Bladewatch.com

I was reading this article and it says new Intel Xeon processors using 8 cores are on their way? Very cool, I wonder how these processors will perform in terms of performance per watt? They might be ideal for virtualization or hpc. I’ll need to check out the Intel blog/site.

Share and Enjoy

With the economic conditions we’ve seen a range of organizations from retail to the investment banks cutting costs and people in order to reduce costs (see this article). Cutting contractors, pulling projects can provide an instant cost reduction but I’ve always wondered if this isn’t cyclical and cost businesses more in the long run? By this I mean not just in terms of stop/start costs, but also in terms of service delivery. The way it works is that as business does well, IT gets more funding, employs more people to help with projects or business as usual support, as business gets harder, business will cut back often to the point at which IT is unable to meet it’s service level agreements, it’s unable to deliver, so we start the cycle again, recruit more people.

The core reason though your IT isn’t delivering is not financial, it’s operational and political.  We’ll go over what I mean in a few minutes, but the key factor is that IT is seen as a cost rather than an investment, in doing so you place IT on the offensive in terms of justifying head count, justifying existence, the old “but you didn’t ask for that, so we didn’t supply it.”, it’s a two way street and as IT gets integral to the business, as we see more restructuring, the business doing more of the IT tasks and IT becoming more architectural, this will change.

A prime example of justifying headcount is this. If I run a windows server team and I keep missing calls, do I not automatically get justification for more head count? Is it more headcount that I need or just people and workload management I need? Are the existing processes working in a way that rewards failure and how is it related to the way you reward your team? Do managers with bigger teams not get more ‘rewards’ or kudos, is this not anti business in terms of cost and delivery?

We need to manage the concept that bigger isn’t always better, there are compliance issues of course – but consider the difference between having 10 engineers and five really good expensive ones that you can call all the time. The five really good ones will adapt your infrastructure, your monitoring to make sure they don’t need called. We need to abstract our businesses from ‘industry standards’ in terms of rewards and switch to one that works, that delivers for our business, industry standard is fine, but if it means that I am unable to get the right people, it’s costing your business more in the long term.

In the meantime:

Ineffective communication internally and with external users:

• What are our standards – why do we have them, how do I deploy an application which does not conform to these standards
• How do I request a desktop – what do I need, what questions will I be asked
• I want a free desktop, how can I do this?
• How do I request a server – what do I need, what questions will I be asked
• What do I need in order to request remote access for support/remote working?
• Why does IT enforce replacing the server after three years – cost illustration example/marginal benefit etc
• Why can’t I get a bigger user/departmental drive? I can get a drive from PC world which is 200GB
• I want to move a server – what do I need to know

We need to use blogs, wiki’s, team information, accountability, support issues – how do I health check a server? How do I health check applicationbatch7? We need the Pre-Answers and a tool to connect and communicate with end users managed by the help desk.

Inventory information is inaccurate

The main issue is the number of inventories, the asset purchasing database, the server list, the application list, the internal team databases/inventories or spreadsheets. Nothing typically ties up, if I ask the asset team how many DL380′s I have they can’t tell me that, they can tell me how many purchases we made in 2002 though, if I ask the server guys, they might say 300, the application team might say 7. But getting an overall picture one, that works isn’t always that easy. Just like the NHS, moving towards a central platform for data reporting/presentation becomes ever more important – for your server, your application to be supported, it needs to be in the live production system list.

But consider that your inventory, your list of servers affects:

• Billing
• Software licenses purchased
• Hardware support contract coverage
• Cross charging in terms of monitoring or middle-ware tools -  how many SQL/Oracle databases
• Who owns what – who gets called at 3am when it goes bang?

Operational in terms of budgets/accountability

We need budgets and accountability to illustrate what IT is doing, how it’s adding value to working for you, but at the same time, there needs to be built into the budgets a stuff field, where I can simply give someone an infrastructure task and say get on with it.

The prime example of this was Chris down in not so sunny Canary Wharf, he’s been trying to get IBM Director to work with their monitoring tool, however this requires a day or two of monitoring guys’ time, however he can’t get anyone to say yes to book the guy for the day or so. In the meantime, Chris spends an hour a day checking Director manually, when a blade blows up, they know when the monitoring tool reports ‘ping down’ and he reacts accordingly. If we consider the outage cost, the inconvenience and ‘cost of unexpected outage’, the bank has paid about twelve times over in indirect/support cost.

Lack of investment in the right way

This covers a number of things, the way we deploy IT, the short term thinking, the strategies which are written in stone but then overriden by pressing button seven, (I’ll explain in a minute).

Time to build

One of the regular complaints we get is the time of build, how long it takes to ‘build’ a server, but then is this not a result of a number of issues, lack of team work between the teams deploying that server, a server build that is designed for everyone and no-one at the same time. You’ll often find the build is created and managed for a specific site, in a Canadian bank, Canada make the build, but it will work for their region, where they might have DHCP boot, where there’s a build network, but the office in Scunthorpe doesn’t have a build network, so the guy has to sit manually and switch the cables or boot into the bios and change the network card boot order. The site in London is Front Office trading, so there is no dhcp, and they need a hardended Internet facing build – but that isn’t deemed as a priority by Canada, because Canada don’t have that requirement. In the meantime, as IT we have to tinker with the build, harden the existing build, test it, adapt it, all of which delays your go-live.

Different organizational standards

There needs to be different standards between enterprise, but can we consider the cost of support, that by validating more platforms means more testing, more delays, if we only had to support three types of pc and laptop we could package and update so much more quickly. But this requires a change in concepts, a change in billing and accountability, a move to fixed cost computing, either the remote sites get virtual pc’s/thin clients, or (including the enterprise), the pc is replaced to a set model every two years, anything not that model isn’t supported.

Investment in the application infrastructure but not IT

We need to be including the infrastructure in projects, by that I mean, as we bring online new applications, new systems, we need to include extra capacity and investment for performance monitoring, for application and infrastructure monitoring for management. To avoid those situations where an application goes live on infrastructure that is difficult to monitor or manage because there was no budget for time to configure the firewalls or have a management server in place.

Spend on support

The spend on support can often be too high for a number of reasons, your teams are reactive and the queues are reacted to not managed, your teams do not effectively work together and there can easily be the avoid speaking to Mike as he’s emotional. Related to this requires a key statement which I remember a CIO telling me about, “GET ON WITH IT”

Focus on delivery – I have a team of 8 windows engineers, I want three doing incidents, one doing infrastructure analysis and improvement (Insight Manager/Inventory updates), the rest doing requests/tasks. I want the manager shouting across how you doing with that call, can you take a look at this one…

Focus on how we work together as a team -  communicate our needs, what our operational goals and issues are – bad air over a lack of access, or a system causing issue can be so easily be resolved.

Delays in delivery or excellence simply because of poor people management, that Mike is emotional means the escalation teams try not to call Mike, is this not a problem on several levels. That escalation don’t have the access and the skills to do the first line? That Mike isn’t approachable, can we not reboot him?

Share and Enjoy

Check out this article, a new site has been launched covering environmental issues. The more we talk about the concepts actively debate the issues as vendors, service providers, consultants or end users, the more we can establish a way forward, the best practice, the way to configure servers/pcs a base of standards. The more we do this, the better in terms of financial/operational savings and in benefits to reducing our carbon footprint or environmental impact.

We need to work towards a wholistic approach covering acquisition, installation management and disposal, covering best practices and configuration as well as contacts or people that can re-use or recycle the computers, the machinery and everything else?

Share and Enjoy

An article talking about what to do with that old desktop, an issue that’s been on the news recently, did a research group not buy a series of pcs and find a whole host of important data on them despite them being wiped?

Regardless recycling the pcs need to continue, I’d like to see more services kind of based on the mopay model. If you haven’t seen it, you basically whack your phone in an envelope and post it, they then send you money for the model. I’m not so much after the money, just for example I have 256MB RAM chips, a few hard drives which I’ve wiped sitting around, they all work. I’d like to get them out the house, as I understand it, I can’t just post it to Dell or anyone so I have to take it to my local council. Could I not just post it to someone or drop it off in PC World locally to have it tested/re-used or recycled by a charity or anyone else?

Share and Enjoy

I was doing some reading up on cloud computing and came along this article, do check it out. How cloud will work within your organization will depend on a number of things, your organization and how it works with IT, the application or service as well as any legislation or ownership issues attached to them. That said moving to a shared computing model, one based on farm type platform using service provisioning and not ‘selling virtual or physical servers’ has to be the way forward. We need to be moving towards a concept in which IT provides a shared infrastructure – you pay for what you use, IT handles capacity planning and resources, it should be like your mobile phone in the respect that I need 9-5 Monday to Thursday, so I only pay for that, back office or HR have the resources the rest of the time.

Moving to a shared platform comprising grid/hpc, web and citrix should be where we’re moving too, but it presents issues of ownership, of trust and co-ordination, hand holding if you will. The I promise statements coupled with delivery and ownership of issues. When something doesn’t work, we need to contain it, resolve it, and illustrate the core issue, highlight that it is not related to shared infrastructure – that the lights were on ready for business.

Moving to VMware is the first step, absolving myself of the individual physical server, the underlying tin, we need to move to the next step, the next generation infrastructure powering the next generation of applications which are server agnostic. They’ll need database components, MDAC components registered, they might even need other bits registered, but crucially, I should be able to develop configuration templates so that I can migrate an application between infrastructures to maintain or improve service, to keep everything online working around failures or business requirements, whether it’s using gird or Blade Logic.

Share and Enjoy

I remain a fan of IBM’s iDataPlex, their modular computing platform, it is very cool, being able to allocate the storage, the processors and the memory. Do check it out, it’s talking about how this organization is rolling out iDataPlex in their data center to accommodate their future computing needs.

Share and Enjoy

I was doing some research about Bull servers, just seeing what they had to offer and read up on their products, when I noticed that they’ve launched a new range of entry level Escala servers based on the POWER6 processor.

It’s always great to see what different vendors offer, what innovations and new products they launch, so that I know what’s out there, maybe even say “… are offering this, when are you going to?”

Share and Enjoy

Check out this article talking about how to choose the best x86 server for your data center, it covers some great topics including the power utilization, price and systems management. Your server choice is going to depend on a number of issues:

• What centralized purchasing agreement you might have – you might get a volume/enterprise discount.
• What level of experience you have with a vendor – if you’re already HP, it might seem better to stay with that vendor for ease of management – you know what to expect, how to behave.
• What level of experience the support teams have – if the support engineers are certified against that vendors kit, switching vendors might require re-training.

Key things for me would be:

• Systems management – the ability to build servers automatically with your management tools – do they have a plug in for your monitoring tools/automated build tools like Altiris or CA NSM?
• Firmware issues – if I upgrade the firmware, does that mean I need to upgrade the drivers? Best practice would recommend but in the real world you might be busy, your build might be set to specific drivers.
• Lights out – you might not necessarily need lights out, but it’s useful to have and check two things, firstly is it included, secondly is there a license needed for the actual screen view using the lights out card.
• Manage price with functionality, you need to pay what you can afford, but consider the cost of switching vendor against direct savings.

I had a similar discussion recently, the client had three on site engineers supporting 200 Compaq/HP servers, they were considering switching to a mixture of Sun and Dell and they asked my advice.

• A few things:
  • You’re engineers are already familiar with the Compaq/HP support process, with the hardware – a DL580G2 is rapidly different against a DL580 – but they understand the concepts.
  • The hardware support contract might need to be adjusted to include spare parts for three vendors – check the cost of doing this, and check if the vendors/service providers want your engineers to be re-trained.
  • Check the management cost – your operating system build, does that need changed? Does your monitoring need changed? Do we need to provide additional hardware monitoring servers/tools?

We need to move with the times, we need to buy servers, storage and network infrastructure that are in line with the business requirements, at the same time, we need to manage the marginal cost of doing so. Switching vendors is not world ending and indeed can be achieved with relative ease, but as you scale out the infrastructure, as you think about it long term consider the impact of today and the near future where you might have several vendors and the associated support requirements, hardware and software.

Share and Enjoy

An interesting article talking about the development and innovation of Windows 2008 including Hyper-V. We need to continue to be enthusiastic about new virtualization platforms, understand that it is possible and reasonable to deploy different hypervisors for different business or operational needs, to lower the perceived technical and non-technical barriers to entry. Anything Microsoft can do to aid this, both with Hyper-V and with the cost of software licensing has to be a good thing for the server industry, the virtualization platform as a concept and the end user community adoption of the platform.

Share and Enjoy

I remain excited about Cisco entering the blade market (check out this article), the more competition and innovation in the blade market, the more accessible and affordable we make the platform. That said, we need to standardize on the basics, improve systems management both in terms of building the blade, but also of managing the hardware, the configuration and the monitoring. Anything we can do in these areas has to be a good thing.

In the meantime, I look forward to seeing how Cisco and now Oracle (owning Sun) are going to innovate these markets, building infrastructure solutions (as HP have with their HP BladeSystems Matrix or IBM/Dell and the others) that can be deployed to meet different business needs, identifying blades as a vehicle for change, for achieving your business goals.

At the same time, I look forward to seeing how we can create unified computing solutions, where the storage, the network and the server can be managed, vendor specific solutions are ideal for many customers, but open standards, open technologies that let me manage my Dell and my HP, my IBM and my EMC is just as important – not everyone is tied to one platform, I’m not asking for the vendors to do the impossible, simply provide the right hooks, the right tools to make it just that little bit easier for multi-platform management.

Share and Enjoy