|
|
December 15
Designer update
The Preen Scene
December 10, 2008 5:55 pm
 It’s been a very good year for Preen’s Thea Bregazzi and Justin Thornton. First, we noticed that both their main collection and commercial Preen Line have quietly become It girl and celebrity staples. It started in April with Gwyneth Paltrow showing off her gorgeous gams in a black Power Bandage dress, a perennial Preen style, during her Iron Man publicity tour. And in short order came more fabulous credits on Anne Hathaway, Leigh Lezark, Chloë Sevigny, Karolina Kurkova, Mandy Moore, and Sienna Miller. Most recently, Fabiola Beracasa wore their Frances dress to Aid for AIDS earlier this month; Cindy Crawford was spotted in the Band dress (Spring’s finale look) last weekend at a Darfur fundraiser in London; and Bond girl Gemma Arterton repeated Gwyneth’s Iron appearance stitch for stitch in a black Power Bandage at the British Academy Children’s Awards. (You can compare and contrast in the image at left.) Second, when we called their PR to confirm all of the above, we learned that the design duo’s first ever resort collection, made exclusively for Net-a-Porter, is close to selling out. (Although for those of you who want to work a Gwyneth or Gemma, there are still a few Power Bandage frocks left in stock, in black and a special Resort-only hot pink.) A good year, indeed.
—Meenal Mistry 
Gear: Tough Crowd
Etón FR500 Solarlink Radio
What good is going off the grid if you can't keep up on the Mets? That seems to be the premise behind Etón's line of rough-duty radios. As its name implies, the new Solarlink is powered by the sun (or, during a nuclear winter, a hand crank). It comes with a flashlight, emergency siren, and—should you need to get back on the grid—a charger for cell phones and MP3 players.
$80, etoncorp.com August 15
|
|
Applications
|
|
|
A blog for the IT professional around applications for Compute Cluster Servers. This blog may include links to white papers, small code samples or scripts, as well as descriptions of application deployment projects and strategies. |
|
2/28/2008
The Subsystem for Unix-based Applications (SUA) is a full-featured POSIX compliant "OS" that runs natively on the Windows kernel. It's actually a subsystem in parity with the Win32 subsystem. SUA is available as an optional Windows feature on OS releases since Server R2 and including Server 2008. SUA has an interesting history which you can find here.
SUA is useful for HPC application development and perhaps even more so for HPC application migration. Many HPC codes have a long history of availability on -nix variants. Making those applications available on Windows with literally no source changes and no performance penalty is where SUA shines.
A recent Platform Adoption Lab engagement reminded us of the value of SUA. The scenario involved migration of a large HPC code-base. The source included C, C++, Fortran, and Python. The project build configuration leveraged Autoconf to enable support for a number of target systems.
To get started with the source configuration we needed to inform Autoconf about our system (i.e. SUA). The configure script first runs a script called config.guess that determines which system configure is running upon. The configure script generated by Autoconf then simply checks functionality on the system, by looking for a compiler, and then compiling and running simple test programs. For reference, you may need to modify your config.guess file with information about SUA (aka "Interix"). Here's the change:
*:Interix*:6*)
echo i586-pc-interix${UNAME_RELEASE}|sed –e ‘s/\..*//’
exit 0 ;;
Essentially, many source packages migrate easily without further change, just a recompile. In fact, a number of open source codes have become available to Windows HPC users using SUA. A recent posting herein provides more information. Additionally, you'll want to review the following online resources:
| | | |
7/14/2008
“Microsoft impressed the supercomputing community last week when a beta of its new Windows High Performance Computing (HPC) Server 2008 scored a top 25 ranking on the Top 500 list of world's biggest supercomputers.” 当我看到新闻上类似于这样的报道时，总会有一阵激动，微软的HPC产品终于开始在市场上显露出了力量，而且也得到了越来越广泛的关注。
在微软推出HPC产品前，高性能服务器领域一直是Linux和Unix的天下，作为高性能计算领域的新加入者，微软HPC研发团队的工程师们可谓是筚路蓝缕，以启山林，一切都要从头干起。通过不断的努力，我们在2006年推出了Windows Compute Cluster Server 2003，初步获得了肯定；而今年5月刚发布的Windows HPC Server 2008 Beta2版本引起了业界的极大关注；6月在TOP500.org公布的世界第31次全球超级计算机五百强排行榜上，位于美国伊利诺大学的超级计算机应用国家中心 (NCSA) 的服务器，浮点运算性能68.48TFlops的成绩登上了排行榜的23名，而他们所使用的操作系统就是Windows HPC Server 2008 Beta 1。
现在，Windows HPC Server 2008已经能够有效管理数千个计算节点，具备高速NetworkDirect RDMA技术、远程直接内存访问界面、高效集群管理工具、面向服务器架构的工作调度、HPCBP规范标准集群互连等等；而且与前一个版本相比，08版产品的易用性，稳定性等都取得了很大的提高。比如我们能在很短时间内在成百上千台的裸机上部署好Windows HPC Server 2008，并可以运行一些诊断程序方便地测试出现有机器的连通性等。此外，通过Heat Map，我们可以直观地看到包括头结点在内所有节点的状态，包括他们的CPU和内存使用率等信息，这些特性都取得了很多合作伙伴的好评。来自NCSA的副主管Pennington在反馈中提到：“当我们在把Windows部署在超过一千个节点的机器上时，我们从这些机器是裸机时开始，到他们开始运行LINPACK测试程序仅仅花费了4个小时， WINDOWS HPC SERVER 2008的性能达到了我见过的同类机器中最高效的之一。”
除了在性能上的提升，我们也非常注重发展合作伙伴，从合作伙伴的反馈中改善我们的产品，就中国而言，我们已经和上海超级计算机中心，曙光信息产业有限公司等建立了紧密的关系，同合作伙伴一起，共同推动高性能计算产品在中国市场的普及与应用。
在产品不断取得成功的背后，我们中国研发团队也功不可没。Windows Compute Cluster Server 2003中的界面管理曾经被很多人批评，但是如何使得我们的HPC Server 2008具备强大复杂的功能的同时，又兼具易用等特性是一件非常有挑战性的工作，这直接关系到用户体验，如果在界面上就难倒了用户，则整个产品就不再让人感到舒适了。经过我们的不懈努力，以及不断吸取客户的反馈，Windows HPC Server 2008在UI（用户界面）方面做了非常大的改进，包括管理界面，配置向导等都做了很大的改善。例如在网络拓扑结构选择部分，我们将抽象的网络拓扑结构转换为直观网络示意图（图1），这样即不容易出错，也易于用户理解和记忆。
图1
此外，前面提到的HPC Server 2008中受到很多客户欢迎的Heat Map功能，能够让管理员实时监控各个节点的状态，如图2中所示，管理员可以看到现在的所有节点中可用的物理内存（Available Physical Memory），硬盘的吞吐量（Disk Throughput）和每秒系统调用数（System Calls / S），并且这些指标都用图形化的方式直观的表现出来。
图2
另外，强大的Reporting功能，也是由我们中国研发团队完成的：集群的管理员可以将一段时间内关于集群上的信息收集起来，并以图标的形式表现出来，以方便管理员对这段时间内集群的状态进行分析，图3中精美的图标就是Reporting功能生成的结果：
图3
在UI上的巨大改进也获得了很多客户的正面回应，来自英国的达斯伯里实验室（Daresbury Laboratory）分布式计算组的计算机科学家Igor Kozin说，“Windows HPC Server 2008的一个很突出的地方是它具备非常可靠的图形化用户界面，从很多方面来看这些仍然是Linux所缺乏的”，利用我们的UI界面，集群管理员能很快的上手并且高效的操作，因此大大降低了花在配置上的时间和精力。
除了在UI方面的改进，为了满足不同客户的需求，我们也支持在PowerShell中用命令行来完成包括管理和提交Job等。对于习惯使用C Shell和Korn Shell的管理员来说，使用PowerShell脚本会相当方便。
我们还增加了对SOA的支持，提供了API甚至在Visual Studio中嵌入了Add-In，使得开发人员能方便的利用SOA去部署并行程序。开发人员不用去关心他的程序会部署在哪些节点上运行，我们提供了WCFBroker机制去自动的将程序部署到合适的节点上，简化了开发人员编写并行程序。
总而言之，我们提到的这些特性都是以给客户提供价值为核心的，目标是实现我们和客户双赢的局面，因此也获得了众多的好评，而用户的意见甚至批评都成为我们不断改善产品的驱动力，期待Windows HPC Server能取得更大的成功！
刘贤斐
微软中国服务器与开发工具事业部(STBC)
项目经理（实习生） | |
|
|
|
|
Tools 
![Picture of [☆_ναłeитıиα_☆]](http://img.wlxrs.com/$live.controls.images/ic/iciconmap24.png)
![Picture of [☆_ναłeитıиα_☆]](http://img.wlxrs.com/$live.controls.images/ic/bluemann.png)
Send a private message
Tell a friend
Add to My MSN
Add to your network
Sign up for alerts
