Archive for May, 2008
May 14, 2008 at 5:15 pm · Filed under blades, rackmounts
I got an email form Jenny asking if she needed to power down a hard drive that has failed before swapping it, to prevent any data loss. I’ve posted the reply:
Hi Jenny,
Basically is your server has reported a failed hard drive (orange light) or the agents have failed the drive it should be safe to remove and plug in your replacement drive. A few things to check:
- Check the drive is the same size and model type
- Check the drive is the same firmware - sometimes the agents might fail to rebuild the drive if the firmware is out of date (depends on the server manufacturer)
- Ensure you have a valid backup before carrying out any work on the server - as ever if you have any questions contact your support vendor/team.
The actions to replace failed hard drive would be:
- Check event logs for errors/agents for any issues
- Remove failed hard drive, pulling the lever that unplugs the drive slowly
- Plug in replacement hard drive and watch for lights on the drive to come online - on the HP servers, the disk light in the middle flashes as the drive rebuilds the array.
- Wait 5 mins or so and check the event logs for any erorrs, then check the agents, IBM Director, HP Insight Manager or Dell tools to see the rebuild status.
May 14, 2008 at 5:26 am · Filed under Bladesystems_insight_conference, blades, rackmounts
http://www.bladesystemsinsight.com and http://www-03.ibm.com/systems/x/hardware/idataplex/index.html
I saw iDataPlex for the first time today and spoke with some of the guys at the IBM stand at BladeSystems Insight. I do love the concept.
That I can buy one rack and then add the modular components I need, is very interesting andideal for HPC or web infrastructure. I could have an array of shared storage for the individual servers storage (os/data) and then rack the number of systems that I need. I got to see one of the server modules, very very cool. Check out the IBM video.
May 14, 2008 at 5:05 am · Filed under blades
http://www.colfax-intl.com/jlrid/SpotLight.asp?IT=0&RID=121
Colfax CX6000 Workgroup Cluster (WC) built on Intel® Multi-Flex Technology is an affordable, simple, and fully integrated supercomputer-in-a-box. Colfax CX6000 WC puts the power of high-performance computing (HPC) in the hands of individuals and workgroups of engineers, scientists, financial, and creative professionals, without the complexity and expense of an enterprise-level cluster system.
I saw one of these for the first time and spoke with a few guys at the Colfax stand at BladeSystems Insight, it looked quite cool. I like the concept of shared storage which I can then allocate to the servers as I see fit, so if the server has a hardware fault you can point the storage to another server.
May 14, 2008 at 12:29 am · Filed under Other things
http://www.crucial.com/uk/store/partspecs.aspx?IMODULE=CT64GBFAA0
The rugged, reliable hard-drive replacement that’s redefining personal computing
- Flash-based storage
- Rugged and reliable
- Low power consumption
- High performance
- Lightweight
- Silent operation
Very, very cool, I can sense I need one for my ThinkPad - it should help performance, but more importantly my battery life. I will need to build up to one due to the price though!
May 13, 2008 at 11:54 pm · Filed under Grid, blades
http://www-03.ibm.com/press/us/en/pressrelease/24180.wss
ARMONK, NY - 13 May 2008: Driven by growing commercial need in areas such as financial services, digital media creation and medical imaging, IBM (NYSE: IBM) today expanded its High Performance Computing (HPC) capabilities for businesses with the introduction of the IBM® BladeCenter® QS22 — a new, economical supercomputing technology inspired by advanced scientific research facilities.
The heart of the QS22 is a new processor compliant with the Cell Broadband Engine™ (Cell/B.E.) Architecture, originally developed by IBM, Sony and Toshiba to provide the computing power for cutting-edge gaming applications. And for the most challenging arithmetic operations, this new processor, the IBM PowerXCell™ 8i, offers five-times the speed of the original Cell/B.E. processor.
This does sound cool. I remain a big fan of the Cell/B.E platform, any innovation of the platform has to be a good thing for the consumer in terms of choice and the blade platform. The extra memory support sounds very cool and should provide further opportunities/uses for those memory intensive HPC applications.
May 13, 2008 at 11:48 pm · Filed under datacenter
http://business.scotsman.com/business/RBSowned-data-centre-eyes-move.4075318.jp
A DATA centre business set up by Royal Bank of Scotland at the height of the dotcom boom is eyeing a push into the west of the country as it looks to double turnover over the next four years.Scolocate, which runs one of Scotland’s largest so-called “internet hotels”, providing corporates and smaller businesses with remotely-hosted computer services, was set up in 1999.
I highlighted this article to illustrate the demand for data center space for large and small enterprises continues. It’s an interesting read, do check it out.
May 13, 2008 at 11:15 pm · Filed under Bladesystems_insight_conference, The Green Grid
http://www.bladesystemsinsight.com and http://www.thegreengrid.org
The Green Grid presentation was quite cool, it was talking about what The Green Grid is working on and included some interesting comments, such as how power utilization is broken down in the average data center:
- IT equipment 57%
- Cooling equipment 34%
- Power distribution losses 7%
- Lighting/other 2%
There was also discussion of how to work out Data Center Productivity:
Useful work produced by data center/resource consumed producing the work=Data Center Productivity - I wonder what scores different data centers/organizations might get?
There was also the five ways to save server power:
- Identify legacy servers/servers not in use
- Enable power saving features
- Right size server farms
- Power down servers not in use
- Decommissions old servers no longer in use.
May 13, 2008 at 11:08 pm · Filed under Other things
I was having a conversation with one of the application teams about migrating their application to be compatible with one of the grid application tools. There was concern that this might affect the control that the application team have over the application, that by using ‘the grid’, they have to work within their constraints, the rules of deploying a new application, of making sure that their code is adequately tested etc. Their argument was that outside of the grid, they controlled the data source, they controlled the calls they made within the application code, and they (the support team) dictated when new application releases were released, when infrastructure could be taken down or not. That IT could take down so many blades during the day for maintenance and not tell the application team, reducing performance and resilience.
Let us step back just two steps. The reason we’re looking at something like DataSynapse, Platform or any of the others is hardware abstraction. Although you might have ownership of the infrastructure and the application, this doesn’t really change within the grid environment, all that changes is that we have a more industrial strength solution, a bit more control and transparent way of doing business.
You want to deploy new code to the grid? Not an issue, raise the relevant service request or change request, and submit it for deployment to the grid. Consider that it is not in IT’s interest to have servers not in the grid, both in terms of resilience and ‘branding’. Branding sounds such PR’ish, (if that’s a word), but what I mean by branding is end user perception, of managing user expectation and service delivery. We can anyway overcome such issues by giving application support teams read/view rights of the grid middleware, or access to the grid logs etc. What we want is ideally a transparent infrastructure, a common shared infrastructure with the right optimizations for the applications running on it, in an approved configuration so that the grid team can co-exist, can do the grid support, ensure that grid is available for applications to submit workload to the grid on demand.
The grid team in effect should be infrastructure based, should be the connector between the application support teams and production. They should manage the expectations of the user groups, the stakeholders and the teams that support the underlying infrastructure, working in line with (or as close to) the ITIL processes, with a clear understanding that grid is in an on demand world, where the processes should reflect this. I’m all for process, you want a change made to grid to fix a problem on the staging grid, we’ll do it in five minutes, but can you log a service request, so that we can log the setting changes in the system and maybe note the changes on our intranet site (one of those collaboration tool, even a blog).
May 13, 2008 at 11:08 pm · Filed under virtualization
http://searchservervirtualization.techtarget.com
/news/article/0,289142,sid94_gci1310403,00.html
Many companies deploy server virtualization with the expectation that reducing the number of physical servers in a data center will reduce power bills as well as payroll expenses. And virtualization companies often tout their technologies as a means to reduce staffing needs as well.
But in reality as several companies confront the various new challenges created by virtualization, data center staffing has involved more of a reshuffling of IT responsibilities rather than a staff reduction per se.
Do check out this interesting article talking about how virtualization might not be the cost saver that people are expecting. I like to think of virtualization as a vehicle for change, a way of changing your infrastructure to have it work around your business, to be dynamic, to avoid those, “the lead time is three weeks for your server”, conversations. We need to abstract headcount from the virtual infrastructure conversation - just because your servers are virtual doesn’t mean you don’t need the Windows/Linux and middleware guys. To reduce your staff, you need to identify the workload, those causal factors, attack the problem at source then move down the support chain to identify failures in service delivery, whether it’s monitoring, pro-active engineering (patching/clearing down log files etc).
May 13, 2008 at 10:47 pm · Filed under Other things, environment
http://www.prnewswire.com/cgi-bin/stories.pl?ACCT=104&STORY=
/www/story/04-22-2008/0004797293&EDATE=
SAN JOSE, Calif., April 22 /PRNewswire/ — Nearly two-thirds of IT and facilities personnel consider their data center energy efficiency “average” or worse — and their development and test environments might be the biggest cause of that, according to a survey conducted by Cassatt(R) Corporation, a leader in providing software to make data centers more efficient.
More than a quarter of survey respondents said that greater than 60 percent of their development and test servers are idle during off-peak hours. There is some good news, though: 62 percent are working on a data center energy-efficiency project now or expect to within the next year, according to the “Cassatt 2008 Data Center Energy Efficiency Survey.” And, contrary to conventional wisdom, 59 percent would consider turning off computers that are idle.
This article is a little old, but I was just doing some research on server utilization and came across it and thought it was relevant nonetheless. Thinking about your server utilization, about what systems (desktop/server/printer) need to be on for business to continue, and which ones can be shut down, can not only reduce your operational support costs, but reduce your energy costs. We can’t make big changes overnight, but examining what we can do within our operational/business constraints can still deliver/savings and (or) empowerment to the end user. That the desktops are rebooted should mean less calls, “my mouse is broken, my account’s locked out, Internet Explorer/Office freezes”, without necessarily causing that much disruption.
Next entries »
