Imagine you could snapshot yourself and generate a clone from that snapshot. On the surface, that would seem like a cool idea, but scratch the surface and you quickly run into problems. These problems could be ethical (what rights does your clone have to your own possessions?), or they could be practical (how can the planet sustain the population explosion?). Fortunately, with zfs datasets, snapshotting and cloning not only are permitted, they are actively encouraged.

Over at the Xerces Tech site, a recent article outlines how to use zfs clones in order to safely do an apt-get on their Nexenta box. Take a look at the article called Unbreakable upgrades, ZFS and apt-get to see how it's done.

Have you ever wanted to try building your own ZFS-based file server? Jermaine Maree has done just that, and blogged about it. Start with part 1.

Even a suboptimal configuration can result in a performance boost. The most interesting thing, I think, is the ease with which the zpool was created.

I wonder what kind of performance numbers this user would see with the 7110 compared with the Dell Powervault. The 7110 can hold 14 146gb sas drives, whereas the Dell uses 14 146gb scsi drives, so comparing power utilization would be interesting as well.

One of the things I'm trying to do, besides have a full time job at Sun and raise three kids (another full-time job, even with my wife's help), is to learn Japanese. I'm not in any great rush, so I make time where I can and I try to spend at least a few minutes a day on it. I've started by trying to learn the kana writing system before I move onto the more complex kanji. There are two primary kinds of kana, hiragana and katakana. The former is used when writing Japanese, and the latter is used when writing "foreign" words, including words that have become part of the Japanese language but were borrowed from, for example, English.

Sun, as you're probably aware, encourages blogging from its employees around the globe, and more than a few of these are from Japan. Here's an example of one: キャンパス アンバサダ (I hope that came through).

Here's how the entry looks in my browser - I'll include it here in case your browser doesn't show the characters correctly

What I've been doing, in order to practice my kana, is to pick out the katakana symbols from the entries and see if I can work out what word it is in English. Here's the title of the entry I linked to, broken down kana by kana:

Spelling it phonetically, we get kyanpasu or kyampasu for the first word, and anbasada or ambasada for the second. And I was puzzled. Contrast that with another katakana word that appears in the blog entry: "オリエンテーション", which is o-ri-e-n-teh-sho-n ... orientehshon ... orientation. Obvious, right? What's kyanpasu? Words that end with ス tend to have the final 'oo' sound dropped, so it becomes kyanpas when saying it aloud. I give up, so Google Translate tells me: Campus. This is where I hit my head on the desk. Why does Campus start with キャ? The a in kya sounds like the a in father, not the a in campus, so maybe that's why kya is used to differentiate it from the カ character ('ka') which also sounds like the a in father?

Of course, once I saw the first word was 'campus', it's easy to figure out the second word is Ambassador. Which I would have spelled, apparently incorrectly, アンバサドル

When we think of a cache, we think of a way of storing information "closer" to the place it's needed. Most general-purpose CPUs, for example, have an on-board cache which is used to avoid accessing main memory - after all, the memory that's on the same die as the CPU is going to be quicker to access than the RAM chips. Filesystems use caches of RAM to make disk accesses appear to be quicker, as RAM chips are much faster than moving a mechanical arm across a spinning disks. If we're lucky, and if we've got a good caching algorithm, we can get an impressive speed boost by keeping the right bits in RAM. Likewise, CPUs get a speed boost by keeping the right instructions and data on chip.

Caching is not limited to CPUs and filesystems, of course. Most browsers maintain a cache of pages, of images, of css files, of javascript, and of any other bit of information that is useful for displaying web pages. By using a local on-disk cache (some of which is going to be in RAM anyway, thanks to the filesystem), browsing appears to be much quicker than it would if the browser had to re-load every single image from a distant web site. The browser does check to see if any files need to be retrieved again (see http's 304 Not Modified response code), so there is some over-the-wire activity, but that's about it.

All of this is a long-winded way of saying I was amused by the second bullet item here (from Apple Insider):

Either the cache is poorly implemented, or the users reporting this information are confused.

This is a great example of doing well AND doing good. Proof positive they're not mutually exclusive.

It's been over a year since I last posted sightings of ZFS around the web, so it's high time I offered another list.

• How It Works: Filesystems - David Chisnall, author of Xen hypervisor book looks at filesystems in general and talks briefly about one of the many facets of ZFS
• Snow Leopart to Support ZFS - Chris Tangora is one of many folks who noticed that Apple's upcoming update of OS X, Snow Leopard, will have ZFS read and write support built in.
• Creating near-enterprise level storage at home with ZFS - ephekt details how he created a home server using ZFS
• HPC Consortium: Storage - Josh Simons talks about the goings-on at the Sun HPC Consortium in Dresden, and summarises the talks that the Lustre folks gave. Lustre, recently acquired by Sun, is going to be using ZFS for the underlying storage in order to provide data integrity.

It was nearly a year ago that I first made this screenshot:

Since then, I have done quite a number of different things, all related in some way to getting zfs to install and boot. Some of these things also involved teaching Live Upgrade to understand zfs datsets.

But now I'm starting to see that screenshot elsewhere, virtually unchanged from that fateful day long ago when I used the original to help design the changes needed for the text based installer.

Here are some: The Sect of Rama | Number 9 | Otmanix' Blog | Osamu Sayama's Weblog

It's really exciting to see it get out there and for it to be used outside of the development and test teams. Of course, there are some CRs being filed, and there are some things we'll need to address, but it's great nevertheless.

It's too bad they weren't using a filesystem that could detect and automatically repair corruption. Ah well.

To all Windows-only admins:

Back in the olden days, I used DOS. I kinda got used to how scripting worked, and wrote many a .BAT file. Then I was shown Unix (an early System V version) and my eyes were opened. I understood what a real scripting language could do.

ONE OF US

One of us

One of us!