As you know we have a cloud presence; we have had one for several months. Now we are evaluating the cloud to determine whether to maintain that cloud presence or move back to our local data center. We also documented some early teething problems within this cloud presence. What should be our evaluation the cloud criteria? Now that our data center is moved and fully functional, should we keep our cloud presence?What should be our criteria for evaluating the cloud? Should we base it on return on investment (expense), should we base it on functionality, should we base it on redundancy? or all of the above.
Evaluating the Cloud: So first does the cloud save us any money?
We are told by vendors and cloud service providers that moving to the cloud will save us money. For our business, I do not see that. It costs the same to run the systems in the cloud as it does to run them locally. Why? Because, our data center has the infrastructure available for the small volume of workloads within the cloud. At some point there is a crossover that will make it more valuable to have a cloud presence, what that is depends on the number of workloads as well as the storage and networking required. We happen to have moderate needs. Our main need is for 100% up time more than masses amount of bandwidth and storage. When we look at this, the cloud costs us more than our local data center.
But what about increases in costs for cooling and heating? Any time you run a data center you will come across cooling and heating costs, regardless of where it is. Since we have to maintain a local data center anyways and we are 100% virtualized, the costs have already dropped as much as they can. So the real costs are based on improving the technology to either provide power, provide cooling, or reducing the number of power supplies present within the data center. Yet this is also a wash for us, we have not turned off a single power supply since we have moved to the cloud. So this savings did not materialize. We did upgrade the cooling to be more efficient so our actual power draw for cooling went down.
In our case the cloud costs us more, because we still need to maintain our local data center.
Evaluating the Cloud: So does the cloud provide 100% uptime?
This is one area a cloud presence could be better than a local data center depending on the local data center. In our case, the uptime percent with a cloud presence matched our old data center’s uptime and as such we once more see no differences. We had to make the same changes when we moved our data center to maintain uptime such as constant power and cooling, multiple network paths, and other physical redundancy methods.
For our local data center this is not quite a wash, we have generators, redundant power, redundant power supplies, but we do not have multiple network feeds into the data center. At least not yet. So in this case staying in the cloud would make sense, purely for network redundancy needs.
Where the cloud, at least our chosen provider, falls down is in the area of uptime, the cloud instances occasionally still have problems. Not very often, so we have increased our monitoring so that we can KNOW when there are issues as quickly as possible and handle them, such as rebooting a node. The root cause of these failures, we have determined and it ends up being related to the hypervisor in use.
In our case, the cloud provides us more networking possibilities but not much else.
Evaluating the Cloud: Can you get your data there and back again?
I always require that we can get our data there and back again via the cloud. However, depending on the cloud, this may be a manual process with no automation built into copy workloads back and forth between your data center and the cloud. Many times you can get your data there in some automated fashion, but back out again is often difficult. In our case, getting our data there required a manual process, and getting our data back out requires a manual but scriptable process.
In our case, the cloud provides us the necessary functionality but not in an automated fashion.
Evaluating the Cloud: Does the Cloud provide Performance?
When moving to a fully virtualized cloud we need to be aware of the fact that we are using shared resources. To that end we also need to know if those resources are being used efficiently and that the workloads perform as expected. Since we monitor everything using various tools we have noticed that there are times when the performance is sub-par within the cloud. Performance is a primary concern for us, as we use fewer resources to get a better bang for the buck however, in the cloud, instead of scaling up, we should scale out, which costs more money, impacts the cloud more, etc. So there is a major trade-off here with respect to performance.
In a local data center, you have complete control of your resources and can tune your resources appropriately, however, in a cloud, you may not have such ability. If you can get that level of ability your cloud is one of the better ones and this often costs quite a bit more. Not something a small enterprise could readily afford.
In our case, the local data center outperforms the cloud.
Evaluating the Cloud: SDDC or Hybrid Cloud
The last question I have is whether or not our current cloud presence provides us with a gateway to a true hybrid cloud implementing a software defined data center in an easy method. Currently to do this would require an expensive period of automation using scripts as there is nothing currently baked into our chosen cloud.
In our case, the local private cloud still looks better from a cost perspective.
In our case, since nearly everything is a wash and we need to cut some costs, we will be moving back to our local data center where we have better control. However, if our costs had gone down appreciably, the costs savings of going to the cloud would have been a wash and the only stand out benefit of the cloud is the multi-homed networking nature of a cloud. We can also be multi-homed in our own data center and that should be one of our own goals given the work we do, the spread of employees all over the globe, and the need to be on the network 24/7.
In the end, our determination to go to the cloud was to move our data center, to maintain up time. Now that our data center is setup and redundant power, etc. is available, staying in the cloud becomes a cost issue more than a technological issue. But not just the costs of using the cloud, but growing within the cloud as we head towards a hybrid cloud that implements a software defined data center.
Share this Article:
Latest posts by Edward Haletky (see all)
- Continuous Integration, Deployment, and Testing - July 22, 2016
- Serverless: Business Plan or an Approach to Technology? - July 21, 2016
- Root Cause Analysis Is Not Dead - July 13, 2016