I recently attended the VMWare vSphere 4 launch, where the presenter said "We are trying to make the operating system as unimportant as possible." I can see his point, with virtual appliances and virtual infrastructures like vSphere, how long will it be before nearly all applications are shipped from the vendor as virtual appliances?
And who will install security upgrades, configure the stuff and integrate it all? Either this is done by the customer, then the sysadmin will be needed anyways, or it is done by the appliance provider, which means he'll charge a premium and thus limit his market spread.
Next problem is that this will only work when the appliance targets the customer's virtual machine architecture. There are currently many more VM architecture than CPU architectures.
I don't want to get started on the security issues of allowing an external entity into the network.
In the end I believe that'll will go the same way as SaaS: big hype, some very nice niche products and no significant immediate market impact. People will have to explore what will work and what not first.
Quite on the opposite spectrum, I see a proliferation of installed OS, since virtualisation makes it very easy to have an OS-instance running per service (web server, database server, load balancer) and have those easily replicated as needed. This added complexity leads to the deployment of automated management tools like puppet which, like all automation, are a much bigger thread to the "common key-presser and mouse-pusher" through replacing manual labour with higher-value programming tasks.
I'm with everyone else who can type faster than I can on a Monday morning...
The OS is not going away nor is OS administration. There will always be some common platform that applications are developed on top of, that is the purpose of the OS. Even if / when cloud computing takes off larger companies are going to demand internal cloud services in some form another due to security implications. There are too many vulnerabilities at the OS level for larger and/or conservative companies to trust to third party OS admins.
To be quite hones, I don't see it happening.
Also, the notion of cloud-computing itself has some major security concerns by virtue of its design, and is not applicable for an entire host of applications. For example, most companies would have valuable/sensitive data that they would not like hosted with some "trusted" party.
This is all of course my subjective view on things and certainly not a definitive answer :).
Managing virtualised platforms makes huge sense but also requires damn good tech skills in all involved, whether that's design, planning, implementation, support, capacity management etc.
I love VMWare but they're still a company full of salesmen so they will come out with sweeping statement like that, bless 'em :)
Virtualization is definitely a growing trend and not just in OS virtualization but in storage technologies as well as networking. But none of these technologies will replace a system admin, storage admin, or network admin. It's just the tools of the trade.
Virtualization will lead to containerization of services as virtual appliances to some extent but I'm not sure how prevalent this will be in an enterprise infrastructure. These services still have to be configured and maintained as well as the OS that it's running. Virtualization last time I checked didn't get rid of security patches :). And it's adding a layer of administration and security with the hypervisor and virtualization layers.
I think our jobs are safe for the foreseeable future.
Virtual appliances will still require maintenance, the way physical appliances do. We've looked at and used various appliances over the years (firewall, spam/virus filter, NAS storage) and they all require maintenance - checking logs, performing updates, etc. For example, we looked at Windows Storage Server a few years ago but quickly realized that it wasn't going to be any easier to maintain than setting up a server and configuring it just to be a file server. If anything, relying on a an appliance (virtual or physical) from a small company makes less sense for mission-critical functions than running another real server.
Virtual servers make a lot of things easier, particularly things like backups, testing and swift deployment/redeployment/cloning, but they're still servers, running OSes, and they'll still need admins to admin them. Most of the usual maintenance tasks still apply, as in Ward's post above, and there are also a bunch of new sysadmin tricks that I've learned specifically for virtual hardware; which bits of hardware need changed when cloning servers for clustering, for a start. 3 servers, all of which have the same MAC addresses? Not pretty.
how long will it be before nearly all applications are shipped from the vendor as virtual appliances?
It'll happen, but only when patches and updates are hardwired into the appliances. When you start shipping an entire operating system as part of your application, the security maintenance environment gets a lot more complex. To some extent this can be automated, sure, but it also has to be flexible enough to fit corporate security policies. Some shops need to have security patches posted within 48 hours of release and can't tolerate 14 days of QA on the part of the vendor to make sure it doesn't break the appliance. Other shops are just fine with a 14 day lag between patch-post and application.
We're still quite some ways away from this. What we WILL see is applications that already are shipped as appliances, start shipping as VM appliances. Additionally, we may see some applications start shipping a VM as a delivery option, along side traditional software sales, to accommodate shops that'd really like to cloud-source their stuff but need either more throughput than their internet connection can handle or have an on-site requirement.
I don't think that "make the OS unimportant" is the same as "make the OS go away". Clearly, there has to be some software that is telling that hardware what to do and otherwise managing the system. That (by definition) is the OS, and it can't go away.
Rather, as I see it their idea is to use the virtualization layer between OS and hardware to allow improvement independent of Windows (or any other OS). The concept is (I believe) not to eliminate Windows, but to remove Windows as a bottleneck to improving performance, reliability and manageability.
Virtualization has been around since the time-sharing mainframe days (~1972). Just because it exists doesn't mean its going to make anything in particular go away. If anything, it increases the value of the sysadmin who understands how it works, how to make it fit business need, and how to automate deployment on virtual platforms.
Making the operating system irrelevant doesn't make the sysadmin irrelevant. You still need someone to handle the muck, and no matter how much abstraction is layered on top of the platform, you must have people who understand the various layers. I've seen an increase in the number of virtualization platforms that have an API. This means sysadmins are going to become programmers in their primary duty; they will program infrastructures with these APIs rather than manage individual components with "glue" like hacked together perl and shell scripts.
If you're a system administrator these days and you don't know how to program in any language, you need to learn. The future of your career depends on your ability to stay current with the programming interfaces to various virtual platforms and systems automation frameworks.
(Links in no particular order, pick your favorite platform)
A virtual appliance still has to run on something, and even the most basic NAS box has an OS beating at it's heart.
I can potentially see a future where certain boxes in more specialised roles have everything controlled via hardware chips (I'm thinking PlayStation but on your server here), but there will always be a requirement for general purpose machines too. At the end of the day, this reminds me of the bright new future that Java and .NET promised: all well and good for line of business apps, but who's gonna write device drivers or a kernel using them? Same principle in operation here.
Of course, the VMWare folks have probably the most - shall we say - "persuasive" salesmen in the market - at least outside of the high-capacity-toner-cartridge area - so anything they say like this should be taken with the appropriate amount of salt.