Free dc stats gone crazy?

[Desslok]

They are reporting that I got close to 90k today(the 20th). That is 4x my average.

 

[Rudy Toody]

They are reporting that I got close to 90k today(the 20th). That is 4x my average.

The server lost an SSD a few days ago. What you are seeing is the credit that accumulated on the projects while the server was being repaired.

 

[Spacehead]

Stat were back up for a bit, but are down again now. Might not be weekly stats again this week

Here's the thread at Free-DC if you want to follow the progress

 

[biodoc]

that's a painful thread to read. I feel bad for bok..

 

[Rudy Toody]

that's a painful thread to read. I feel bad for bok..

Should we do our October sponsorship donation drive early to help him out. Like we did last year?

Hopefully, we can find a donor button somewhere in that mess.

 

[Peter Trend]

Should we do our October sponsorship donation drive early to help him out. Like we did last year?

Hopefully, we can find a donor button somewhere in that mess.

I think that's a great idea! Sounds like they need us

 

[petrusbroder]

I third this idea!

 

[Rudy Toody]

New thread for donations.

I emailed Bok to get the link to the donor button and he should get back to us when he has time. His email is somewhere on the server, so I hope that part is still working. Nothing has bounced back to me.

 

[Fullmetal Chocobo]

The server lost an SSD a few days ago. What you are seeing is the credit that accumulated on the projects while the server was being repaired.

Server with a SSD sounds scary enough. He must have needed crazy access times for the server to require SSDs.

 

[Bok7575]

Server with a SSD sounds scary enough. He must have needed crazy access times for the server to require SSDs.

It's not really the access time as I cache a lot of the queries/pages anyway. The SSD's are a must for the replication though. I'm pushing 10's of Gb's of data across drives many times a day. With Conventional drives it slows the whole process down and totally delays the deployment of updated data to the web facing database.

In it's new home the drive that gets written to is now a full SATAIII with burst write speed of 550Mb/s

Now if only I could put the whole database into memory! I could probably just about squeeze it into 64Gb but more likely I'd need 128Gb. Probably not too far off from doing that. I am going to be upping the Ram to 32Gb soon and want to get two more SSD's. May start replicating over to a database on the Webserver to help with failovers too.

 

[somethingsketchy]

It's not really the access time as I cache a lot of the queries/pages anyway. The SSD's are a must for the replication though. I'm pushing 10's of Gb's of data across drives many times a day. With Conventional drives it slows the whole process down and totally delays the deployment of updated data to the web facing database.

In it's new home the drive that gets written to is now a full SATAIII with burst write speed of 550Mb/s

Any thoughts about going with a RAID 10 array for the performance/redundancy or are you limited to using a single SSD for the DB replication?

 

[Bok7575]

Any thoughts about going with a RAID 10 array for the performance/redundancy or are you limited to using a single SSD for the DB replication?

Love to but it's cost prohibitive when you are talking about two arrays, but that's one of the reasons I'm considering replicating over to a database physically held on the webserver instead. I don't use mysql replication as such either, I wrote my own and copy the raw files (different table names), then once done drop and rename the actual tables. This way you never see a pause or inconsistency from the web side. Probably just use RAID 1 though.

 

[somethingsketchy]

Just read through the discussion page for the downed DB and had some thoughts...

http://www.free-dc.org/forum/showthread.php?32267-Stats-are-down/page2

...I was thinking what Amazon offers for EC2, however I don't see anything they offer that would work for your stats DB. While there would be enough memory (68.4GB - High-Memory Quadruple Extra Large Instance), at $2.00/hr that would get expensive quick.

Just out of curiosity, how often do you download, parse and update for the stats? I'm wondering if you decreased the frequency, if that would ease up on the DB replication or have the opposite effect (wearing down the SSD at a faster rate).

EDIT: Side thought, what about a hybrid drive? I (at my job) have been testing Hybrid drives with the developers for SQL Server/Visual Studio work and we've noticed comparable performance to a standalone SSD. They appear to work very well, but then again, we ain't running DBs 24/7 on the drives.

 

[Bok7575]

As often as I can. If I'm not getting the stats out there as soon as possible after they are published then I ain't doing my intended service

Honestly, as discussed, I don't think I'm wearing down the drives. It's the SATA Controller that went bad in the last box. The current box has a pretty good ASUS board with 4 SATAIII connections and a few more SATAII ones. Should be good for a while I think.

I'll see if I can build a second webserver and start replicating to it once I order some extra hardware.. Got some parts for it already, an Intel board and a core i5.