How To Reduce Cloud Costs
Public cloud costs are spiralling out of control
“5/8 instances on public cloud isn’t workload it was designed for.”
With the movement of data and services in to public cloud at an ever increasing rate, using this one-size-fits-all strategy means 60% of IT workload is using the wrong cost model. Here we discuss how to reduce cloud costs.
Techies globally are using public cloud for purposes it was never intended to be used for. It was initially designed as an elastic e-commerce platform for the on-demand type of compute processing required at sale time
when compute demand was high for a fraction of the lifetime of the servers. On-premise e-commerce systems were rightly deemed too expensive to own when there were much more affordable public offerings becoming widely available.
Once the move to public cloud proved it’s worth for e-commerce, businesses looked for other areas where they could outsource their infrastructure. Technical staff began to move core infrastructure to their public cloud environments too. This could work in some instances when executed in conjunction with on-premise core infrastructure, but when 100% of it is transferred to the public cloud costs quickly begin to escalate. Sure, creating a new server instance in the cloud is pretty straight forward, but when it’s this easy to create new instances it makes it that much easier to totally lose control of costs. And reading or transferring the data can prove very expensive.
So, why is it less cost effective to move core infrastructure to public cloud. Why is the model so right for one scenario and not all? How can you reduce cloud costs? Here’s just one example. Your company is designing and implementing Microsoft Active Directory application solution.
The High Availability Production Environment
High Availability needs to still be built in to your application solutions and there are many reasons why public cloud is right for this essential workload. Take Microsoft’s Active Directory for example. Microsoft best practice and vendor supportability dictates AD is an active / active resilient solution, but only when multiple domain controllers are deployed in each domain. You’re going to be deploying at least 2 but probably 3 per domain. That’s production sorted. You’ve not had to purchase enterprise equipment, you expect an always-on service and you still have the flexibility to make changes and updates yourself as necessary.
The Pre-production Environment
Now you also need a pre-production environment that should mirror production, (another 3 servers then). But does this need to be in the public cloud? No. It isn’t going to need the peak demand elasticity, it doesn’t need to provide access for multiple hundreds of users and you are going to be transferring data in and out at additional cost. That’s 3 server instances that are more suited to private cloud.
The Development & Testing Environment
You should also have a dev/test environment that doesn’t necessarily need to mirror production. You cannot afford for it to fail, so another you need 2 instances to be safe. This where the most flexibility is required, but when workload is directly related to costs, how can you sensibly budget? Being able to create server instances without requiring budgetary authority is a strange place for any employer to put an employee. And on top of that, there is no technical advantage of it being in public cloud.
That’s a total of 8 server instances for AD.
I will not bring up the subject of Disaster Recovery because there are options available that could see an active site on warm standby waiting to be used at a moments notice, or alternatively you may just clone your production site and keep it offline, ready to boot up in case of emergency. In the latter case it will not be burning compute power so would be fairly inexpensive to maintain.
So, your average monthly bill for just 8 core infrastructure servers is a large enough number, but 5/8ths of the workload isn’t what public cloud was designed for. You’ve not started implementing VPN instances, firewalls, application instances, database instances, etc. Already there is an obvious argument for keeping production servers in public cloud and everything else in an on-premise private cloud.
In conclusion, seriously consider the whether you need to transition 100% of your infrastructure services to public cloud before committing the decision to your strategy, 40% of the workload may be all that you need to make the required savings. Monthly subscriptions can get out of control very quickly and before you know it, your CFO will be asking how to reduce cloud costs. Repatriating some of the services back on-premise will be the best option. With an on-premise private cloud, coupled with the public cloud for your production environment, you’ll be able to make full use of everything cloud has to offer in the most cost-effective way.
For information on a low cost private cloud, take a look at ScaleCloud. With capacity for 400 linux servers or 60 windows instances hyper converged into a 1U chassis, we’re re-defining private cloud. Contact us.