Cloud storage and its main cost components

Object storage

These are large-scale data repositories designed to cover wide geographies and provide high capacity. Object is at the heart of most cloud storage. Normally, object storage is not accessed directly but via a third-party tool. Backup, archive and global file shares are all common uses of back-end object storage.

Disk

A more traditional form of storage. Usually only used as underlying storage for cloud-based virtual machines and unlikely to be integrated with your on-premise servers.

File services

Provide NAS-like cloud storage with flexible capacity accessed via standard NAS protocols such as NFS or SMB. Normally presented to users as cloud home directories or used to present data to other cloud services. Special note should be paid to bandwidth allocation when considering file services.

Where do we incur costs?

Understanding storage costs may seem complex at first, but it doesn’t have to be. Most storage costs are calculated using a set of elements that are standard across the main providers and their offerings. Key areas of cost are:

Capacity

This is usually the easiest to determine and normally calculated per unit over a period. Per gigabyte per month is standard and charges are prorated, so you are charged only for capacity consumed over a given time.

Capacity charges mean it is extremely import to understand our data. Paying to hold data that nobody accesses may not be an issue in your datacentre, but when you are presented with a monthly bill, it can be a very different consideration.

Transactions

Often the least obvious of the charges and more difficult to estimate. Most commonly seen with object storage, transactions are the interactions between applications and storage, for example a read, write or scan of your repository.

Individual transaction costs are not high, usually fractions of a penny per 10,000 transactions. Even though a large amount of transactions is unlikely to incur significant cost, an understanding of your applications’ transaction profile can be valuable because high-transaction applications can leave you with unexpected costs.

Network

While the major providers are happy to ingest your data for free, all of them charge to move your data, whether between their own datacentres or back to yours.

If you are looking to move your data between datacentres within a single vendor region, normally this can be done for free. But if you want your data replicated between regions (or your own datacentre) to provide increased resilience, this will incur charges. Networking is crucial in our cloud analysis both technically and commercially, with unaccounted costs potentially having a major impact.

Snapshots and backups

Taken for granted in most enterprise storage solutions, snapshots and backups must be considered and modelled in our cost profile. Snapshots and backups not only take up additional capacity (in Microsoft’s case, with no incremental snapshot model, this can be high) but moving that data from one repository to another as part of your protection policy will also incur network charges.

This is not an exhaustive list, but shows the core elements of your cost model.

It must be noted that some services have additional cost-related nuances, such as I/O performances, bandwidth allocations and costs associated with more performant disk types, high availability and regional variation. However, the above components should provide a sound basis for your initial model.

So how do these elements affect your cloud storage bill? Below are two high-level illustrations:

Object storage

In this example, we are going to use cloud object storage to hold 1TB of monthly backups, with an application that will create 100,000 transactions plus the assumption that 50GB of data will be restored back to our datacentre.

As you can see, capacity pricing is similar across the providers, but an understanding of the transactions your local tools will make will be valuable in your cost modelling.

Disk storage

In this example, let’s consider 1TB of disk assigned to a virtual machine plus 500GB allocated for snapshots.

There are some differences between provider to consider when it comes to dedicated disks. Notably, Microsoft’s model does not utilise a “used capacity” model, but charges for a dedicated disk. It should also be noted that transaction costs are incurred with its standard tier, but not premium.

Although costs are not significantly different, the impact of Microsoft’s model should be understood. It provides increased deployment flexibility, but be aware that you pay for disks assigned, not capacity consumed.

Other costs

There are other cost considerations, such as connectivity to your own infrastructure as well as external access. Costs such as VPNs to your datacentre, firewalls and appropriate identity management for secure and reliable access must be understood.

Summary

Choosing cloud storage can seem complex, with a multitude of storage types, options and chargeable items, but understanding these standard elements should provide a good basis for modelling.

But knowing all this is only valuable if we fully understand the profile of the data we want to move to the cloud. Capacity, performance demands, levels of data protection and resilience should all be well-defined. Although we have covered some base cost elements here, each provider will have further nuances to consider, depending on your workloads and types of services available.

Understanding your data demands should be at the heart of your cloud storage decision-making. But support, performance guarantees, data protection and integration with other services on-premise and cloud are equally, if not more, important in your choice of cloud storage solution.