Some interesting feedback on the post on Friday – wanted to share some thoughts and ideas on those.
Specifically, David wrote in with several things that have been his experience in working with the cloud. I have to say, I’ve seen these quite a lot, both personally with project migration attempts (and failures) of our own, and those of people we’ve been called in to review.
Specifically, David writes: “Sorry, never seen a cloud implementation be as fast as a physical implementation in a well chosen data center.
Never saw that cloud providers actually have any vested interest in my company’s best interests.
Never saw then respond to a problem as fast as the local system admins would.
Never saw any cost savings using a cloud provider.
Never trust and never use one. You can save a lot of money with a well chosen data center and buying you own hardware.“
As I said, this isn’t a rare experience, unfortunately. I wanted to offer a few ideas that might help alleviate the pain, or at perhaps even point out that hosted services aren’t a good fit for your particular application (or environment.
To the first point – that of performance. This may depend somewhat on the hardware you’re electing to use, and certainly will depend on something that many of the hosting companies refer to as “noisy neighbors.” Noisy neighbors refers to the others sharing your physical infrastructure and most certanly can impair your performance. We’ve seen issues with cabinet saturation, with physical equipment over-subscription and with surge issues and the noisy neighbor issue.
To find out how this may impact you – ask. A lot of questions. Like how this is monitored, what the standards are for perfomance, what you can expect when issues arise, what steps will be taken by the provider (if any) and in what timeframe. How do you escalate their response and get attention when you’re in the thick of it?
One thing I’ve seen very frequently is that the answer to these questions can vary wildly depending on the systems you’re using. If you’re on a dedicated bit of hardware, obviously you don’t want to be fighting with others on your network trunk for bandwidth just because they’re having a killer sale on their product. You’re more likely to get a finite, specific response answer in a dedicated situation. If you’re on a shared server, the answers get more fuzzy and difficult to quantify. I think the most frustrating answer I see frequently is “best efforts.” As in – “the provider will use their best efforts to monitor and address performance…” — that type of thing.
What can you do? Force as much transparency as you can muster. Keep asking until you understand their response. You may not like it, even. But at least you’ll know.
On cloud providers not having a similar to yours vested interest. Very true. You’ll be fallling back on their support group, their escalation options and the overall stability of their platform. The level of communication you receive, and priority and assistance, will be directly related to your support level. It’s just a fact. So, again, ask questions. “What happens when *I* see things go offline and need assistance?” – this is a good question – it suggests to them that you are paying closer attention in the moment than they may be, and, though their first response is often “we have a full team…” you need to push beyond that and find out what their response is. Again, you may not like the answer. It may not be sufficient for your tastes; but that knoweldge is just as critical in your decision making.
On cost savings – In my experience, the cost savings can be there, but it takes a lot more work to manage costs. Storage costs, bandwidth, little add-ons – they all add up and slip under the radar. Temporary servers that are brought online for testing don’t get shut off. General growth that is easier to allow than manage. These are things that are hard to manage in the limitless environment of the cloud. And the bill is where you’ll see this. Ask about controls, ask about alerts – see if you can get an alert when the bill is exceeding a certain amount, or if new services are added, or both. It can really help. It’s also a great idea to regularly review your usage and each of the servics you’re involved with. Sometimes it can be very clear and very easy what you can to do trim those expenses in signficant ways.
Trust. That’s a biggie. In my mind, if you can’t prove trust and can’t feel good about that relationship and the security and trustworthiness of the service, you have no business even considering it. It’s just not worth it. You have to feel good about the measures in place, you have to have a good understanding of what is being done and you have to LIKE those things. It’s not on you to make up things you like, it’s on them to show why they do what they do, and why that’s great for you. But I think this one is far more prevalent than we may think when people are considering cloud solutions. I’ve talked with so many people that give me a laundry list of things that are issues or questions and if you net it all out, it’s trust.
Hopefully this will help – the big thing is communication (it almost sounds like relationship counseling, sorry). It comes down to making sure you have answers to your questions that move beyond the standard, canned answers. You need to feel like you really understand what you’re considering relying on and how it would play in your own world.
There are huge benefits in the right mix. There are huge pitfalls in the wrong one.