy, but as more and more sites start using it, it's gotten bad. Very bad.

The long term solution is to cluster it and use replication to mirror accounts onto other machines to load balance. But that's going to take time.

The short term solution is to get a mongo server in there to do the work. And that mongo server has arrived. It's got 6 hard drives in it to do both RAID 0 and RAID 1.  It's got SMP Duo Core 64-bit processors.  It's got (I think) 10 Gigabytes of RAM in it.

Once we have that up, the sites will be faster and definitely more reliable. Then we can work on the long term solution which is via replication and mirroring.

Update: The relief has arrived! No more unscheduled down time anticipated.


Comments (Page 1)
on Mar 14, 2006
Whatever you can do is appreciated. As an aside, have we ever had a JU Mirc channel on #stardock?
on Mar 14, 2006

We should set that up.

A #JoeUser on irc.stardock.com.

Actually, just looked. There is a #joeuser already on irc.stardock.com

 

on Mar 14, 2006
Really? I'll look again, thanks.


Ah, I just looked at the list on the welcome statement when you join stardock's server. Might wanna add it.
on Mar 14, 2006
Hey, it's back up. Cool. Thanks.
on Mar 14, 2006
Thanks for the news. Best of luck with the new servers as well.
on Mar 15, 2006
Best of luck on the new servers for sure, and please tell us techies how things go during the transition to clustering, NLB (network load balancing) and such. It'll be interesting to read about how things progress during the process
on Mar 15, 2006

And here I thought the reason for the down time was Trudy's new picture.

But an interesting aside.  Many wonder why we need to go to 64bit processing, when very few 32bit processors are maxed out.  The Answer is simple.  Memory!  Yep, memory is going to drive 64 bit processing down to the desktop in the future.  You just cannot have too much memory.

on Mar 15, 2006
Baker - Yeah, Danny and I setup the JU channel on the irc server several months back. Here's the article on it
http://zoomba.joeuser.com/index.asp?AID=73899

Feel free to stop on by whenever JU is acting up
on Mar 15, 2006
Great! Thx for the update, although to a "lay"person like me you're just talking gibberish!! But thanks!
on Mar 15, 2006
Hey brad?......................... thanx for caring so much about JU......... spending so much time keeping it up, having your crew spend time doing same.


Many thanks to you and yours,


elie

MM
on Mar 15, 2006
Sounds like a (really really) nice server, but...

RAID0 and 1? Why not go with a RAID5 solution with a hot swap backup? You get the best of both worlds in redundancy and performance and it damn sure has the drive capacity.

- B
on Mar 15, 2006
And here I thought the reason for the down time was Trudy's new picture


OH, I heard that one!
hehehehe, thanks Doc, awwwwww, now I'm blushin......or should I be angry?? hmmmm
heheh, havin an emotional meltdown here...

Thanks to the boss man for the site and good luck with all the new stuff
like forever said it's greek to me too!
on Mar 15, 2006
Yeah, happy new server day
on Mar 15, 2006

OH, I heard that one!
hehehehe, thanks Doc, awwwwww, now I'm blushin......or should I be angry?? hmmmm
heheh, havin an emotional meltdown here...

I was going to blog on the new picture, but after the trouble they were having, I figured another link would only add to it!

And that is the reason I have not commented on it.  Looking very good!  2thumbsup!  Yes!  The bionic lady is here to stay!

on Mar 15, 2006
It's 2 sets of RAID 1 mirrors that are then striped to improve performance while maintaining fault-tolerance (sometimes called RAID 10). And there's a hotspare in there ready to go. Short of some mainframes from years ago that I had limited access to, the new machine is the most powerful (raw specs) I've ever used.

Clustering is liable to have to wait until later this year. Though we are round-robining some of the websites and that will increase with new machines.