Here's a collection of highlights, selected totally subjectively, from this week's HPC news stream as reported at insideHPC.com and HPCwire.

10 words and a link

2008 Cluster Challenge results

TotalView available on RHEL 5.3

Intel's Parallel Composer in beta

OpenCL 1.0 spec ratified, announced at SIGGRAPH Asia

Open MPI featured on tv

Four paths to parallelism with Java

AMD warns Q4's sales will be weaker than expected

IBM discounts to win new Power business

Sun launches X6640 HPC blade server

Computerworld wonders if HPC can save the economy

NCSA and Nimbis sign on the line

Sun's Network.com closes to new users

Joe Landman writes on clouds and HPC in Linux Mag

Lloyds invests in HPC, Windows inside

SGI delists, lays off 15 percent of workforce

A raft of bad news this week for SGI. First, Silicon Graphics announced that it received a letter from the Nasdaq Stock Market on Dec. 2 indicating that SGI's stock sat below the minimum market value required for inclusion on this market's listing for too long. SGI's stock sat below the $35 million minimum requirement for 10 consecutive trading days leading up to the notification.

Silicon Graphics has thirty calendar days, or until Jan. 2, 2009, to regain compliance with the aforementioned financial benchmark.

SGI confirmed layoff rumors Thursday afternoon with the release of an official announcement:

The company has refined its focus and is restructuring to adjust to changing customer and financial market realities. As part of its restructuring, the company will reduce its workforce by approximately 225 positions, or roughly 15 percent. The reductions will include several executive and senior-level positions.

Although the release doesn't specify where the cuts happened, a source close to the company says that the cuts were primarily on the commercial side of the business, while the federal business was left mostly intact. Those that were let go left the company Wednesday.

The executives who were dismissed are not named, and I assume we'll have to wait for subsequent filings with the SEC to know for sure who has hit the bricks. But comparing today's leadership pages at SGI.com with earlier versions insideHPC notes that Shahin Khan, VP of marketing, Barbara Stinnett, VP of Services, and Bill Trestrail, VP of APAC, are all missing. Presumably victims of the cuts.

According to the announcement, the company will retain and focus its efforts on servers, visualization, and innovations like the Industrial Strength Linux Environment.

Cray goes back to the debt sale barn, loads up for Christmas

With all the very bad news in the economy these days, it can be hard to remember that every crisis is someone else's opportunity.

Lately, that someone has been Cray. Back on October 1 Cray announced that it had repurchased $25M in debt at the discounted price of 92 percent of par value, plus interest.

What does this mean to Joe the HPC guy? It means that Cray got to borrow $100 and only pay back $92, plus whatever interest had accrued. Pretty sweet deal, made possible by banks willing to deal hard for fast cash to cover tough times on other parts of their business. It was also a signal that Cray had real confidence in its business position, cash on hand, and in its ability to borrow in the future if needed.

As the financial problems have deepened over the past month or so, it seems that banks have gotten even hungrier for cash. Cray announced in a Form 8K filed with the SEC today that they've bought another bolus of debt back at a steeper discount:

On December 10, 2008, we repurchased $6,073,000 in principal amount of our 3.0% Convertible Senior Subordinated Notes due in 2024 (the "Notes") at a price equal to 88.75% of par value plus accrued interest and commissions.

After this repurchase the company has roughly $34M in convertible notes outstanding.

Go team rocket! This, along with the raft of Form 4's filed in the past month are strong positive signals in an industry going through a pretty tough business cycle.

(Nota bene: A Form 4 signals stock buying by company insiders, and there have been 8 filed since Nov. 12.)

Univa UD's UniCloud offers toolset to access Amazon's pay-per-cycle model

Found at HPCwire:

UniCloud enables organizations to provision and scale HPC capacity on the proven computing environment of Amazon Web Services, expanding baseline computing resources through the dynamic provisioning of capacity to meet peak demand. An extension to Univa UD's leading UniCluster and Grid MP products, UniCloud allows organizations to establish workload policies and requirements which dynamically trigger the setup of virtual compute nodes in Amazon Elastic Compute Cloud (Amazon EC2), the Web service that provides resizable compute capacity in the cloud. It is designed to make Web-scale computing easier for developers.

Univa UD is also offering a whitepaper that describes how the whole thing works. You do have to buy UniCloud as an extension to your existing UniCluster or Grid MP licensing.

Univa UD recently published a how-to white paper addressing how to run UniCluster in Amazon EC2. The paper, along with a Q&A webinar with the author, is available for download at http://univaud.com/hpc/webinar20080917.php.

Amazon EC2 pricing, which you'll have to pay for after your UniCloud license, starts at $0.10 per instance hour for a small Linux instance ($0.125 per hour for Windows), and goes up to $0.80 per hour ($1.20 for Windows) for the extra large instance. In looking over the Amazon EC2 instance descriptions, it looks like traditional HPTC users would probably need the XL high-cpu instance, based on this description

Amazon EC2 instances are grouped into two families: Standard and High-CPU. Standard Instances have memory to CPU ratios suitable for most general purpose applications; High-CPU instances have proportionally more CPU resources than memory (RAM) and are well suited for compute-intensive applications. When choosing instance types, you should consider the characteristics of your application with regards to resource utilization and select the optimal instance family and size.

This instance features 20 EC2 Compute Units (8 virtual cores with 2.5 EC2 Compute Units each). What's a Compute Unit?

In order to make it easy for developers to compare CPU capacity between different instance types, we have defined an Amazon EC2 Compute Unit. The amount of CPU that is allocated to a particular instance is expressed in terms of these EC2 Compute Units. We use several benchmarks and tests to manage the consistency and predictability of the performance of an EC2 Compute Unit. One EC2 Compute Unit provides the equivalent CPU capacity of a 1.0-1.2 GHz 2007 Opteron or 2007 Xeon processor. This is also the equivalent to an early-2006 1.7 GHz Xeon processor referenced in our original documentation. Over time, we may add or substitute measures that go into the definition of an EC2 Compute Unit, if we find metrics that will give you a clearer picture of compute capacity.

So I think you'd be getting 20 processors for an hour for $.80. Amazon recommends benchmarking the different instances to figure out which one you need.

-----

John West is part of the team that summarizes the headlines in HPC news every day at insideHPC.com. You can contact him at john@insidehpc.com.