iNODE

Some time back I posted a few notes on using APC (Alternative PHP Cache) on a BSD-based system. The important idea then was to increase the size of the shared memory segment from the default 4MB.

Yesterday, I decided it was time to implement an APC cache on a Leopard-based OS X machine to improve php performance on that box as I added a new research portal to the server.

A few updates to the process for Leopard (10.5) servers:

1. Down latest version of APC code.
2. Use this command line to configure the code before issuing the make and make install commands (thanks to Pascal Opitz’s Content with Style for this tip):

MACOSX_DEPLOYMENT_TARGET=10.5 CFLAGS="-arch ppc -arch ppc64 -arch i386 -arch x86_64 -g -Os -pipe -no-cpp-precomp" CCFLAGS="-arch ppc -arch ppc64 -arch i386 -arch x86_64 -g -Os -pipe" CXXFLAGS="-arch ppc -arch ppc64 -arch i386 -arch x86_64 -g -Os -pipe" LDFLAGS="-arch ppc -arch ppc64 -arch i386 -arch x86_64 -bind_at_load" ./configure

You’ll also want to expand your shared memory segment on the server. Under 10.5, you create a file /etc/sysctl.conf and add the appropriate values. This example makes a 128Mb shared memory segment:

kern.sysv.shmmax=134217728
kern.sysv.shmmin=1
kern.sysv.shmmni=32
kern.sysv.shmseg=8
kern.sysv.shmall=32768

Not sure why I find this picture so reaffirming.

http://i39.tinypic.com/24w7ed0.jpg

Update (5/20/09) Now that I’ve started twittering, I guess this isn’t as funny as I thought….

@grotophorst

Spent an hour or so today participating in a WebEx presentation hosted by Sun, the DSpace Foundation and the Fedora Commons. Billed as “A Collaboration Update” this event basically served to offer an introduction, rationale and development path for DuraSpace (sounds familiar doesn’t it?).

Where to begin?   On some level, I guess the headline of today’s session was:

Trusted Repositories are So Yesterday.  Welcome to the Cloud

I use cloud storage today to store extra backup copies of some of our library websites and other data but I don’t think of it as an overly reliable solution.  At least, not in the sense that it’s the only place my data lives.   That sort of dependability isn’t the IT problem the cloud was created to solve.  If you doubt that, take a look at the Service Level Agreement Amazon offers for their S3 service.   It’s quite specific in some areas but vague when it comes to total data loss:

If availability is impacted by factors other than those used in our calculation of the Error Rate, we may issue a Service Credit considering such factors in our sole discretion.

Sorry, your data’s gone.  But we will give you a credit in case you want to give us some more data.   Comforting, isn’t it?

It’s possible that DuraSpace might eventually morph into something useful:

• maybe a LOCKSS-inspired dispersal of IR content across multiple cloud-based systems
• or a cloud-based aggregation of all the world’s repository content under a common set of APIs

but that doesn’t seem to be the current focus.   Instead, there was prominent mention made of “value-add” services that could be performed on the data coming from your site on its way to the cloud.   And I also heard an allusion to the fact that with DuraSpace those sites that have run out of storage or computer power will now have an option for their continued existence.     I didn’t find either  terribly convincing.    I can see that it might be nice to have my data “improved” on it’s way to the cloud but wouldn’t I then want to replicate those improvements on my local copy of the data?  And given today’s pricing, what site running DSpace can’t afford another terabyte drive?

While it may be buzzword-compliant,  I’m afraid at this point DuraSpace sounds like a “let’s try to monitize our involvement in this thing” solution searching for a problem.   Maybe I missed the point.  I hope so.   I also hope that the development of DuraSpace won’t steal too many cycles from those currently working on improving the core software assets of the DSpace and Fedora projects.

A few quick items to help bring things up to date:

Netbook Experiment

Had an idea the other day that instead of purchasing regular size notebooks for staff loaners (machines people use when attending a conference or making a presentation), I should see if a netbook would meet the need (costs and weighs much less). After reading nice things about the MSI Wind, I ordered a couple U-100 models (1.6GHz Atom processor, 1GB RAM, wireless b/g/n, and 160GB hard drive) for approximately $450 each.

Well another reason I went with the MSI Wind is this chart from Boing Boing:

The chart is quite correct. Putting OSX on the MSI Wind is almost trivial.

Federated Search Update

Working through a proposed contract from Deep Web Technologies. When signed and the service is underway, I’m hoping we can build “boutique” federated search engines for our various research portals.

Unlike MetaLib, 360 Search, and other federated search products, Deep Web prices their product with a matrix that includes: search pages (e.g., boxes where a user types something in) and sources (e.g., Something as small as an individual e-journal can count as a source or a system as large as ProQuest). This flexibility in configuration is exactly what we need for our application.

My goal is something like this: A librarian who manages a particular portal can go to the departmental faculty and say “let’s talk about the 9 or 10 resources you’d like for me to combine into a federated search service for you.” We then put that search box on the portal and good things follow:

• By working with the faculty to identify and build the search system, we give the librarian a useful “carrot” with which to attract faculty interaction.
• We end up delivering a product that our target audience (researchers) will actually use and value.
• It will help drive users to our portal(s) which gives the librarian an incentive to keep it fresh and informative.

More on this as it develops over the coming months.

If you’re not familiar with Deep Web’s search software, I’ll suggest you take a look at Biznar (a business topics search service) or Scitopia (a search portal for the libraries of twenty Science and Technology societies and more).   My ultimate goal is to figure out why Mason should send me out to visit the Deep Web Tech home office (in Santa Fe).

Fix it Till It Breaks [latest update in a continuing series]

The other day I received a copy of Drive Genius 2 from Other World Computing (buy a drive and it comes at a really reduced price), so I thought, “Hey, why not take my main machine that’s running really well and test out the Drive Genius defrag utility.”

Went to get some lunch while it worked and when I got back it was waiting for a restart command.    I didn’t expect to see a huge difference but you never know…

Hit restart…got the grey screen…then the spinning daisy and apple silhouette …then about 45 seconds in the multilingual kernel panic message began rolling down the screen.   Holding down the option key, I was able to attempt a boot off a Leopard install disk but even that ended in a kernel panic.  Who ever heard of a degrag killing a motherboard?

From a second machine I checked the serial number at Apple’s site and found the MacPro was still under warranty.    Took out my drives and add-on memory, put in a empty (new) boot drive and schlepped the big tuna (it must weigh 50+ pounds) over to the Apple Store in the local mall.   The “genius” took it in the back and returned about ten minutes later.

Turns out the machine was OK. It wouldn’t boot off the DVD I was using because I had upgraded the video from the stock ATI card to an Nvidia GeForce 8800GT.  My 10.5.0 install disk didn’t have a driver for that card (they appear on 10.5.4 disks or later).   Booting from a new 10.5.6 disk it came right up.   Brought it back to the office, inserted and  repartitioned my defragged trashed drive, put the memory back in place and restored from a week-old SuperDuper! image file.  Back in business.

Yes, I should have thought of the video driver problem.  Perhaps if Apple were to publish “Hardware release” numbers like Sun does with Solaris revs I would have thought about it.  Of course,  you don’t go with a Mac because of great enterprise-level support.