This accidentally fell out of her pocket when I bumped into her. Took me four goes.

Sunday, April 05

Geek

The Internet Has Everything

Even this.

(via)

Posted by: Pixy Misa at 06:21 AM | No Comments | Add Comment | Trackbacks (Suck)
Post contains 7 words, total size 1 kb.

Geek

While We're Waiting

[root@akane ~]# hdparm -t /dev/sda /dev/sdb

/dev/sda:
 Timing buffered disk reads:  628 MB in  3.00 seconds = 209.27 MB/sec

/dev/sdb:
 Timing buffered disk reads:  786 MB in  3.00 seconds = 261.77 MB/sec

sda is a 3-disk RAID-5 volume.  sdb is the Intel X25-E SSD.

Vroom!

Posted by: Pixy Misa at 01:04 AM | No Comments | Add Comment | Trackbacks (Suck)
Post contains 47 words, total size 1 kb.

Saturday, April 04

Geek

I Have Bad News, And Expensive News

Installed OpenVZ on Akane (installing OpenVZ is a piece of cake). 

Rebooted.

No server.

This is not good.

It could be a whole lot worse, since there's nothing on the server, no files, no users.  Still, I can't reach it at all.

One of the things I love about SoftLayer is that IPMI - remote monitoring and KVM - is standard on all servers.  Many other providers are heading that way now, but SoftLayer was one of the first.

So I log in via IPMI - which is good in itself, because it means the server is up.  And I poke around a bit, and find that I have no network cards.  Funny, they were around here a minute ago.

So I restart the networking subsystem and am informed:
Bringing up interface eth0: igb device eth0 does not seem to be present, delaying initialization   [FAILED]
That's not good.  The fact that this message does not appear anywhere on the internet is extra not good with cream.

Some more digging revealed the fact that the network chips on Akane's motherboard - the Intel 82576 - was first supported in the 2.6.25 kernel.  OpenVZ runs 2.6.18.  So do CentOS 5.3 and RedHat Enterprise 5, but someone there must have backported the driver.

I downloaded the OpenVZ 2.6.26 kernel - a development version, not recommended for production - and that was, if anything, worse.  Definitely no network, anyway.

So it looks like we're getting Virtuozzo after all.  Which will cost me extra, but...  Well, mostly it will just cost me extra.  It has some nice features, like a user control panel so you can look after your own virtual machine.  I don't have any users who actually need that...  Well, I'm sure I could find some, but it's not what I'm here for.

Anyway, I've sent an email off to sales to get the move to Virtuozzo-land under way - which will involve wiping Akane clean and reinstalling her again, probably at least twice - and then I can get things under way once more.

Posted by: Pixy Misa at 11:46 PM | Comments (4) | Add Comment | Trackbacks (Suck)
Post contains 347 words, total size 2 kb.

Geek

Reboot, Reinstall... CentOS 5.3

Just reinstalling the operating system on Akane from scratch.  I always do this at SoftLayer because (a) it's easy, (b) it's a good test; if I can't do a clean reinstall, there's something wrong, and (c) I can make sure that it's installed the way I want.

Then on with the OpenVZ.

Then I order a bunch of VLAN-routable IPs.

Then I start creating VEs, and installing CPanel, and so on.

Then I test-copy the whole of Aoi / mee.nu across.  Official migration probably over Easter, but I'll do a quick test first.

Then migrate people off Sakura.

Then off Midori.

Then back to migrating people off Movable Type and WordPress to Minx.

Then upgrade to Minx 1.2.

Then Pixy goes to happy land.

Update: Well, that didn't work. sad

Posted by: Pixy Misa at 08:35 PM | No Comments | Add Comment | Trackbacks (Suck)
Post contains 134 words, total size 1 kb.

Geek

New Toys Indeed!

Look what I've got!

top - 22:12:36 up 14 min,  1 user,  load average: 0.01, 0.02, 0.01
Tasks: 280 total,   1 running, 279 sleeping,   0 stopped,   0 zombie
Cpu0  :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu1  :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu2  :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu3  :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu4  :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu5  :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu6  :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu7  :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu8  :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu9  :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu10 :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu11 :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu12 :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu13 :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu14 :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Cpu15 :  0.0%us,  0.0%sy,  0.0%ni,100.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Mem:  24677364k total,   374464k used, 24302900k free,    19348k buffers
Swap:  1052248k total,        0k used,  1052248k free,   187340k cached

Posted by: Pixy Misa at 01:13 PM | No Comments | Add Comment | Trackbacks (Suck)
Post contains 193 words, total size 2 kb.

Geek

Um, Waitaminute

SoftLayer sales got back to me, with a quote that was more than what I could get just by clicking around their website.

Oh well, guess I'll just click around their website then.

Sleep first, tho'.

Update: Oh, they quoted me on the 2.66GHz 5550, not the 2.26GHz 5520.  There's about $1200 difference in the RRP; no wonder the price was high.

Posted by: Pixy Misa at 03:47 AM | No Comments | Add Comment | Trackbacks (Suck)
Post contains 64 words, total size 1 kb.

Geek

Also Also Wikout

Aaaand - if I use Xapian - which I shall - the Minx database need not support full-text indexing.  So for instance PostgreSQL, which I rejected because its full-text indexing is hideous, becomes viable again.

Actually, PostgreSQL's full-text indexing is no more arduous or inelegant to implement than Xapian's, and perhaps nearly as flexible.  The difference is that Xapian lives off to one side of your database, doing search-y stuff and nothing else, and doesn't clog up your tables or your SQL.  The PostgreSQL full-text search, by working exactly the same way, turns the virtue of integration into a vice.  The pre-computed search terms field that you need for the full-text index is present in both, but in Xapian it's out of the way and doesn't bother you.

And that really does matter.

Posted by: Pixy Misa at 12:55 AM | No Comments | Add Comment | Trackbacks (Suck)
Post contains 136 words, total size 1 kb.

Geek

Awards And Stuffs

The coveted (and rarely-given) Doesn't Suck Award today goes to the Xapian search engine.  I've been testing it out for my day job, and it's quite impressive.  I'll definitely be using it in a future version of Minx - our current search function works by brute force, which works surprisingly well, but will become a problem sooner or later.

The pre-release versions of Xapian feature not just full text and tag-based indexing, but geospatial and image indexing too.  And when I say image indexing, I mean "find me all the images that look similar to this one".  It does - as far as I can tell - a search in 49 dimensions to pull this trick off.  Seriously.

Posted by: Pixy Misa at 12:37 AM | No Comments | Add Comment | Trackbacks (Suck)
Post contains 121 words, total size 1 kb.

Geek

Ultra Plus, Ne?

By the way, all those SpecIntThingy 2006 numbers in the previous post?

They're normalised to the Sun Ultra Enterprise 2.

I used to work on one of those, albeit the 200MHz version rather than the 300MHz (actually 296MHz) version used as the Spec '06 baseline.  I even have an Ultra 5 in my closet.  Right next to the SGI O2...

For the baseline tests, they only used one CPU.  Even so, this new server is equal, for the purposes of general data crunching, to one hundred dual-processor Sun Ultra Enterprise 2's.

Which was 1997 hardware, sure, but they used the 2005 compilers and OS for the baseline testing, so the software isn't that far removed from today.

We have 4x the number of cores, and ~8x the clock speed, so there's a factor of 3 improvement in efficiency.  That's not too shabby; the Ultra 2 was a pretty solid machine in its day.

Posted by: Pixy Misa at 12:24 AM | No Comments | Add Comment | Trackbacks (Suck)
Post contains 156 words, total size 1 kb.

Geek

Attack Of The Killer Akanes

Still waiting to hear back from SoftLayer sales - apparently they're swamped with orders for the new servers.

Meanwhile, doing a bit of poking around.  The SpecInt 2006 score for the Xeon 5520 - the CPUs we'll have in the new server - is 27.9.  That's in a SuperMicro system, exactly what we'll be using.  The score for the Core i7 920 I was using for comparison is 28.6, so we won't be far short of that.  The score for our current Xeon 3220 is 20.3.  No scores yet for the Xeon 3520, which is what I'd planned to upgrade to, but it should be similar to the 920, because it's the exact same chip.

So I'm pleased and a little surprised by that.

Floating point is 32.0 for the 5520 vs. 19.0 for the 3220.  And 29.3 for the 920...  Odd.  Maybe a new compiler revision?  Too lazy right now to read the detailed test reports.

For SpecIntRate (the multi-threaded version), the score is 202 for the 5520 vs. 59.2 for the 3220 and 109 for the 920.

And finally SpecFPRate is 160 for the 5520, 42.2 for the 3220, and 78.9 for the 920.

So the new server will be three times as fast as the old servers (but it will replace two and a half servers), and have three times the memory (and will replace three servers worth of memory).  It will be cheaper, easier to manage, more robust, and more expandable - I can take it to 48GB of memory if needed, where all three of the current servers are full.

No room for extra disks, though, unless I go for a 2U model, but that costs...  Huh.  That's not that much more expensive at all, and comes with redundant power supplies, and 18 DIMM slots and 12 drive bays (instead of 12 and 4).  That's definitely worth considering.

Intel will be releasing 6-core versions of the new Xeons in the next 12 months, so that will get us up to 24 virtual cores, 72GB of memory, and 12 disks.

That should do.  Yes, I think that should do. wink

Posted by: Pixy Misa at 12:04 AM | No Comments | Add Comment | Trackbacks (Suck)
Post contains 359 words, total size 2 kb.

<< Page 2 of 3 >>
63kb generated in CPU 0.0173, elapsed 0.2662 seconds.
51 queries taking 0.2473 seconds, 349 records returned.
Powered by Minx 1.1.6c-pink.
Using https / https://ai.mee.nu / 347