On ‘the cloud’ – Why sometimes it just doesn’t add up

I got back from TechEd last week and had an absolute blast! The Cireson team and I set up an amazing booth, had some excellent presentations from some of our partners, and we, as a company, got to do some team building for the first time ever! Nothing like being able to shoot your coworkers with paintballs to build team spirit 🙂 I think I only cursed out my boss, oh, 100 times or so. Thankfully, I’m still employed.

Of course, Microsoft’s big theme this year was ‘Cloud.’ It was Azure this, or Office 365 that, or “look what you can do with our cloud” again and again and again. Microsoft’s ‘hard’ tools, including a good chunk of the System Center suite, was left to play second fiddle to whatever was being offered via a cloud subscription model.

Now, I’m not outright opposed to the cloud, at all. I think there’s some awesome uses for it, but I had some really interesting conversations with a few people, and did some research of my own, that really puts the whole cloud vision in perspective – and I think it’s perspective worth sharing.

One of our developers lives in New Zealand and loves it. When he wasn’t being the butt end of one of our jokes (“Hey, whose that guy with an Australian accent over there!”), he and I had some awesome conversations about food, life, and of course, technology.

Somehow, the topic of internet access came up, and I expressed my love for my 100mbps down/ 5 mbps up connection that I get here in Baltimore, for a decent price. My Kiwi friend’s jaw nearly dropped. His connection, and keep in mind he’s one of our developers, is 500k down, who knows what/up, and is capped at 5GB per month. Sure, I get it, he’s on an island, but New Zealand is far from isolated compared to some of the markets Microsoft wants to get into, and if you’re using a cloud solution for resources, I can see that not only will your connection be spotty on a 500k link, but you’re apt to run up to that 5GB bandwidth cap really quickly! I’m sure businesses in NZ have better connections than a lowly developer, but putting all your eggs in a ‘cloud’ basket seems like a rather expensive, and potentially slow or limited, solution to your IT needs. It’s especially terrifying as we see Microsoft move more of their offerings to cloud only (PowerBI anyone?) with no on-premise solution.

Let’s head to another island, this time, one rather well connected, Great Britain. Another one of my co-workers, the ever-so-talented Shaun Laughton, happens to live on this very island! He joined in on this conversation and lamented his own internet situation – that he can see his local telecom box from his house, and yet his internet speed was only moderate at best, expensive, and had a data cap. If he lived in a major city, say London (where Google happens to have a massive campus with super-fast internet) then internet access would be cheap, fast, and without a cap.

This scenario isn’t uncommon and puts cloud solutions in a really interesting situation. For people in very well connected areas (read: Urban, 1st world) then cloud makes some sense. Why own the infrastructure when someone else can do it for you, and then just lease the time and resources you need? In this same instance, Microsoft’s own direction makes a lot of sense, as they are positioned to provide the *best* platform for their products in the cloud, and can really benefit from the subscription licencing model – hopefully attracting smaller customers with more reasonable, monthly or annual pricing.

As soon as you leave one of those major cities however, this plan breaks down. Who is going to depend on a cloud service when it’s going to eat at their monthly data cap, and even then, not be accessible nearly as fast as a local server or instance of a software application would be? For these customers, not only does the cloud not make sense from a costing perspective (as relying on it would require multiple, redundant, unlimited, internet lines, likely costing a fortune themselves) but their users are far less likely to have reasonable access to high-bandwidth connections from outside the office, thereby breaking down the ‘work anywhere’ principles that cloud relies on.

One final example that I thought was really interesting comes from my own situation. I currently reside in good ole’ Baltimore, MD. I’m therefore conveniently positioned in the densely populated ‘Northeast Corridor’ of the USA that spans from Boston, MA down to Washington D.C. This area of the USA boasts the most dense concentration of transportation infrastructure in the States, both physical and telecom. I have the luxury of fast internet at a reasonable price, and if I really wanted to, I could hop on over to a University library and get on the Internet-2 bandwagon for some really insane speeds.

Recently, I’ve been having issues with my Data Warehouse for SCSM remaining intact on my laptop lab. I’m sure it’s a case of too many reboots and restarts, too many ups, then downs, and so it’s had me looking at setting up a more permanent lab somewhere that isn’t on my laptop. I started looking at cloud solutions because hey, why not! Everyone’s doing it, it’s got to be cost effective… right?

Azure (US East, USD):

1, Medium (A2) Basic Instance Annual cost: $110/month = ~$1300/year

I’m going to cheap out, but let’s say I go with a Medium (A2) instance for my SCSM server. On that single server, I could toss the SCSM Workflow Server and a SQL Standard instance. It would be stupid to do, since an A2 instance only gets me 2 cores and 3.5Gb RAM, but I’m trying to keep costs low. Right. So that’s just one server, let’s scale up a bit.

1, Medium (A2) Basic Instance Annual Cost: $110/month = ~$1300/year

1, Large (A3) Basic Instance Annual Cost: $219/month = ~$2628/year

Alright, now I’ve got a large instance for my SCSM DW server, and the medium one for my Workflow server. I’m looking at almost $4000 USD a year to run two servers. This doesn’t even start to include an SCCM Server, Domain Controller (Though I think Azure has other services in play for that) or any client machines or servers for hosting demo web-portals. Wowza.

Amazon (US East, USD):

1, m3.large Instance Annual Cost: $197/month = ~$2365/year

Now this isn’t exactly apples to apples, since this m3.large instance has 7.5Gb RAM and can flex up to 6.5 ‘Elastic CPUs’, but humor me here. Double that for two servers, like above, and we’re at about $4700 a year to run two instances of this, again, with no additional machines for clients or other servers. Oh yea, and if you don’t have licenses for the MSFT software, good luck (Thankfully I have MSDN, phew!).

So, imagine I’m not even paying for my internet access here at home, it’s still a lot of money to use cloud services for a small guy like me, running an instance 24/7. What’s my alternative? Well, it is, of course, my favorite – do it yourself!

I priced out a rough machine that would do what I wanted from a hardware perspective:

1, Supermicro MBD-H8DCL-6F-O – $360

1, AMD Opteron 4386 8 core, 3.1 GHZ – $350

64GB Kingston RAM – $670

1, 1TB Samsung SSD, $470

Case, power supply, and other things: $250

Total hardware cost: $2100

Now let’s factor in power – I tried to do this with a pretty low power requirements, but I’m going to estimate on the high end.

600 Watts * .12 per kWh * 24 hours/day  = $52 per month = ~$630 per year

So for a total cost of $2730 in the first year, and $630 every year thereafter, I can have a server that can run an entire lab of VM’s (I’ve got 13 running on my laptop right now with 4 cores and 32GB RAM – this server could double that easily). There’s no point in me going to the cloud, at all.

The cloud may be the future, but the future isn’t now, at least not for everyone. Thankfully, Microsoft hasn’t totally killed off their on-premise solutions yet. Let’s just hope they don’t get around to doing that for a long, long time.