Virtualization and Grid Computing

Emerging Filesystems is a LinuxWorld article that Chris Samuel just put online (see his blog entry about the article).

It has a lot of great overview information, benchmarks, and interesting insight into ChunkFS, NILFS, btrfs, ext4, Reiser4, and ZFS.

I’ve been interested in how people are tackling the fsck problem which is one of the things he discusses. In my mind, it is already pain enough regardless of where it’s heading. The problem is also relevant to virtualization and grid computing. If you’re persisting and reusing VMs (not using the copy+throwaway model), the filesystems inside the VM are going to fsck eventually. Perhaps you have some timeout code in the meta-client layer that will only tolerate a certain wait for the VM to boot and report on its status somehow (perhaps using workspace contextualization technology). Because of the possible fsck delay you wouldn’t want to set this timeout very low (where it’s useful) unless you disabled filesystem checking which is not a great idea. A fsck friendly filesystem might become a popular VM choice? I’ll surely use a stable one for my local VMs and laptop.

The Ceph Distributed Network File System is an interesting, new project announced on the lkml

“Ceph is a distributed network file system designed to provide excellent performance, reliability, and scalability with POSIX semantics.”

(Note the comments at the kerneltrap link, there is an update on the GPFS part of the announcement)

The Apache Synapse team is pleased to announce the project is releasing 1.1 of the Open Source Enterprise Service Bus (ESB).

Apache Synapse offers a wide range of connectivity, including HTTP/S, JMS, (S)FTP, SOAP, WS-ReliableMessaging and WS-Security. The project has been designed to allow administrators to support advanced functionality such as load-balancing, throttling, and failover without writing code.

The new Apache Synapse 1.1 release includes many enhancements based on feedback from users, including

• Apache VFS based file transport - supports File System, FTP, SFTP, JAR, ZIP, TAR, GZIP, Mime
• Scheduled Task support makes it simple to run repetitive tasks
• XQuery mediator - simplifies XML transformation with the XQuery standard
• POJO Command mediator - allows the creation of message-independent mediation logic
• DB Report and DB Lookup mediators - support message augmentation and database logging
• Cache and Throttle mediators/enhancements - improve performance and manage load on existing services
• Split/Clone/Aggregate mediators - support batch processing of large messages
• Improved logging and tracing support

With asynchronous support for HTTP and HTTPS, Apache Synapse has very high scalability under load, supporting thousands of concurrent connections even with limited threads. Combined with streaming XML support, Apache Synapse has class-leading performance results.

The Apache Synapse code and binaries are available from the website at http://ws.apache.org/synapse

Thanks
The Apache Synapse Team

I’m looking forward to reading more about Blue Gene/P (coming to Argonne!). Excerpt from the press release:

ARMONK, NY - 26 Jun 2007: IBM (NYSE: IBM) today announced Blue Gene/P, the second generation of the world’s most powerful supercomputer. Blue Gene/P nearly triples the performance of its predecessor, Blue Gene/L — currently the world’s fastest computer — while remaining the most energy-efficient and space-saving computing package ever built.

The IBM® System Blue Gene®/P Solution scales to operate continuously at speeds exceeding one “petaflop” — or one-quadrillion operations per second. The system is 100,000 times more powerful than a home PC and can process more operations in one second than the combined power of a stack of laptop computers nearly 1.5 miles high. The Blue Gene/P supercomputer can be configured to reach speeds in excess of three petaflops, a performance level that many thought unattainable only a few short years ago.

I will be too busy in the coming month to post much here. I will keep adding sites to the gridindex.org search engines and updating the event list. In all likelihood will keep making brief news item entries. Enjoy your summer (or winter)!

From Quantum Scoop: The Holy Grail of particle physics may already have been found which was linked on Slashdot yesterday.

The current rumor, which comes in time for the summer conference circuit, may be different. It claims an experiment at the Tevatron has found a peak twice as high as the previous rumors’ bumps. And unlike the other rumors, this one includes details: the new particle’s mass, for instance, which fits within theoretical bounds on the standard model Higgs. Some versions include a decay chain, which describes what the new particle turned into as the experiment progressed, and which may be consistent with the standard model’s predictions.

Here is an interesting video of a talk given by Ari Zilka (CTO and founder of Terracotta Technologies) to employees of Google. Terracotta was recently open sourced and is receiving good reviews and attention. I am excited to experiment with it.

• Direct link to Cluster your JVM to Simplify Application Architecture.
• Online introduction by Ari Zilka, the speaker.

Now that Java 6 has been released, people are starting to talk more about the future of the language. I noticed Kirk Pepperdine’s comments in Java 7.0 language feature considered harmful:

“I’ve just watched the Danny Corward presentation at the Prague JUG on what could be coming in the JSE 7.0. In that presentation he demonstrated the “->” operator. The purpose of the operator is to provide a short hand notatino for properties. The syntax that we would normally use is;

a.setFoo( b->getFoo());

With the “->” operator we would see;

a->foo = b->foo;

How we’d currrently coding this is;

a.foo = b.foo where foo would have to be more visiable than private.”

He goes on to comment on how this violates OO data encapsulation principles and I agree with his arguments about these principles themselves, but I don’t see what the problem with syntactic sugar for very standard method calls is (I am assuming this proposed syntax is, as he suggests, just syntactic sugar for getFoo() and setFoo(x) and that the resulting bytecode is the same). This is just a compiler facilitating something already possible, like using:

String[] x = {"one", "two"};

instead of:

String[] x = new String[2];
x[0] = "one";
x[1] = "two";

As someone who likes Python a lot, it’s a natural thing for me to see “a.foo = x” and understand that a set method could have been executed. You can use “__getattr__” and “__setattr__” to override member access/mutation via instance.property syntax. This can be employed to do useful things behind the scenes while maintaining simple, legible code (such as in SQLObject).

If only it weren’t far too late for Java to eliminate the ability to do direct instance.property access altogether and have that mean get/set instead :-).