+ Reply to Thread
Page 3 of 3 FirstFirst 123
Results 21 to 30 of 30

Thread: Server upgrade 04/08 in the AM

  1. #21
    The race is back! John's Avatar
    Join Date
    Sep 2002
    Location
    On the mat
    Age
    44
    Posts
    40,432
    The site will be going down shortly for the RAM upgrade. It should be down 10-15 minutes, and be back up in time for AI and Bachelor.

  2. #22
    The race is back! John's Avatar
    Join Date
    Sep 2002
    Location
    On the mat
    Age
    44
    Posts
    40,432
    The RAM upgrade was not done last night, but will instead be done this morning. Sorry for any lag you may have experienced last night.

  3. #23
    The race is back! John's Avatar
    Join Date
    Sep 2002
    Location
    On the mat
    Age
    44
    Posts
    40,432
    Live from NYC, here's last night's bandwidth graph - I thought you might find it interesting.

    [IMG]images/img/bandwidth_1.jpg[/IMG]

  4. #24
    Soccer Kicks Balls cali's Avatar
    Join Date
    Sep 2002
    Location
    right behind you
    Age
    47
    Posts
    15,258
    That's purdy John.


    Um, I guess the fact that the green drops so much towards the end is good?

    **total geek wannabe, but still clueless**
    "Rice is great when you're hungry and want 2,000 of something' -- Mitch Hedberg

  5. #25
    The race is back! John's Avatar
    Join Date
    Sep 2002
    Location
    On the mat
    Age
    44
    Posts
    40,432
    Well, that's the middle of the night. The graph runs from 7am yesterday to 7am today. You can see how we're pretty busy during the day, then things slow down a little around dinner time, then spike back up during prime time to our highest levels (also the times when the site is lagging), and then drop off as people go to bed.

  6. #26
    The race is back! John's Avatar
    Join Date
    Sep 2002
    Location
    On the mat
    Age
    44
    Posts
    40,432
    As you may have noticed, the site was just down. That was the memory upgrade. Tonight's Apprentice finale, coupled with the AI3 elimination, should push the upgraded server to new limits.

  7. #27
    FORT Fogey eldee's Avatar
    Join Date
    Mar 2003
    Posts
    7,545
    Did someone just trip over the plug just now? J/K!

  8. #28
    Princess
    Join Date
    Oct 2002
    Location
    Wonderland
    Posts
    2,777
    "Currently Active Users: 719 (195 members and 524 guests) "
    And still fast!

  9. #29
    Wonky snarkmistress Lucy's Avatar
    Join Date
    Aug 2003
    Location
    Playing kickball for the beer
    Age
    38
    Posts
    8,870
    Cool. Thanks, John.
    It's such a fine line between stupid, and clever. -- David St. Hubbins

  10. #30
    FORT Fan Live from NYC's Avatar
    Join Date
    Mar 2004
    Posts
    221
    Quote Originally Posted by John
    Live from NYC, here's last night's bandwidth graph - I thought you might find it interesting.
    I just remembered to check back on this thread after posting some stuff on the Apprentice Forum tonight.

    That is a good looking bandwidth graph. Interesting that FORT is getting the peak around 9 PM - that must be when the shows are on. Ours (e-commerce stuff) had three load peaks at roughly 8 AM, noon, and 6-8 PM. People logged on when they got to work, spent time on the company computer over lunch, and logged on again when they left work and/or got home, it seemed.

    We had a 4 AM dip too. Ours went nearly to zero between 3-4 AM.

    5 Mbps peak should be no challenge for 100 Mbps full duplex Ethernet in the data center cage [hopefully you are not on 10 Mbps Ethernet ]. As for the disk I/O...

    5 Mbps / 8 = 625 KBytes/sec outbound

    I'll take back what I said about moving to RAID 0. Doesn't look at all necessary, at least for speed, you may still need to for size at some point. Seems like most modern IDE drives will do around 30-40 MBytes/sec burst or about half that continuous. Still way more than than the current outbound + inbound network requirement.

    Simply going to a better disk/controller interface protocol - SCSI or the new SATA-IDE - would probably do a lot more for database throughput than RAID, I would think. The better protocols seems to handle lots of small inbound/outbound data requests from a database better than the standard parallel IDE protocols. The old parallel IDE disk protocol is known to be pretty clunky for multiple small requests. Lots of overhead.

    I know from some past experience that a modern OS (WIN 2000 server, WIN XP 2003 server, WINNT server, and Solaris in my case) with dual-channel SCSI controller (always Adpatec for me) on a dual-processor box will do a pretty good job about allowing near-simultaneous dual access (one processor talking to one disk controller channel and the other processor talking to the other disk controller channel) by very effeciently queueing up disk I/O requests over the PCI bus. I'll be curious to see how the traditional parallel IDE controller does. I'll bet there will still be a pretty good similar-sort-of dual access speed up with that second IDE disk and the database migration.

    The new memory really seems to have helped! The site screams now. I just read the thread about the new disk being installed. I'll be curious to see what the database migration does for things.

    EDIT: Good luck!!!!

+ Reply to Thread
Page 3 of 3 FirstFirst 123

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts

SEO by vBSEO 3.6.0 ©2011, Crawlability, Inc.