The key differences between caching and tiering

Aron Brand, CTO, CTERA, discusses in detail two different approaches to managing data migration across numerous Edge and cloud locations – caching and tiering. We hear about the benefits of both strategies and how they differ.

Cloud caching and tiering are not synonymous. These two commonly misunderstood terms are both data management techniques that combine the advantages of local and cloud storage. Tiering and caching both move the majority of data from on-premises storage to the cloud while maintaining local access, but this is where the resemblances end.  

As organisations increasingly adopt a hybrid cloud approach, IT administrators must understand how the two technologies differ in order to ensure their chosen data management solutions meet their requirements. Read on for the key differences between caching and tiering:

Copying vs. moving data

Data is transferred between the local, on-premises storage tier and the cloud storage tier with both caching and tiering, but in different ways: caching copies data between tiers, whereas tiering moves it.

Edge-centric vs cloud-centric approach

Tiering is a technique that focuses on the Edge. In this scenario, portions of locally-stored data are migrated, according to pre-established criteria, from the Edge to a slower and cheaper tier, such as the cloud, and retrieved on demand. The cloud is where cold data is archived at a lower cost, with local storage acting as the primary storage. At any one time, data is stored in a single tier.

Caching, on the other hand, is a cloud-centric strategy in which the cloud stores a ‘gold copy’ of all data. For rapid and efficient access, on-premises caching devices keep a local copy of frequently-accessed data.

Let’s look at how caching and tiering stack up against four use cases for hybrid cloud storage, as defined by Gartner in its Market Guide for Hybrid Cloud Storage: burst for capacity, disaster recovery, burst for compute, and data orchestration.

Caching vs. tiering: Burst for capacity

Burst for capacity allows Edge devices to expand their storage capacity indefinitely and elastically, leaking excess data into a low-cost cloud storage tier. Cloud storage is particularly cost-effective for capacity bursting because it is elastic and organisations only pay for the capacity they utilise. Both tiering and caching are well suited to this use case.

Caching vs. tiering: Burst for compute

When a dataset is created locally but needs to be accessed in the cloud for processing or analytics, burst for compute is employed. A visual effects company, for example, may run 1,000 cloud servers for eight hours to render 3D models developed by a team of artists working locally. Tiering is not appropriate for this use case, as live data processing (i.e. rendering) cannot take place in the cloud. Caching saves both hot and cold data in the cloud, allowing data analysis and processing to make use of the cloud’s high-performance compute capabilities.

Caching vs. tiering: Disaster Recovery

Local data is backed up to the cloud for Disaster Recovery and Business Continuity. Caching enables Disaster Recovery capabilities and, more critically, fast recovery by keeping all data in highly robust and redundant cloud storage. In the event of a disaster, a new caching device can be launched anywhere in minutes to provide data access instantly while the cache is warmed up in the background. Tiering, on the other hand, only stores cold data in the cloud; safeguarding local data is outside the purview of tiering and necessitates the use of a separate backup solution.

Caching vs. tiering: Data orchestration

In hybrid cloud deployments, data orchestration is utilised to obtain a consolidated view of data across several clouds employing a single protocol or interface. Consider a company that wishes to display a single view of data that can be read and written from a number of Edge and cloud locations, as well as transport data across them and manage access through a single namespace. This use case is not supported by tiering because only cold data is managed in the cloud. Cloud caching, on the other hand, exposes a global multi-cloud file system that consolidates data from several backend storage clouds and Edge locations into a single namespace that can be accessed from anywhere.

To sum up, caching and tiering are two different approaches to manage data migration across numerous Edge and cloud locations. Tiering keeps live data at the Edge, while stale data is moved to the cloud. In contrast, all data is stored in the cloud and cached at the Edge for quick access via cloud caching. While tiering can help end-user organisations save money on storage, it is only useful for one hybrid cloud use case, capacity bursting. Caching is a preferable option for hybrid cloud architectures because it supports a wide range of use cases, including Disaster Recovery, compute bursting and data orchestration, in addition to capacity bursting.