Thin or fat – when is a cloud not really a cloud? Steve discovers two companies with very different views of the future.
Thin or fat – when is a cloud not really a cloud? Steve Cassidy discovers two companies with very different views of the future
This has been a month of battles with modern business platforms that turn out to have ancient and surprising underpinnings. It’s long been a tradition in IT that no matter the roots of your system, it will always be described in the most modern marketing idioms, even when circumstances prevent most of those idioms from actually applying.
Yes, I typed that with gritted teeth. I have a couple of very new networks, commissioned in the final months of 2017, and both clients are terrible practitioners of what I call “top-down diagnosis”. They believe that the more important the job is to the health of the business, the faster and more accurately the IT infrastructure will perform while running those jobs. They frequently exclaim, “oh don’t run slow now, I have a meeting!”, or like to delay any possible fix by explaining in excruciating detail how important their job is.
Trying to tell them which part of the system is misbehaving is almost impossible, because it’s only their viewpoint that really matters. And, of course, no IT worker has ever been in this situation before, or understands things such as deadlines or profit motives. They’re also quite sure that new kit is inherently unreliable and will have “bugs ironed out of it” only by the courageous and unstinting testing by real users… can you tell I find this irritating, yet?
Top-down diagnosis requires that the most complicated and innovative parts of a system are tested when there is a delivery issue, before the simpler components are looked at. So when a cloudresident accounting product slows to a crawl and finally logs the user out, this type of user will reboot the local server, the PCs, the VPN router, and the fibre modem – re-logging in every time – before they call the cloud supplier and ask if perhaps something was wrong with their session.
It’s important to encourage people to think about the right experiment, too. Normally, I don’t bother those people whose understanding of the world is filled with malevolent spirits, cartoon physics, animism and worship. However, when they start exporting all that in a request to a techie to make it all better, to my mind they become fair game. A case of “the customer is always right, until they try to think like the supplier”.
So with this incident, I sat at the next-door PC and roundly rejected all appeals to reboot the server farm, unplug the router, turn vegetarian, change electricity supplier, or any other absurd responses to a basic error. My defence consisted of continually flipping around YouTube videos at high speed, copying files across the LAN, and generally refuting all the incoming hypotheses.
This is mostly because the “cloud service” this guy was using was a Citrix session. So far as he was concerned, because it wasn’t in the building, it must be “In The Cloud Somewhere” – and that made it inherently good. Yet, quite plainly, Citrix Receiver was failing to
“Database woes that take down multi-user Citrix servers are more likely to arise from abuse than over-use”
reconnect to the service: he had no idea if his accounting session was open, closed, incomplete or crashed.
Don’t get me wrong. I like Citrix, and it behoves us to remember that it’s the daddy of remote desktop platforms. The acquisition of Xen produced one of the sharpest 90-degree turns in the history of computing. To this day, Citrix is unapologetic about its specific, niche-orientated approach to cloud software provision. And it hasn’t deserted its traditional multi-user remote access marketplace, either. Full marks to Citrix for all of this stuff.
However! What Citrix is presenting is mostly about Windows servers, and remote complete VMs of individual PCs. This tempts suppliers and customers to do a “lift and shift” – pick up the complete software manifest of a business and shift it all onto a cloud-resident hypervisor platform, offering secure access by way of Receiver. To the naïve user, they click on a cloud shortcut, they get a window, they do their work. See, cloud! Zero server! Wooo!
Um – actually, no. It’s certainly possible to run a cloud service on Citrix – but a Windows server isn’t capable of doing the things for which cloud sessions are famous, especially not when it’s configured to provide access to a software package that has no understanding of cloud itself. So my client had bought into this idea of cloud computing, skimming lightly over the bit labelled “legacy support”, and focusing entirely on all the sexy stuff about scalability, reliability, and variable-cost-based – instead of capital-cost-based – computing.
Not one of those features of cloud is relevant to what this guy is doing. He wants to keep his accounts app open all day, typing stuff in and getting reports back. The software opens many windows, uses multiple databases to represent multiple separate companies, and doesn’t like it if you leave transactions open.
To achieve magical cloud scaling with this kind of legacy product requires some serious infrastructure skills, with multiple servers in a farm, centralised user management, tiered storage with virtualised connection – and the evidence suggested this “cloud platform” had none of those. The main giveaway was that the provider was asking its clients to hang back from a pending update, because some of them weren’t ready for the changes to their accounting practices that taking on this product would create.
So what we were looking at wasn’t cloudy at all. In fact, it wasn’t all that Citrix-smart, either. It was a Windows server remote-working environment, with secure connections from something like 50 separate smallbusiness users. It had to be that way because of the stick-in-the-mud attitude of the developer: if you wanted the full-fat feature set then you better stick with the Windows version, is its position. This made a hosted version inherently unscalable, and not that good at segregating the impact of one user session on another – hence the crashout.
This is a hoster abusing its privileged position within a market of small businesses. Rather then spend its money on a properly configured Citrix deployment, it threw it together with about 5% of the feature set being used. For its customers, this meant several retrograde steps in overall reliability. A small florist might only clean up his accounts at the end of every month; a garage with 50 staff might be doing the same job every day. The kind of database woes or misuse that take down multi-user Citrix servers are more likely to arise from abuse than over-use. But top-down diagnosis rules meant I had to sit there for some time while it crawled and crashed, before my client decided to ring the service provider.
Once the distant hosted server had a classic Windows reboot and repair process, normal service was resumed, followed by a rather committed conversation on why there was a need for a cloud solution – and why they shouldn’t allow their local installations of the accounts package to fall out of subscription just because the other option was “in the cloud”.
This is one of those very difficult jobs, where it’s all too easy to get drawn into the fine diagnostics of what’s really going on at the far end. And if you get that diagnosis right, you’ll have successfully avoided being paid by 50 people, all of whom are impacted by the poor implementation of the hoster. Nobody should be supporting a deployment such as this, which manages to make the worst use of an excellent product in the dumbest possible configuration, because to keep it running with ever-smarter collective fixes simply pushes back the day when the right solution arrives.
Playing in the big boys’ game
Right at the other end of the scale, I found myself in Madrid in January (pretty much as cold as London, I must report) with Software AG. It was mostly making noises about a much deeper investment in the IoT marketplace by way of a not-reallytakeover of Silicon Valley startup, Cumulocity. I like Software AG events: it’s easy to get to the hardcore techies; there isn’t a lot of software business rhetoric to wade through; and there’s always a short route to a practical justification for what’s going on.
Not so much this time. Try as we might to make sense of the takeover, it just kept eluding us. Cumulocity will be gifted with a whole lot of sales and support staff, beefing up its ability to sell in markets where it hasn’t before had a strong presence. Most of these come from existing Software AG offices and functions. Yet Cumulocity’s own pages make it read a lot more like a cash injection. What on earth could be going on here?
The hints were pretty sparse, but I believe I understand the logic now. The whole purpose of taking Software AG’s IoT business and putting it inside the Cumulocity business unit is, it said, about being able to pitch for larger IoT projects, in a market projected to show 70% growth in 2018. Yes – 70% growth. Whatever the size was in 2017, measured in dollars or in number of devices brought online, there will be a little over half as many again during this year. That’s a lot of
growing. And it seems plain this will come from businesses that aren’t exactly a mom-and-pop shop doing e-bike hire (thanks to IGEL for that particular worked example).
Working on projects for large multinational companies is a bit of a deal with the devil. They’re not here to make you especially rich; they’re here to cut the deal as close to the bone as they can, without actually crippling themselves. Crippling you is okay, because they’ll be able to hand most of your materials over to the bidder that lost out to you in the first place.
There’s probably a management treatise somewhere that relates the average size of IT contract to the likelihood that one party or the other will resort to legal action at some point in the relationship. While I can’t find an easily digestible infographic for this, I’m quite happy to assert that if you think you’re going into a business sector with explosive growth in it, and likely risks of a lawsuit even in the course of a successful project, then it makes sense to take the risks associated with working in those fields and move them off somewhere that won’t put your main brand or main bank account in peril.
This isn’t a snide remark, incidentally. It’s a rather unexpected outcome of the fact that IoT has become something of a fifth wave in the big-time world of global business. It’s another try for IT businesses to get inside the hearts, minds and wallets of their largest customers. It won’t be a smooth ride, and not for the usual reasons that surround software houses. This is more a matter of who’s the best diplomat than it is about who’s the best coder. Unfamiliar territory – but definitely exciting stuff. Where else have you heard of an IT sector with this amount of growth in it?
The key to thin client computing
A bit of a snowy ride around Europe to start 2018, hopping from plane to plane to catch up with various bits of the tech sector. I was especially intrigued to hear from IGEL, which is hardly a household name when it comes to thin-client computing. This hasn’t made any difference to its outlook: while most of us have been unconcerned by the thinness of our clients, IGEL has been beavering away, strengthening the range of its activities in this odd little backwater.
Not without good reason, too. It’s odd to put the thin-client world alongside others that have been burgeoning in adjacent sectors. Most people wouldn’t like the idea of coughing up more for a thin client than for a business-grade PC, yet think nothing of spending rather more than either on iPads for staff. Similarly, there’s a lot of lip service paid to the idea that IoT deployments need to be based on new, often enormous management and monitoring software. Yet, typically, the work required from an estate of thin clients comes out looking pretty similar to that generated by even thinner IoT endpoint devices.
Understandably, this presents IGEL with diversionary temptation. Could it make it big in generalpurpose IoT device management, which is, if anything, a subset of what it’s been into with its own devices in the thin-client continuum? Or is that diluting its proposition in a market that’s been threatening to burst through for several decades now? (Yes, it’s decades. I got several drinks out of several exhibitors by mentioning that I reviewed Citrix Metaframe in PC Pro around 1998...)
The oddness of the market is driven entirely by the demand for managing what can often seem like the unmanageable. Thin-client computing was all about reducing the chances that people would keep their holiday pictures on their PCs, or be so lazy with their clicks on embedded links that they catch something nasty. But the PC business – including Apple – just kept moving the goalposts.
It’s almost insane to try to use a Windows UI on a remote desktop server on a tablet device of any kind, and Retina screens make the problem 100 times worse. On a Retina screen, my thumb covers both the OK and the Cancel button on a RDP session to a remote host.
Nonetheless, the momentum of the tablet in business is such that there’s considerable pressure to not use thin clients coming “bottom up”– users who just want to be allowed in. The main successes in thin client are “top down” – not in terms of diagnostics, but rather in terms of who decides what happens to a business’s technology investment.
And top-down planning is a lot more productive than top-down diagnostics! The best story IGEL had was told in secret, slightly; not up on the stage in the keynotes. One senior chap from IGEL mentioned that the company has a product that’s just a USB stick. It will boot pretty much any nasty old lump of iron into being a fully manageable thin-client workstation. Including, rather amusingly, competitors’ thin clients. Somewhere in the USA, there are a whole lot of Wyse terminals – longstanding, capable devices, to be sure – equipped with the IGEL USB key. From the point of view of the network operations centre, they look and feel just the same in the software control suite as the newer deployments of IGEL boxes.
Of course, this isn’t quite as heavenly as it might sound, because the whole idea of a thin client is that it shouldn’t require lots of attention in a regular working week. The action in thin clients is mostly up in the server farm, with a side-order of endless fiddling about with printers. Nonetheless, the idea of being able to make every machine – fat or thin – in your entire network look and feel identical is immensely appealing.
“Multinationals will cut a deal as close to the bone as they can without crippling themselves”