Virtualising Everything in a Data Centre

Looking in to virtualising everything for some clients and moving their desktops and server in to the cloud.

Every person would dial in from where they are to their "Desktop" and do their work.

I've had a quick look around and found that Citrix and VMware seem to have varying offering's and they state that remote users can dial in with ease but not much is mentioned about performance if all users are in a different physical location to their desktop and servers.

Just looking for thoughts and opinions at this stage.

 

Prepare to piss off your users.

 

What's the rationalization for this?

 

A number of things are being considered:
Desktop Provisioning
Reducing Capital Expenditure and Converting it to Operating Expenditure
Uptime
Storage
Centralised Management
Distributed Users - With a number of different environments
Snapshots of work environments

 

Out of curiosity, can you elaborate on how this move to the DC will achieve a higher uptime? The others I sort of get but this one has me stumped.

I don't work in desktop/server land so it's not a loaded question.

 

Remote users currently rely on single adsl connections in the office and non-redundant servers.

 

My biggest problem with uptime is that if you have an issue with your desktop virtual platform, everyone loses out. I've heard of at least one example where a company which was using virtual desktops had an issue which caused everyone to be unable to use the desktop. They didn't have backup "fat" desktops either, so no apps would run, and the company (a fairly large one) was at a standstill until the virtual desktop infrastructure was fixed.

I love the idea of a desktop I can use at any location, and it has all my applications and performs similarly to a local "fat' machine, but I'm yet to see one which meets my expectations.

 

Not necessarily.

Depends on your VM arrangement.

But most of the time when people are doing this kind of thing, its because they want to save money, and to really save money, you end up skimping on everything - which in turns lends to a really shitty, over-subscribed deployment that runs like shit.

 

Exactly, there is a reason people left Mainframes and went to fat client systems.

Unless your users are lightweight (they use a single web app, maybe a little bit of MS Word, some light smh.com.au browsing) a Virtual Desktop infrastructure is going to struggle with a none generic environment. It's also a very expensive operation but on paper it looks very cheap from day 0.

VDI is a beast that can grow out of control very quickly once you release it into the wild (it's cheap initially, looks great for DC consolidation, energy footprint etc). But once your users need the power that only a dedicate system can provide (especially engineers, or power users, even watching YouTube can be a slow struggle) you will need to start dropping big money on specialised Server gear (storage [IOPS], Network, Memory, CPU) to get them to a decent state.

Your best bet is to find the "select" users who want this feature and provide them with a method of reaching their desktops instead of moving the whole company (if they don't all want this feature).

As said above, VDI is a substandard Desktop experience (if your users think their systems are slow now, just wait )

P.S. this is the same argument that everyone had for blades and we can all see how well that went.

 

Also depends on your patching, upgrades and a bunch of other stuff even if you are in an environment with both prod and dr infrastructure. An event where you patch all servers both in prod and dr, or you have a single point of failure with both could lead to a very bad situation. A virtual desktop environment needs to be properly planned and executed as it does introduce more risk into the organisation.

Also need to look at things like copying files. Going from a USB to your virtual desktop involves copying across the WAN in most places. So if your site is connected via ADSL to the head office where your desktop resides, you can easily max out your link. flash also tends to run like shite on virtual desktops, so you can count out a lot of the "nice" web interfaces.

 

My girlfriend works for a company where this is implemented poorly. They experience down time daily and simply loading Word is a, go make a cup of coffee, kind of deal.

Not to say your system will be like this but its easy to skimp on the costs and then not listen to the "whingey users" when they complain they can't send emails fast. Its costing this business literally thousands of dollars daily when their over subscribed system dies. I know quite a lot of talent has left the company with the stress of just typing up a document.

Its hard to see her come home nearly in tears with frustration and as an IT consultant I can do nothing to help but tell her to find a new job with better IT.

 

If your doing it as a cost cutting exercise, your planning for failure.

Be careful, as most I.T. Managers will dress up this in a thin vale of "easy access" & "computing everywhere", If this is true tell your project manager to take your current Desktop budget (times it by 5) and that should be the project budget. If anything less... return to the DC, lock the door... Cry loudly.

 

We have a 95% Virtual Environment.... but we didn't do it cheap.

 

What's the link like? That's your first issue.

 

That was sort of what was rattling around my brain but I couldn't put it to words. Unless you shelled out on nice everything-else-around-your-virtual-platform, aren't you at the mercy of:
- data links (sizing, latency, etc)
- yet another provider that has downtime when you really don't need it
- stability of your virtual platform / experience of your virtualisation/network/security platform guys to not break stuff
- your DC provider not being idiots

I've seen a company who had most of their workforce remote, all run on non-virtualised Citrix, wyse terminals and adsl and they were mostly fine, but I've also seen it from my own employer who trialled this (except hosted on site, not in a DC) for a small user base and it was backed out pretty quickly when they realised it wasn't as good IRL as it was on paper. No judgement on the theory of the plan, I just assumed someone fucked something up internally.

Anyway, as I say, I don't do servers/desktops for a living. I'm sure it can work, but as others said... only if you don't get all tightarse on the other stuff when you see the bill for the virtualisation platform.

 

Oh wow, and you think that putting that single ADSL connection in between the users and their servers is a GOOD thing???

Sure, get rid of the shitty old white boxes with exchange '97 (or whatever, you get the point), but implement a solution that results in what you're trying to achieve. All this talk about shifting costs to opex sounds like the accountant/business owner doesn't want to spend any money and thinks cloud is cheap.

 

Lots of interesting points. Just a few add in's not looking at this as a cost cutting measure. We know it is not going to be cheap and are happy to invest in the hardware and software to get it done.

Also aware that link speed will need to go up dramatically. Fibre is what we're looking at. Just don't know how much bandwidth would be required on a per user basis.

 

What sort of workforce and how many staff are you talking about?

We manage a number of clients who access everything via centrally hosted servers and they've all been fairly happy with the solution. They are all in the SME environment and have many remote staff, which is why the solution is a good fit.

However, the single biggest complaint is the reliability of the end user's internet. This will be your single biggest bugbear, without reliable internet staff can't work.

 

Another thing to think about that some often forget is printing, and don't size for it or don't have appropriate QoS to prioritize it over the real-time required citrix traffic. And have the whole site go down every time someone prints a document.

It also comes down to your applications as well, things like MS word are so much better than some custom application built for your company that thinks its a great idea to re-draw the entire screen on every click.

 

i wouldnt be going for it unless you have fast reliable internet (and maybe redundant links). Hear way too many times that people move certain things to the 'cloud' and all goes to shit because of connectivity issues.