Bladewatch.com

GridwiseTech

GridwiseTech has developed AdHoc, an advanced framework for sharing geographically distributed data and compute resources. It simplifies the resource management and makes cooperation secure and effective.

The premise of AdHoc is to enable each member of the associated institution to control access to his or her resources without an IT administrator’s help, and with high security level of any exposed data or applications assured.

It takes 3 easy steps to establish cooperation within AdHoc: create a virtual organization, add resources and share them. The application can be implemented within any organization to exchange data and resources or between institutions to join forces for more efficient results.

AdHoc was initially created for a consortium of hospitals and institutions to share medical data sets. As a technical partner in that project, GridwiseTech implemented the Security Framework to provide access to that data and designed a graphical tool to facilitate the administration of the entire system.

I got emailed about this, it certainly sounds interesting, anything we can do to make the management of resources around the geographical or logical areas of the enterprise has to be a good thing, I’m off to read more.

HP.com

PALO ALTO, Calif., Aug. 24, 2009 – HP today announced the fourth in a series of webcasts that provide an in-depth customer view into the inner workings of the technology used in today’s data centers and the business benefits they deliver.This week’s webcast, titled “Deploying a Cloud Computing Infrastructure,” is scheduled for Thursday, Aug. 27.The HP “CIO Real Talk” webcast series features chief information officers (CIOs) and IT executives from leading companies on a panel moderated by Jonathan Eunice, principal IT advisor at Illuminata. During each audio webcast, IT executives will discuss important industry topics and share real-world experiences as to how they transformed their technology infrastructures to support and drive business objectives.

Any discussion along the supports, the process and best practice would be very interesting, saying we’ll move that on to cloud next Wednesday is fine, but fitting that within the corporate environment, making the billing work, identifying the lines of support, the liabilities and service level agreements, even deciding what is ‘cloud compatible’, what is core to my business operationally/technically, whether I want an internal cloud, or buy services from Amazon etc. It’ll be good to check out, I’m off to read more.

http://www.hp.com/hpinfo/newsroom/press/2009/090824xb.html

HP today announced along with the Canadian arm of GS1 – a leading global nonprofit organization dedicated to improving supply chain efficiencies – a cloud-based recall service that traces and removes potentially harmful food products from the supply chain.

The GS1 Canada Product Recall service will run on the HP cloud computing platform for manufacturing, which allows companies to see and share information across the supply chain. Food and consumer products organizations can use the service to reduce errors, decrease the amount of time it takes to respond to a recall, and mitigate the costs associated with managing the recall process.

“Global supply chain standards are foundational to effective product recall,” said Art Smith, president and chief executive officer at GS1 Canada. “GS1 global standards are used by millions of companies around the world to enhance the safety, security and efficiency of their supply chains. With the integration of GS1 standards and HP’s cloud computing platform, we are reinventing the way that recall information is exchanged between businesses, and further supporting industry efforts to improve consumer safety.”

It’s always interesting to see how organizations in the enterprise and non-profit communities are benefiting from the technology, whether it’s revenue generation, cost reduction or service improvement. An interesting read, do check it out.

http://www.finextra.com/fullstory.asp?id=20400

California-based Tibco Software has acquired privately-held grid and cloud computing vendor DataSynapse in a cash deal worth around $28 million.

Very cool, I wonder if this might further the possibilities of DataSynapse in the market data or rendezvous or multicast space. I wonder if we’ll see any new products or services resulting from this deal or pre-configured solutions for the enterprise. I was on their site this morning and it’s been updated already, do check it out! I’m genuinely excited to see if this might extend the possibilitied of grid within the enterprise, we’ll have to see.

I first started server engineering in 2002, I was young, just out of university and had installed Windows on a server only once. I had hardly ever supported one server, let alone a room full of them, and when I joined a bank, I started working with Compaq servers.  I saw them evolve to HP, worked with Dell and IBM servers  as I moved around the city and I have been following their evolution ever since both personally and here at bladewatch.com. So it was with no small amount of excitement that I received a Dell R710 blade system on trial which I could play with, break and fix, tinker with until I understood how it all worked.

Dell has been making many significant improvements and innovations to their server product range, from the great blade enclosure that they released (the M1000e) to up to the PowerEdge 1950 III which I recently installed at a local small business as a file server.

The R710 is Dells 11th Generation 2u rackmount server, and they have made some smart updates to the range in response to user feedback, with a focus on customer driven design, energy efficiency and systems management. The metal handles on the new hard drives are an illustration of this. The 2u rackmount server is increasingly getting a lot of exposure because of the versatility of the platform. I can now have a 2u server with two quad core processors, a massive 144GB RAM and 6TB of local storage which strongly extends the use case possibilities.  For example, the R710 could be an excellent stand alone application server, a remote virtualization solution, or an easy way to deploy an ESX farm of servers in an enterprise, making the server a processor in a box and having everything hosted on SAN.

With this in mind there has been a lot of competition and innovation in the 2u/3u format as vendors try to extend the possibilities of their platform, for more memory, more storage, network cards and power supplies, to improve virtualization, storage or high availability offerings. The vendors have additionally continued to increase their management and support offerings, to make deploying and supporting their products that bit easier in line with customer needs.

The specifications then:

• 2x processors (either dual or quad socket)
• Support for 144GB of RAM
• Support for up to 6.0TB of local storage
• iDRAC lights out card
• 4x on-board 1GB network ports
• On board SCSI controller

The server comes with a new front cover, an improved LCD display which allows you to view system information like the DRAC configuration, and a new embedded system, the LifeCycle Controller (which is rather good). This allows you to boot straight into the tool and configure set options using their GUI. The amount of time I would save in the data centre with this kind of offering is hugely compelling.

With four on board network ports, impressive memory and storage support, the server will meet most application and virtualization requirements, particularly with the expansion slots to plug in your iSCSI or SAN storage or network cards.

The onboard lights out or iDRAC6 management controller furthers the ability to remotely manage and support the server, offering support for remote media and KVM over IP.

From a server guys perspective two thoughts come to mind.

1. Dell have certainly improved their design and attention to detail, the R710 seems generations ahead of their recent PowerEdge 2950 III.
2. The R710 also highlights how scalable the 2u rack server has become and makes me wonder about those four socket boxes, with multi core:  Am I better buying relatively disposable low cost 2u servers, replacing them every few years than investing in high performance higher cost 5u boxes? We will have to see, in the meantime, well done to Dell for their innovation on the R710, long may it continue.

Now, theres a final issue that I had to examine. A lot of the people I come into contact with around the city bleed blue blood. That is, they are HP through-and-through.  Offer them an alternative manufacturer and they simply wont give it any attention at all.  How does an engineer whos a total HP fan react to the Dell 710? Could it ever match or exceed an HP DL380 G6 in his eyes?

Well I thought we should find out  So I got hold of Barry (definitely an HP fanatic) and asked his viewpoint after showing him the demo unit.  Heres what I asked:

Barry, as a server guy raised on HP, does it match or exceed the HP DL380 G6?

Not quite, but I wonder if that is not more down to me as a result of my knowledge and experience with HP servers, coupled with their SmartStart product, their Proliant Support pack (which makes driver maintenance just that little bit easier).

That said a well put together package, manufactured to a high level, an effective server offering which would suit many a business or application requirement. Put bluntly, as my good friend Barry said many months ago when I asked for his views on the different server vendors, Dell? Almost there, just not HP, but on price, on value it’s just what the server CFO ordered.

Goodness me, thems fighting words. Nicely done Dell.

Finally, it is time for the Bladewatch rating:

Performance: 4/5
Build Quality:  4/5
Usability:  4/5

Summary: Excellent price, super 2u performance, fantastic stand-alone or networked workhorse, smart innovations (we particularly love the LifeCycle Controller function).

About gosh it must have been 17 years ago, my dad was buying a new car, it was between at the time a Vauxhall Carlton (or GM in the US), and a Volvo. The Volvo man said basically, “it comes with a stereo, steering wheel and seatbelts…”, the Vauxhall (or GM if you like) came with everything, leather seats, air conditioning, so a Vauxhall Carlton was ordered and delivered to a very pleased daddy.

A few years later, it was car time again, dad looked around, another Vauxhall, a BMW or a Volvo? As a result of re-structuring some more interesting products, dad turned up looking at these Volvo 850′s, and was astounded when the man said, “if you want that, buy the SE pack, it’s £2000 but you get……..(a very long list of stuff)” basically everything my dad could ever think off (though my little brother noticed something called electric seats), but I’ll move on.

They’d changed the pricing structure, they had re-branded, re-badged or compartmentalized the cost. This is the base (but we’ve only sold 3 base ones world wide), it’s supplied with these features and functions. If you like you then choose bundles for enhanced features, (like the S or the SE pack in the Volvo 850 world at the time), which brings us on to my concerns with pricing in the IT world.

I remain a fan of the concept of compartmentalizing the pricing structure, it allows the end user to choose the features, the components they need to deliver the service/features they need. However, the issues ignores two challenges, marginal cost, I wonder if it’s not cheaper to supply one unified configuration or a very limited variant on that theme, a light and full version rather than effectively a per customer configuration.  Secondly everyone loves a bargain, by focussing on the low initial cost, you run the risk of disappointing the customer when by the time they go through everything, add on the bundles the price is significantly higher and everyone feels disappointed. I’m also concerned that in lowering the cost you might reduce the user experience for what is in real terms reduced savings?  Can we not have a McDonald’s style meal, or like the Volvo, standard, S and SE? Entry, middle and Club class?

There is something though that I’ve noticed more recently the concept of licensing, of packs, of bundles/extras.  I appreciate that software companies need to earn revenue, they need to protect their intellectual property, but as the CIO and the specialist both mentioned to me on separate occaisions that the list of what’s not included in the box seems to be getting that little bit longer than what is deemed an extra function, an add on which you do not need to use if you wish to use the product.

That’s fine, but we need to address and consider what is a core function, whether it’s virtualization, operating systems, database or monitoring, there are core aspects of the product which I feel should be included, put finely, I’ll buy the SE pack, but the per socket/per instance/per data center/per ‘excuse number seven’ as to why the base price is £2 million but an extra £700,000 for this feature of the bundle pack and that part is a fine line.  I wonder if this has not increased the adoption of open source?

As the CIO had complained to me, the cost of an extra processor on their public facing database server in licensing costs was more than replacing the hardware – that CIO was now wondering is it that database he needs, or whatever open source downloaded written by a guy called Marvin on Tuesdays version.  Or the specialist who illustrated wonderfully how the lines had been re-drawn so that in the new version, what was core is now ‘an extra feature’.

The reason for the article was to highlight that as we virtualize, as we consolidate, as we achieve more with less, using smart technologies, even smarter applications and open source products, we should be making it cheaper to use the ‘original products’, so end users want it. Windows is THE example of this, if Windows XP was £40 even £50, I wonder how many small businesses would be found running illegal copies of Windows, or indeed might be thinking you know what stuff it, I’ll get a mac, I’ll try this linux stuff, or even application streaming/Citrix.

By making it expensive, by making bits extra, you make the end user think:

1. Do I need it?
2. Can I buy it from someone else
3. Can I get it done in house or buy in a product that’s cheaper

Could it lead to the end user experience being as one end user had said to me, “…frankly it’s pants, everything’s extra, designed by people from Neptune, I’ll do everything I can not to use it or recommend it.”

I could be accused of being emotional, but what you need to consider is internal mathematics, yes cloud might arise, outsourcing could transform the landscape, but eventually we can only shift the cost, play with the maths so long.  Like the CIO had highlighted to me, the recent quote he got to monitor 1200 blades in his data center in sunny Wales, was £20 per blade, fine if you’re deploying say 50 servers, but remember that’s a monthly cost. Now the monitor people might say “but look at what you get for that”, but consider it’s not them that have to convince the application owner, the business sponsor that they need all that functionality for a bunch of blades running in a grid farm which is going to cost him £24,000 per month in order to conform with IT standards for support.

Within the realms of IT, there is often the phrase ‘in support’, this means that the IT teams officially support it, and in some cases are able to obtain support from the original vendor or service provider.  There are of course those systems that have been out of official support, these are the ones that we tend to put on the ‘risk register’ that require a code upgrade or hardware refresh. Please note the dates are from the UK Microsoft lifecycle site, check your region and with your vendor/service provider or IT team to establish what is supported, as an example:

• Windows NT 3.51 – Mainstream support – September 2000 – Extended support – September 2002.
• Windows NT 4.0 – Mainstream support – December 2002 – Extended support – December 2004.
• Windows 2000 – Mainstream support – June 2005 – Extended Support – July 2010.
• Windows 2003 standard edition – Mainstream support – July 2010 – Extended support – July 2015.
• Windows XP – Mainstream support – April 2009 – April 2014 – Extended support
• IIS 5 – Mainstream support 06/2005  – Extended support – 07/2010.
• SQL 2000 – Mainstream support – April 2008 – Extended support April 2013.
• SQL 2005 – Mainstream support – April 2001 – Extended Support April 2011.
• Exchange 5.5 – Mainstream support – December 2003 – Extended Support January 2006.
• Exchange 2000 – Mainstream support – December 2005 – Extended Support January 2011.
• Exchange 2003 – Mainstream support – April 2009 – Extended Support April 2014.
• SMS 2.0 – Mainstream support – April 2004 – Extended Support April 2011.
• SMS 2003 – Mainstream support – January 2010 – Extended Support January 2015.

Do remember that what the vendor supports and what you actually use can so easily differ, that Windows NT has been out of support does not use mean many enterprises are not still using it, whether through choice or because their application, their system is validated for that platform. I was speaking with a client the other day that had said virtualization had allowed him to run legacy systems a little longer whilst the code was updated, by having their systems run on a virtual machine as the new hardware wouldn’t run the older operating systems.

Some ideas about setting up or optimizing the configuration of your file server:

• File shares should be on a separate disk to the operating system, so if you need to re-install windows you shouldn’t have to impact the data that is already there – C for Windows, D/E/F etc for data.
• Network card set to the correct speed and duplex to avoid any network issues, coupled with switching on the buffers/time out settings on the card if necessary.
• Drivers and firmware up to date – it shouldn’t have any impact, but let’s not forget that if it’s done, we can rule that out.
• Format the file system in line with the files that are going to be there if possible – if we’re going to have millions of 1k files, then 1k cluster size it shall be.
• Backup the share names – so we can quickly restore them in the event of failure – a net view \\servername will do to a text file using a scheduled task once a week?
• Consider separating the file shares/drives per business line – all sales on d drive, all IT on E, and shared files on f – that way if we want to take action on a drive, we minimize the people impacted.
• Think about how you permission the file shares – should we be using file level permissioning or share permissioning, but use groups for easy administration and auditing.
• Defragment the file system regularly, it can make a tremendous difference to end user experience and reduce the workload of the server itself as it tries to locate and serve files. Remember that the defrag works best when there is free space on the drive.
• Archiving – what is it that really needs kept online?  Could we offload some data to an ‘archive file server’, or have IT restores on a separate box so we’re not mixing client data with back office support work?
• Anti virus configuration – if it’s a high volume file server, do we need both the client and the server scanning on both read and write, or can we turn off scanning when reading a file on the server?
• Backups – the defragmentation of the file system should help improve backup times, but what needs backed up and how regularly, can we stagger the backup? Backup the IT drive during the day, sales after 6pm and shared at 8pm with the system itself at 10?
  
• Remember that even a Pentium 200/400 server running Windows 2000 can easily serve hundreds of users running Windows NT4/2000 if everything is optimized, it’s not the performance of the box as much as it is the configuration, the steps you take to optimize it.

When you do run a defragmentation the first time it might take a long time, but once you’ve done it a few times it should fly through, you could schedule it using a scheduled task just running a command prompt – defrag d: -f -v, defrag d drive, force the defrag with the verbose switch so if there’s an issue you can see what’s happening on screen. The defrag might be run during the day, providing you consider the impact on server performance and monitor it.

Over the last few months and since I’ve began writing this blog, one of the key things that has always been on my mind and that of the end user community is inventory. Inventory is key, knowing what you have, who owns it, and what it is underpins everything, your support matrix, your decision making process – what can be virtualized, what’s an operational risk and what isn’t etc.

With this in mind, with discussion with many teams and individuals at different levels I requested a bladewatch CMDB be written, an inventory tool a tool that can be used in partnership with whatever in house or bought in systems you have to aid in the management of your infrastructure – something that becomes ever more crucial as the infrastructure gets more complex.

From feedback from end users in the CMDB management/reporting space were:

• Too many systems that didn’t talk to each other – there were always just a few systems that we couldn’t talk too
• Support documentation was in a separate location to the server – so at 3am when server Manchester18 goes down, the engineer might spend 35 minutes looking for an 80 page support document in our SharePoint site when he only needs to know if he can reboot it as it’s out of memory – making a simple support call an issue that got escalated unnecessarily
• The high entry cost of a CMDB, they all seem to include a ‘consultancy fee; a development and set up fee and even worse sometimes a per server or per data center site – oh you’ve got virtual machines, you’ll need our virtual plug in, our different user view plug in.
• The proprietary design – “that adding a field could require a re-design or weeks of development time”
• Ownership issues – “you can’t add an application field, we own the code..”

With this in mind we asked some CEO’s/CIO’s and CFO’s what their core requirements were from a CMDB. We outlined what was possible at a fixed cost, what functionality was crucial, what was a nice to have, and most importantly what we actually needed. I’ve paraphrased them:

• How many servers do we have – How many are virtual/physical
• How many are in data center in Swansea (our BCP site) What ratio of servers in the data center in Swansea are production and who owns the most
• Can we have a list of applications and servers with their application or business contacts
• Can we have a per application view of the data center, the server estate etc
• Centralization of the data – a gold source which everyone agrees on

Oh I can get the information, that’s not an issue, but it’s having it in a format that is reportable, that does not require the world’s most complicated vlookup with access to the 8 different monitoring and inventory tools that I might have. The crucial concept being that accuracy to the nth degree is not as important as having an overall system view for noting changes to individual systems, for reporting where we are with our IT, for giving a per application, business line or even IT team view – a networks view, a Windows server or trading system view, with the right level of access to protect our data as required.

With this in mind I’ve had a developer working on a CMDB. Naturally what I view as a CMDB will of course differ from what others may think, and in the typical enterprise there may be an industrial strength solution which crosses all networks, which inventories everything automatically on Tuesdays at 3:17am. But for everyone else, who needs a tool that allows me to, keep an inventory, manage our server estate and add value to the support function, we have the bladewatch cmdb.

I’ve seen the first draft and frankly it rocks, I’ve requested some detailed changes once they’re done, within the next few weeks, we’ll be making it live and launching an online demo, I’ve yet to get the development bill for all the code, but regardless the viewpoint was always to develop a mixed business and IT tool, to reduce the barriers to support in terms of inventory and in empowering your technical teams to have one central repository for notes about systems and applications. More soon. If you have any questions do email me.

http://www.bladewatch.com/2009/08/05/dell-r710

I’ve taken delivery of a Dell PowerEdge R710 from Dell in order to do a review which follows shortly.

In the meantime, here’s a few pictures, which you can see on my flickr feed.

This PowerPoint document illustrates the ports and expansion support, do check it out.