Why does Longhorn Server need a Desktop Experience? Windows Media Player has no more place on your mission critical server than Pinball. At least with Longhorn Server it’s an optional extra, presumably for developers who haven’t yet got the hang of running the OS they’re targeting in a virtual environment and want a little entertainment between compile cycles.

User interface matters as much for administrators as for any user - perhaps more. After all , clicking the wrong checkbox on the server can be considerably more catastrophic than doing the same thing on your desktop. But why are we clicking checkboxes on servers at all?

Time was, you had to know what you were doing before you started. Novell Netware and the various flavours of Unix you ran on RISC servers were as user friendly as a cornered rat. No encouragement to click an option ‘to see what it does’.

Putting a GUI on Windows NT Server made it accessible. It meant anyone who had mastered Minesweeper had a chance of setting up a file and print server. That was fine as long as NT was essentially used for toy servers in small offices where the worst that could happen was losing a print job. But once you move to complicated tools like Active Directory and SQL Server databases the potential for disaster with what looks like a familiar operating system is far greater. A wizard makes you feel you can work out what you need to do next as you go along, but even with the roles we’ll get with Longhorn Server need planning in advance. Making something complicated accessible isn’t the same as making it simple. And setting it to music does nothing but chew processor cycles.

After hitting three airports in 14 hours - and in retrospect  flying to San Francisco via LA isn’t as good an idea as it sounds when you’re checking ticket prices - I’m struck by the similarity. You might prefer to fly from the Apple of airports, San Francisco International; clean, white, shiny, with some excellent services, some very neat design touches (look up above the check in desk) and the feel of the future. You have to sacrifice some things; very few departure lounges, not very comfortable gates and if you want to run a full range of applications you have to pretend you’re in another airport. But mainly, think about the switching cost for making another airport - any airport - your main departure point. Either you move there wholesale or you go a very long way round to fit it into your journey. Sure Heathrow is building a shiny new terminal and has been for years; it’s clear, connected and has a nice vista. But it doesn’t have that exotic feeling of flying off into the unknown…

Not to stretch the metaphor any further, there’s one service that an airport that makes more sense if you look at in computing terms; baggage handling. The flow of luggage through the system is clear in the abstract, but following one bag is like following a single packet of data. You know where it’s been and who scanned it at what point at what time. But when it’s going to make it  onto your luggage carousel? There’s nothing in the system responsible for tracking and reporting the delivery. It’s like waiting for the date of a new OS release; you won’t really know until it trundles down the ramp and falls over.

Mozy (www.mozy.com) is a free backup service that uses viral marketing rather well. In exchange for a free 2GB account you have to accept a regular email newsletter with the kind of down-to-earth ‘I’m your mate’ chumvertising that you get on the side of an Innocent smoothie bottle, but if you pimp the service to a friend who signs up for their own 2GB account, you both get an extra 256MB of space free (and you can keep doing that until you run out of friends with common sense).

2GB isn’t enough for any kind of full backup, but you can point it at key files and get an off-site backup that you can restore on any PC via the Mozy website. You can retrieve individual files or the whole bundle; a friend of mine has already used it to recover her dissertation after some over-enthusiastic spring cleaning.

It’s really quite painless. Once you pick the files or file types you want, backup is automatic. Mozy runs as a service - and here’s where it’s handy for a server - whether anyone is logged in to the PC or not. The first backup takes a long time because the files are compressed, encrypted (with your own passphrase if you prefer, a random key stored with your account if you don’t want to keep track of a passphrase) and trickle uploaded. Differential backups are much faster as long as the files haven’t changed too much. The main drawback with Mozy is that it can’t yet backup a mapped drive or a UNC path, but you can get round that by running it on the server.

Obviously your server already has a full backup solution and you test the restore path regularly. Key data goes onto tape, or a NAS box, or optical disk, and you have offsite arrangements. Everyone protects their important data, right?

Back in the real world, if you don’t have a backup solution already, online services are worth considering and for an SME the $4.95 a month Mozy charges for 30GB is quite reasonable. Even if your server is fully backed up, putting a handful of key documents and configuration files on the free Mozy service is a useful belt-and-braces option because restoring is so simple. It’s also a handy way to share files between PCs without having to remember to update a USB stick every time you edit them.

Mary Branscombe

I like to practice what I preach, so I use virtual machines as a development and test sandbox. They let me push software a little harder, and give me a consistent (if virtual) hardware platform for all my tests. It also means that I don’t have to invest in more hardware for our already crowded office.

It’s a great way of working. Your host OS is isolated from any mistakes you make, and with the latest tools you can cut and paste from guest to host and back again (as well as keeping a copy of a clean image so you don’t have to reinstall everything). But sometimes something goes wrong that leaves you trawling through Google in search of some crumb of comfort (or even a fix).

I’ve recently been working with Windows Server “Longhorn”, and decided to test it out in VMware Workstation. It’s long been my preferred VM for testing software. I’m familiar with it, and the guest OS tools make it easy to work with. I made sure to be running the latest version, after all, I was going to be using it with beta code.

Setting up a basic VM isn’t difficult. VMware’s wizards walk you through the process quickly and easily, leaving you to sit back with a cup of tea while it prepares your virtual disk. It’s also able to host ISOs, so you don’t even need to burn a DVD from the OS image Microsoft provide on MSDN.

I fired up the VM, and watch the installer boot. It loaded the WinPE install environemnt, and then hung. I tried again, and again. No joy - and no graphical installer. It was time to hit the search engines.

It took a little while to track down the solution, somewhere on an Ars Technica blog. It turns out there’s a subtle interaction between VMware Workstation and the Windows Server Longhorn installer (there’s no problem with VMware Server). The fix is quite simple. All you need to do is add two lines to the VMware .vmx file for your VM:

svga.maxWidth = “640″
svga.maxHeight = “480″

These will restrict the display size for the Longhorn Server installer.

Once you’ve made the change, start the install, and everything should work. You can edit the .vmx file once you’ve complete your installation to remove the display size restrictions.

Simon Bisson

For us, a pearl is something that’s small, beautiful and desirable. For the unfortunate oyster it’s a a response to considerable irritation. It’s probably Microsoft’s success in the smartphone market rather than the lawsuit that counts as RIM’s irritating grit, but the Pearl is a suitably shiny response. It looks good, it has the features we want, it’s small and desirable - and it’s not hard to use, which is where some of the other shiny little phones fall down. Interface isn’t always Motorola’s strong point, for instance; in what you could call an admission of guilt, the latest RAZR phones come with a help system and Ron Garriques (president of Mototola’s mobile devices division) has promised to reduce the number of clicks it takes to get to key functions.

That hasn’t stopped the company selling 50 million RAZRs; as many as the iPod in a third of the time. Is that about form or function? If you look back at the history of Motorola (from the first radios in police cars to the first mass market television set to the radio that let Neil Armstrong phone home from the moon surface) and forward at their plans for seamless mobility, the current focus on desirable design seems something of a stopgap; a way of clawing back the sales the company missed when it lost momentum after developing the whole idea of cellular mobile networks.

It’s also an acknowledgement of the power of the consumer in what were once IT and business decisions. In smaller companies without a strong IT department, , the users choose the mobile devices they want to use for messaging and email and they tend to choose on looks, according to Nick Spencer of Canalys. That can cost the company or the user dear; a high percentage of mobile phones are returned as faulty because it’s just too hard to get email working. Nobody will reveal which phone manufacturers have the worst returns, but RIM seems to score very well, because the setup is either simple and streamlined for the user, or up to the IT department completely.

The look of the Pearl certainly competes with style icons like the RAZR. But RIM can’t sit on its laurels; one survey claimed that users don’t always find the interface easy to learn. The Pearl’s trackball is one response to the problem. Another is the pages of help and tips you have to sit through to set the phone up; most users will only see them once but many journalists have memorised the screen taps it takes to get a Pocket PC up and running and instinctively shrink from anything that smacks of My First Smartphone.

The Treo 750W is a very different beast. Bigger and heavier, not nearly as sleek and desirable - but it has 3G and some ‘it just works’ touches like a VCR-style interface for voice mail systems so you don’t have to remember that 3 repeats the message on Orange but deletes it on BT. A Google search box on the front screen, being able to just start typing a name to phone someone instead of pressing the phone button first; they’re little things that make a big difference to how much you like using a smartphone. If you wish Pocket PC was more like Windows Smartphone, this is the device for you. If what you want is Windows Smartphone with a keyboard, wait another week or so; smartphones are like buses at the moment.

RIM wants to fit into the life you already have, or the one you’d like; if you can’t leave the office at 5 without looking back, at least you can leave at 6 and finish your email on the train instead of staying past 7 and you have music to listen to on the way home as well. Palm just wants you to find its handhelds so useful you carry on buying them. Motorola has bigger plans, from cars that phone in a warning about road conditions to safety jackets that monitor vital signs for firemen and soldiers to a set-top box that pushes the programme you want to watch to your phone or the nearest TV set. 

Is it too soon to be strategic and look at the bigger picture beyond the handset? RIM and Microsoft see the bigger picture as being based on PCs and servers. Motorola wants to embed the phone in a lot more places than that, which means persuading a lot of people and a lot of companies to buy into the Motorola vision - which means a lot of people have to like Motorola. Being a byword for desirability will be a selling point for the consumer side of things. Being innovative gets you some new customers and some confused looks. But being reliable and easy to use never goes out of fashion.

Mary Branscombe

A recent article has caused more than a little controversy, as it seems to predict the end for Microsoft’s OS market. But read a little more carefully, and what you find is not a dire prediction of the end of the computing world as we know it - instead it’s an optimistic prediction of a world where the virtualisation technologies in the latest generation of desktop and server processors allow the development of modular operating systems.

We’re already used to the idea of information appliances. Your PVR, your iPod, your NAS box, and your firewall are all appliances. Embedded versions of Linux and WIndows Server allow hardware manufacturers to mix and match the components they need - so they can build systems that do one thing, and do it well. Take Buffalo’s TeraStation NAS appliance. It runs a Linux kernel, but the rest of the software platform is focussed on its role as a file server - with a simple web server acting as a management agent. Linksys does the same with its wireless routers.

So why do we persist in thinking of the PC as a monolithic entity? We even make them look like the monolith from Kubrick’s 2001. There’s one under my desk - a black slab of a tower PC case. It’s time to take a lesson from the mainframe world, and walk away from the monolithic OS.

Virtualisation hardware allows us to add in another layer of abstraction. We don’t need to think about the PC hardware when we write software. Future virtualisation tools will use thin hypervisors to manage software interactions with hardware, leaving operating systems to do what they do best: managing software. Why should an OS be handling a memory leaking graphics driver at the same time as it’s looking after a transactional database handling terabytes of data? Surely it’s your data that matters - you don’t want to lose any when an OS kernel faults in the middle of a transaction.

Isolation and abstraction mean that your OS can be broken up into vertical silos - perhaps a web server, a database manager, a file server, a security monitor. Each will communicate through a DMA-based virtual bus managed by the hypervisor, and each will operate in its own virutal box of memory. One silo fails, the rest carry on…

Microsoft is already thinking ahead. At this year’s WinHEC, engineers talked about how they could take advantage of virtualisation. Research projects like Singularity point the way to a loosely coupled OS future, and Steve Ballmer’s vision of a regularly updated OS (behaving almost like a Salesforce.com) will demand a move to a component approach.

The Internet’s model of a world of small pieces, loosely connected, will move into the PC. We’ll still have Windows and Linux and OS X and Solaris. They’ll just be different - taking advantage of secure operating environments, hardware abstraction, and memory isolation. Meanwhile, the same companies will also be selling hypervisors that will promise to run any (and all) of the new OS personality layers, controlling application specific OS silos, and giving us desktop PCs and servers that will be truly non-stop and very very secure indeed.

The end of the Microsoft Monolith? Yes. In fact, the end of all monoliths,

But at the end of 2001, the Star Child returns home to change the direction of human evolution. Tomorrow’s hypervisors will do the same for the way we build and run servers and desktops PCs.

It’s a Brave New World out there…

Simon Bisson

Later this month we’ll set off for a business trip that will be a mix of conferences, meetings, interviews and writing in hotel rooms. For several years I’ve travelled with the HP Compaq TC1000 Tablet PC which is a near-perfect combination of a lightweight slate (with a really comfortable pen that’s much more Parker than Bic) and a usable, detachable keyboard. I need a keyboard for writing articles, but for an interview where I want to be able to make eye contact and wave my hands around when I’m explaining my question, I want to be writing on my screen rather than lurking behind it. I can draw diagrams and lay out information quickly and easily. And I can take notes standing up, walking around an exhibition hall. A detachable keyboard means I can do it without my wrist buckling under the strain

But it’s time for a new machine and at the moment, I just can’t find what I want.

The main reason I can’t carry on with the TC1000 is that the Transmeta Crusoe processor was always underpowered and when I’m working on articles as well as taking notes, I’d like the power to run all the applications I need. I want to run 2007 Microsoft Office because even in beta OneNote 2007 is the place I take all my notes, whether I’m collating research from white papers and Web sites, scribbling down notes in a keynote or an interview or bashing out a to do list or an article structure; if it had a word count and the same AutoCorrect features as Word I’d probably switch from Word almost entirely. I keep the notebook on our SBS 2003 server and I can open notes on any connected machine, edit them when I’m on the road and automatically synchronise back when I’m on the network again. And I’ll want to run Vista; the enhancements to tablet PC alone are worth upgrading for as soon as the ACPI support for your tablet is robust. There are some very nice Core Duo tablets like the Toshiba M400 and the HP TC4400and many of them will move to Core 2 Duo soon. Multicore does wonders for handwriting recognition.

Battery life on the TC1000 wasn’t ideal - three hours when I’m recording audio in OneNote - but I could carry a second battery in my bag and have a full afternoon of working time. This time I’d like an extended battery that will give me 6-8 hours; many tablets have these.

But the problem I have is getting the form factor I want. My main portable notebook for the last five years has been one version or another of the ultaportable Toshiba Portege; first the 2000 and now the R100. the R200 is very tempting; fast, sleek, superb battery life, nice keyboard and very thin and light. If it was a tablet, even with a swivelling keyboard rather than a detachable, I’d be ordering one now. But Toshiba’s thinnest, lightest tablet is the M400 and that’s twice as thick and half again as heavy. Acer and HP have machines that perform well, but they’re not really ultraportable. NEC’s super-slimline slate doesn’t have the battery life, the clip-on keyboard - or a UK price.

I’ve experimented with pocket-size models. The Fujitsu Siemens LifeBook (the teeny P1510) is ultra-small, ultra-light and it has a keyboard. I could learn to cope with the pocket-size keyboard (I’ve written a thousand words a day on a BlackBerry keyboard when I was travelling around New Zealand), but I can’t cope with the passive digitiser screen; I need to lean my hand on the screen as I write to get anything legible. I’m a big fan of the Motion LS800, which has an 8″ screen that screams ultra-mobile PC, and the active digitser that tablets need. It’s the size of a hardback book so I can fit it on the table at lunch, but with the extended battery I can take notes and record my way through a three and a half hour keynote. There’s even a kickstand for propping it up when you plug in a keyboard, but you couldn’t use it on your lap on the train (where I’m typing this now on my R100). I want to try a Samsung Q1 in the real world rather than at a demo, but like the LS800, it would probably be a secondary machine rather than my only travel PC.

The LS800’s big brother is the closest machine on the market to what I want; that’s not too surprising as Peter Hunt from the TC1000 design team now works at Motion. The 14″ screen is a little larger than I really want for a portable. There’s no Core Duo version yet. And I haven’t been able to try the snap-on keyboard to see how good the typing action is. But the main reason I haven’t succumbed to the lure of the LE1600, is that at this year’s WinHEC I saw a model tricked out with a combination digitiser. Write with the pen and you get smooth accurate inking; and when the pen is in range, you can rest your hand on the screen without having any effect. Put the pen away and you can tap the OK button, drag a window around the screen or scroll through a Web page with your finger. Perfect.

We’ve had several years of tablet PC design now, and a decade of tablet design for custom operating systems before that. There are plenty of great ideas; the direct hinge that hides the keyboard on an Acer, the way you can scroll using the fingerprint sensor on the LS800, the magnetic catch on the Toshiba M400. But what about a really portable tablet? Here’s my shopping list:

- 12″ widescreen (go to 14″ and the weight goes up too much).

- Lightweight - 1.2kg with the keyboard fitted would be the same as the R100, but I could live with a little more weight if the slate was detachable (the TC1000 was 14.kg without the keyboard and 1.8kg with).

- Core 2 Duo - 1GB of RAM on the motherboard and a free memory slot.

- 100GB hard drive.

- Active or combo digitiser.

So, who’s going to build it for me?

Mary Branscombe

Simon and I are fans of Las Vegas - the city and the show. For one thing it’s where Simon was when I proposed. By text message (why yes, we are geeks). He was in a keynote with a Palm Treo 650, I was at home in London with a Blackberry. Where else can you see lions, eat sushi, watch a volcano, take a gondola down the Grand Canal, go up the Eiffel Tower, hurtle past the Empire State Building and finish the day by taking a limo into the desert? And Vegas is one of the most advanced consumers of technology outside the military and adult entertainment sites. Steve Wynn’s new hotel has a single sign-on system many enterprises would be proud of; your room key is your player’s card and you play the slots with it, swipe it in the restaurant and flash it at the blackjack table. VOIP phones in the rooms are reprogrammed with better services for the high rollers and you can plug your iPod into the HDTV screen if the Internet radio channels and photo screensavers provided don’t suit you.

Las Vegas the series has plenty of technology, although as usual the facial recognition systems the casinos really use aren’t quite as impressive as the ones they have in the fictional Montecito. Sometimes it’s very realistic; when James Caan’s character wants the surveillance tape from a supermarket and diner that will prove he didn’t run a red light he uses social engineering (implying he’s from a security agency in one case and outright bribery in the other). Sometimes rather less so; one much-hated character has recently been blown off the roof by the enormous sleeves on her kimono. But the Montecito is all about managing risk by surveillance; the security team watch out for problems and aim to deal with them without closing down the gaming tables. A parallel set of corridors and control rooms snake behind the public areas so staff can get around more quickly than the punters. And while the flashing lights and spinning wheels look very fancy, most of what goes on is hard work hidden behind closed doors.

It’s a bit like Intel’s new vPro platform. Security and network services running in the chipset should be able to deal with problems without you needing to reboot, but if a virus gets in the virtual network connection can close you down before you break the bank. Integrating services and virtualising the network as part of the platform in a standard way that all the management tools can work with is going to be more useful to businesses than the raw horsepower of Intel’s Montecito (the new dual core Itaniums that HP is about to ship).

Itanium has always been a gamble for Intel; it’s one of the most expensive and complex chips the company has ever made and Intel has spent a lot of money betting that businesses would move back end systems to a new architecture rather than relying on Opterons, Xeons and other x86 servers continuing to get faster and cheaper. So far the returns haven’t been good and the delay while Intel worked to reduce the power used and heat generated by Montecito hasn’t helped. Now Itanium has the same VT layer as the x86 chips. That doesn’t mean software doesn’t need porting any more; when you put in something that replaces the RISC server that runs your business or the time you’ve been renting on a mainframe, you’re going to be planning the software development as part of the cycle. But it does give Intel a better chance as it takes Itanium back to the roulette wheel. Next season, maybe the racks in the state of the art control room at the Montecito will have an Itanium running the Monte Carlo simulations on the risk they’re carrying.

I spent much of this morning talking to VMware’s Raghu Raghuram, its VP in charge of Data Centre and Desktop Platform Products. A long job title, but one that means he’s the man to talk about when you want to find about the future of the data centre - and of the server operating system.

We began talking about the recent release of its 3.0 product, a range of tools to help virtualise and manage an entire infrastructure, so that a single computing surface (made up of the commodity hardware that everyone from Dell to HP to Sun is now selling) could run an entire business. Add a new box (running VMware’s ESX), and it’s just more compute resource. You can give each part of the business a target utilisation - perhaps 20% for sales, 50% for web, 20% for ERP, 10% for everyone else - and if there’s spare capacity, hand it over to whoever needs it the most.

(As a side note - apparently only 5% of VMware’s code handles virtualisation - the rest is management tools…)

So what does it mean for the next generation of the server OS?

VMware’s virtualisation vision for the hypervisor is to use it to control the hardware, and then give the guest OS a set of common hardware APIs, no matter whether it’s an Opteron with ATI hardware, or an Intel - NVidia combination. This then leaves the OS doing what it should be doing, looking after the applications that it’s running.

Raghuram uses the example of VMware’s recent appliance competition as a vision of the future. Virtual appliances are customised OS and application packages, focused on a single task. The competition winners included a network monitoring tool and a simple NAS client. ALl had customised Linux kernels, along with only the components and applications needed to fulfil the task.

So why not do this for other OSes, and other tasks. VMware’s vision of the future is of hypervisors abstracting the OS from the hardware, and vertical silos of functionality, with componentised OSes handling specific functions in specific silos. After all, if you’re running a web server in one partition, you don’t need to have all the services that a database in another silo may use (and definitely less than a ERP system or CRM…).

So is the monolithic OS doomed? I suspect so. I’d even go so far as to say that Microsoft’s Longhorn Server will be the last traditional server operating system. The future will be componentised OSes, coming from new directions. Take a look at MSR’s Singularity project, and some of the other speculative OS research that’s under way. There’s a sea change coming, and you’ll need to be ready.

But whatever happens, it looks as though VMware will be ready. It’s got a vision of the future that’s exciting and enticing - but is it alone? What does Microsoft think, or Xen, or Parallels?

[Simon Bisson]