In recent months, I’ve been poking around various clouds. Along the way, I realized that they were not working the way I expected. Virtual machines are not as interchangeable or as cheap as they seem. Moving to the cloud is not as simple as it should be. In other words, anyone who has thought about the word “cloud” as a synonym for “perfection” or “painless” will be very disappointed.
You cannot say that there is no truth in what the companies claim about cloud, but there is much exaggeration and also plenty of complicated details that are not immediately obvious. In essence, the cloud are not miracle workers. The improvements are incremental, not revolutionary.
To keep our expectations in more realistic level, here’s a list of what we should really expect from the clouds.
1: Uneven Performance of Virtual Machines
Cloud uncomplicated many of the steps related to the purchase of a server. The desire? Press a button, choose your operating system and get the root password. Everything else should be treated by the cloud, which takes care of all computational tasks behind the curtain.
But the benchmarks taught me that virtual machines behave quite differently. Even if you buy a body having the same amount of RAM running the same version of the operating system, you will find surprisingly different performances. There are different chips and hypervisors running underneath it.
2: Many Choices
A great promise from cloud? Rent a Cloud and see what it can do. Your boss may want to invest in buying a rack or collocating the architecture to data center. But spending a few hours in understanding cloud can help in easy decision making.
To rent Cloud under pay per use model is an ideal way for people interested in trying out some features. But with the increasing choice also increases the complexity of analysis and uncertainty about what is really needed.
3: Eternal Instances
After spending some time with the installation of software and settings, many people give up on shut down a virtual machine that costs so little per hour, leaving them in the background waiting for work.
Generally, an audit of the lists of virtual machines takes longer than the cost of leaving them running another month.
I think the cloud companies will make big money with servers that are standing there, waiting for further instructions.
4: Difficulty Dealing with SaaS Pricing
Software as a service is another temptation in the cloud. You do not need to buy the license and install. You send your bits to an API and it does everything for you. But calculate the cost of software as a service involves analyzing several variables.
One of the ways for a company to be aware of the costs is to start the execution of applications in the cloud computing platform, and then calculate the actual costs. But this raises a number of additional issues, such as how to account for the characteristic fluctuating prices of an immature market.
To prove the cloud can be a rather nebulous process, even comparing a private cloud with the traditional IT infrastructure model.
5: Embedded Solutions
When Google announced Google App Engine, it seemed that the device would make the simplest cloud computing. The problem is that the product’s owner needs to use it. This means that you will be tied to it until you can rewrite the software. And who has time to do that?
Not surprisingly, the standard OpenStack is gaining momentum. Everyone is frightened by the possibility of seeing tied to a provider, no matter how good he can be. It takes more flexibility.
6: Security Is Still A Mystery
At first glance, it seems that you completely control your machine. You and only you set the root password. If the OS is secure and patches are installed, that’s OK, right?
All the clouds are far from clear what actually happens under the hypervisor. Cloud providers are far from offering as safe environments like a cage in the center of the room where you can lock your server.
7: The Cost Estimate Is Not Easy
You should buy a faster machine to 7 cents per hour or three slower machines to 2.5 cents per hour? As each vendor has its own way to charge for bandwidth, storage and other resources, expect to spend hours analyzing the use of various sizes of servers. Then put all this data into a spreadsheet to determine the cheapest configuration.
8: Moving Data Is Not Easy
You like the idea of buying computing power per hour? But often the purchase is the smallest part of the work. Getting your data in the cloud can be a substantial task. If you are loaded log files or large data sets, you may be spending too much time in just moving the data you need.
So some suppliers are making it easy to store data locally and then buy computation time when you need it.
9: The Data Is Not Guaranteed As Well
Cloud contracts rarely have disaster recovery. Some vendors are starting to be more clear about its guarantees. Some have terms of service that explain a little better what they cover and do not cover. Others are responding faster and better questions about the physical location of the data, knowing that the answer may be crucial for regulatory compliance or security issues. The geographical distribution is critical for disaster recovery.
One of the most important items in the hiring of public cloud is the agreement of the level of service quality (SLA). Users need better answers from cloud providers on the finer points of management of availability and vulnerability of services before signing contracts. It is important for customers to know not just where your data is stored as who will access them.
10: No One Knows What Laws Apply
It is easy to imagine that the cloud is living in a Shangri-La, away from those pesky laws and rules that complicate the lives of companies. We would all like to believe that cyberspace is a beautiful place, full of harmony and mutual respect, which dispenses lawyers. What is a half-truth, because no one really knows what laws apply.
Not all the laws apply, because the Web extends throughout. As services and solutions are delivered anywhere in the world, this feature of the concept of cloud computing challenges the current legal model, which is based on local laws. As a result, the legal risks are even greater than those of other traditional IT outsourcing contracts, experts in digital right.
Join this borderless world requires more caution in drawing up contracts with service providers. It is important that the contract contains clauses on privacy issues and data availability. Companies should know the risks of hosting information outside of India. In case of a court order, the data confidentiality may be broken, depending on the privacy law and data protection applied by the country where the server is installed.
11: Extras Will Reach
The cloud business seems to be following the same model for collection of hotel companies and airlines. They will do anything to keep the cost of the main service as low as possible, because they know that price is the determining factor for the purchase of the service. But will try to compensate for this cheap price with add-ons.
The problem for most customers is that it is increasingly difficult to predict how many extra services are used. You can estimate the amount of data that will flow between their machines and the server in the cloud? Some cloud companies charge for it. If a programmer uses a data structure such as XML, can quadruple their bandwidth costs.
12: Responsibility For Backup Still Rests on You
It is tempting to buy the marketing hype and think of the cloud as a giant collection of computing resources. When you need to chew numbers, you cast your spell across the ocean and the answers are reborn from the mists.
If you think the cloud will save the responsibility of backing up your data to you, you are mistaken. Underneath it all, the virtual machines are so fragile as the machines on your desk.
In practice, the machines are machines only. If you build a backup plan for your server today, then you should build one for your software in the cloud as well. But It can also fail.