tag line

moving IT to the cloud with service not servers

Sunday, 24 September 2017

Chromebooks and Office365 - surprise!

Chromebooks can be a surprisingly useful tool in reducing the device management overhead for schools that are standardising on SaaS applications such as Microsoft Office365.

This is mainly due to the fact that Chromebooks handle imaging and updates in a completely different way to traditional Windows clients which traditionally depend on  a suite of local software such as WSUS, MDS and SCCM.

The Chromebook model can be a bit of a culture shock for Microsoft admins who are used to managing infrastructure and controlling every aspect of the update process. However once the basic concepts are understood, the initial scepticism is normally replaced with a sense of relief that one of the more tedious elements of desktop management, imaging and updates has at last been fixed.

So from a Microsoft sysadmin’s perspective what are the surprises when working with Chromebooks?

Surprise One : There’s no image.
Unlike a Windows client the local admin does not build, maintain or deploy a Chromebook image. The work of integrating, patching and regression testing still occurs but that’s all done by the the Google smarts. The local admin only sees a fully formed download. Each Chromebook is delivered from the manufacturer with a running version of the operating system (ChromeOS) which will automatically update to the latest release once the Chromebook connects back to Google. Behind the scenes each version of ChromeOS has a specific image matched to the hardware so everything stays in step with the model of the Chromebook. 

Surprise Two : There’s no patch set.
ChromeOS moves forward as a series of minor and major point releases. There is no concept of a security or application patch which is independent of a point release. This system avoids the possibility that a security patch or application upgrade could introduce instability into the system when combined with a specific OS version. For this reason there’s no requirement to invest in auditing tools to provide a patching profile of each client device. If a Chromebook is running version 61.0.3163.101 that’s all you need to know.

Surprise Three; There’s no gold build or extra security requirements.
As an administrator you are committed to move forward on Google's six week release schedule . There are some breaks and controls that you can use to affect the pace of the change but the direction is always forward. It’s not possible to hold on a gold image which remains static for an extended period and because security updates are integral to point releases they can't be side stepped. This ensures that all Chromebook’s are running with the latest security policy rather than an out-of-date set that's vulnerable to current threats. There no requirement to slipstream extra software into the build to protect against viruses and malware which, apart from the benefit of reduced licence costs, just makes things easier and more secure.

Surprise Four;  There’s no update scheduling.
There's no Configuration Update message or patching schedule because the update process does not impact on the user experience. The user does not see any spinning graphics, percentage dialogs or delays on boot. Each Chromebook has two copies of ChromeOS that act in a similar way to the “Last Known Good” on a Windows client, although the security model is entirely different. Updates are applied to the passive version of the OS as a background process. Once it’s complete and verified the two versions are swapped on the next power cycle. Because the updates can occur without affecting the running workload and from any location is there’s no real point in attempting to impose a schedule that has to be checked and independently managed.

Surprise Five;  There’s no profile management.
Chromebooks can run in an entirely ‘stateless’ manner. Each time a user logs off all personal is removed and the Chromebook returns to exactly the same state it was in before the user session. The user profile that’s delivered from Google is extremely light. User data files can be forced into cloud storage (Google Drive/OneDrive) so moving between devices is a seamless experience. There are however some operational benefits to maintaining a cache of local profiles but even here Google has your back. Each local user profile is protected by an encryption layer that only the owner of the profile can pass through and the Chromebook will automatically purge degraded profiles to free up space in a shared device deployment.

Surprise Six: There’s no requirement for any local infrastructure.
Managing a suite of Chromebooks does not require any local resource and for this reason is an ideal partner for a serverless approach or any strategy built apon SaaS applications such as Microsoft Office356 or Google GSuite.  However this exposes one potential weakness in the solution. In a Microsoft environment the on-premises WSUS server acts a cache for downloads and updates but no such facility exists for Chromebooks. So what happens to your internet bandwidth when 500 Chromebooks all decide to update at same time.

Fortunately this situation can be addressed in a number of ways.

First the admin console provides a option to stagger the updates over period up to two weeks so not all the Chromebooks will request the data at the same time. Chromebooks are also capable of updating through a peer-to-peer process although this function has a limited role on more secure wireless networks due to the reliance on mDNS as the discovery mechanism.

The most common solution for large Chromebook deployments is to use an edge security device that has a file transfer proxy function. Updates are one of the few Google functions that can operate outside of the https security envelope so the patch files can be easy cached using software such a Squid.  File integrity is maintained by the fact that every update is digitally signed by Google and this signature must be verified before the update is applied. 

So basically there’s far less hassle and and fewer things that can go wrong from the security and general management perspective. Because Google does all the background work it’s possible for schools and districts to manage hundreds and sometimes thousands of devices without being bogged down by the overhead of imaging and patching and, because it’s a serverless model, hardware dependencies can’t put a break on the deployment. 

Although a Chromebook will never run a local copy of MS Office they make great platforms for thin client deployments using the Citrix Receiver or MS Remote desktop client and of course the MS Office webapps.

Although you still need a Chromebook management licence and a Google GSuite for Education account it's role is reduced to that of an policy management engine for your Chromebooks as there's no actual requirement to take any of the core GSuite services such as GMail, you simply turn them off and replace them with links to Office365 and your SaaS applications.

You can even replace the authentication service with Active Directory by directing the Chromebooks to a local federation service (ADFS) if you have a strong desire to be presented by a familiar logon dialog.
Even if you're planning a Single Sign-On solution you still need to synchronise user accounts between Active Directory and GSuite for Education but there are tools available to help with this.
So as web-based applications become increasingly integrated with MS Office365 there are fewer reasons not to look at Chromebooks as the student device of choice. Try it out, it might be a pleasant surprise

Saturday, 12 August 2017

Going serverless with Microsoft

Over the last few months Microsoft have been developing a blueprint for a fully serverless cloud architecture based on Office365 and InTune for Education.

The individual elements for a serverless school have existed for some time but we now have a Microsoft strategy document that brings all the pieces together with a clear technical direction.

The document is updated regularly so there’s little to be gained from summarising it, other than to note it includes the two core elements mentioned above plus School Data Sync, One Note, Whiteboard and Teams while avoiding any mention of local servers, Active Directory and the System Center Configuration Manager (SCCM) management suite.

Sounds great, but how practical would it be migrate to this his model today?

First, it’s clear that the Microsoft vision of a serverless school requires Windows 10 clients in order to link into the security and management features of the Azure cloud based directory.

Therefore Step 1 is migrate all clients to Windows 10 and when that’s done you can move onto Step 2.  A full client upgrade program would be a good sized step for Neil Armstrong never mind an school with a mixed set of legacy hardware but currently it’s a prerequisite for a Microsoft cloud solution.

However, let's assume we’re already at Step 2. What other obstacles do we face?

The first is the same stumbling block that challenges other initiatives in this area  - how to support locally installed Windows applications ?

In this instance Redmonds approach has an advantage since we have a fully featured Microsoft operating system and the ability to deploy and maintain applications using InTune.

Things become less clear when we consider how well this model applies to shared devices in a teaching environment. If the toolset is fairly static across the user base it might be practical but if you have applications required for specific classes, students moving between computers and large installation packages being pulled across an internet connection, it could get messy quite quickly.

Strangely there is no mention of Windows 10 S in the document. This is the Windows OS which works exclusively with apps from the Windows Store and is aimed directly at  educational deployments.  This might be because the post is focused on a migration scenario but I would still expect a mention, if only to position Windows 10 S within the overall strategy.

Perhaps the idea is not present too many disruptive concepts all at once.

A school that has moved to Azure AD automatically gains access to Microsoft's ecosystem of Single Sign On (SSO) web applications. While this is mainly focused on the workplace the directory already contains over one hundred web resources marked for education including well known names such as Khan Academy, Discovery Education, My Homework, Edmodo and ClassDojo.

Once a school starts to take advantage of the rapidly evolving pool of SaaS applications with built-in SSO  the deployment issue disappears and Windows 10 S becomes a good news story for everyone, with perhaps the exception of software houses still shipping an .msi file on an annual release cycle.

Locally installed applications of any type do not work well in shared device deployments that require a degree of differentiation. Until 1:1 rollouts are commonplace, SaaS will win out every time and a cloud based directory with integrated SSO can only accelerate this process, unless of course your students are really looking forward to next years release of SameOldProg V8.

It’s also worth examining how the integration with the Azure directory will be managed.

Third party software such as classroom control, content filtering, payment schemes and print management need to read data from the user directory. In the future this will be in the cloud and not on a local domain controller. All this is fine except that Azure AD does not support LDAP or Kerberos, the two access methods that every management tool sold to education in the last twenty years expects to use.
Azure AD has it’s own convention (Microsoft Graph API) which is better suited to modern internet protocols than either LDAP or Kerberos.
Therefore vendors of firewalls and content filters will need to embed support for this new directory source before schools can consider moving to the cloud.
In a completely unscientific survey I recorded the Lightspeed content filter as capable of working with an Azure directory.  If you know of any  others please let me know and I’ll compile a list.

Wireless might also have a problem with a Microsoft serverless school. A common security method uses the RADIUS protocol to query group and user information and in the past this was normally provided by a local Windows server that accessed information from a domain controller.

The problem is - not only are we a server short, we don’t have a domain controller either !

Anybody know of any vendor initiatives in this area ?

Microsoft and Google are going head to head for this market and now both vendors are essentially proposing the same serverless approach which will only drive innovation at an even faster rate.

In the short term Microsoft has the advantage because they are are the incumbents in this space and now have an offering which appears to match Google GSuite for Education in certain areas.

However these are early days and few would describe the Microsoft strategy as fully defined offering. A number of roadblocks remain but over the next few months we should expect new features to emerge at a rapid rate to fill the gaps. Overall the outlook is pretty exciting and whatever your technical point of view, schools will benefit massively from the one upmanship as the two tech giants slug it out.

The real challenge is convincing education to assess the alternatives with an open mind and then invest some time in constructing a development plan that will take advantage of this unique opportunity to get things right.

Friday, 4 August 2017

Why BYOD could soon be BYOC.

Bring Your Own Device (BYOD) has always been an attractive idea for education.

The possibility that students could use personal devices in a learning environment without the school having to make a financial investment sounds beguiling but there are some fundamental problems that have never really been overcome.
  • How to integrate a variety of devices, all with different capabilities into a lesson plan.
  • How to securely manage school data on a range of devices.
  • How to onboard the devices onto the wireless network without a management overhead.
  • How to provide secure web filtering without additional licencing costs.
  • How to answer the question “It’s my device, why can’t I have Facebook”?
Unfortunately the big advantage of BYOD is also it’s biggest weakness.

Because the device remains the property of the pupil or a parent/guardian there’s a reasonable expectation that, outside of school hours the device could shared with other members of the family to access Facebook, eBay, NetFlix and various game sites.

Of course this creates a host of e-safety issues that’s almost too long to list. Therefore the cautious response is to apply the school security policy at all times even though for the majority of the year the device is at home and belongs to somebody else.

So in a BYOD EDU environment how do you answer the question;

“It’s my device, why can’t I have Facebook” ?

Let’s look at this from another angle. Ideally, how would you fix this problem?
  • During school time the device is under management control with all the usual policies applied. 
  • Outside of school hours the device takes on a personal policy which allows access to Facebook.
  • The two worlds must never meet.
Elements of this approach are already possible using web filtering rules that can be updated based on a schedule but this doesn’t address the fundamental problem of device security.

When the device is under personal control how do you ensure security isn’t compromised by malware, keyloggers, trojans, inappropriate software or images which are then brought into school and propagate across the network ?

This problem doesn’t rest with the management platform, it lies with the nature of the user device.

The device has to maintain a set of isolated user profiles without any possibility that information or activity could bleed from one to the other. It would also have to have built-in security that would ensure that the operating system image was clean and verified as secure. Once you throw in the requirement for centralised web based policy control it's clear that the device you describing is a chromebook.

Let's imagine what this new form of BYOD, admittedly limited to Chromebooks, might look like.

Each pupil would need to bring a Chromebook to school. It could be an existing device, newly purchased or sourced through a payment scheme. Any finance schemes would be independent of the school because the device remains the users personal property at all times.
Students would be able to choose a form factor that best suits their requirements (touch / size / price), shop around for the best deals and personalise them as much as they like. I suspect the ‘missing key’ problem will disappear appear overnight.

Personal Chromebooks could be enrolled into the schools Google organisation using the home broadband or a simple phone tether, it’s really not that difficult. The school will be buying a batch of non-recoverable Chromebook management licences but this is small cost when you consider that this would enable a 1:1 programme with very little management overhead.

The big selling point of this approach is that during school hours a policy is applied that restricts access to all the fun stuff and locks down the device. Out of hours this restriction is lifted - hello Facebook.

This could be done in a number of ways. It could be as simple as enabling guest access or lifting the restriction on organisational only logins. As the schedule moves back into school hours the standard policy is re-applied.

This doesn’t mean that the student could access social-media using an organisational login only that the Chromebook would allow a logon using a consumer account to which the filter does not apply. Nor does it mean that an out-of-hours policy would apply to all devices. It could be operated as an opt-in scheme that requires parental consent or be subject to an acceptable use policy
Integrating personal Chromebooks into the classroom is easy because although you might have 57 varieties they will all be running the same OS version and all have the same basic capabilities. Because they remain personal devices the school doesn’t need to get involved with insurance or warranty repairs, although a loan pool of utility Chromebooks all covered in a massively uncool school laminate might encourage careful handling and long term memory.

Expensive trolleys aren’t required. Ad hoc charging could be problem but only until USB C becomes commonplace. There’s no issue with respect to software licencing as this likely to be SaaS based and linked to the organizational account or installed within the user's encrypted profile.

Sounds interesting ?

Unfortunately none of this is possible because the basic mechanism to relax the Chromebook management profile on a timed schedule doesn’t currently exist.

However ChromeUnboxed recently reported a new commit to the Chromium repository described below.
“Allow unrestricted using of parent-funded Chrome OS EDU devices (Chromebooks) that are managed by school, while the device is not at school (“off-hours”).”
While we are unlikely to see this feature until ChromeOS V62 the fact that it’s even in the pipeline is a significant development.

Currently there is little indication how this might work other that the fact that it uses an "Off-Hours" flag in the device policy but it’s clear that this initiative could accelerate the drive towards 1:1 devices in education and be an important new way of getting Chromebooks into the classroom.

BYOC perhaps.