Cloud storage is a popular subject these days, with some good and some bad examples out there.  The appeal is obvious – data needs to be retained somewhere, whether it is of the primary, backup or archive flavour.  Being able to do that at low cost (especially when a large percentage of that data is probably inactive anyway) is attractive, when a full TCO is taken into account, including data centre space, power, cooling, acquisition, refresh and administration.  The latest news from Avere is that their NAS acceleration technology has been adapted and can integrate with cloud solutions from the major SSPs (storage service providers).

Background

Figure 1: Avere NAS Optimisation

Avere has been around since 2008 and were initially focused on NAS optimisation; that is being able to place an appliance some distance away from a set of core filers and cache active content locally to the user.  This core-edge idea allows businesses to concentrate their NAS deployments into a smaller number of locations, where they can reduce TCO through consolidation and still deliver the performance expected by end users.  Over time, releases of the AvereOS have implemented a global namespace (version 2.0) and enterprise-class data management features such as data mirroring and migration.  These are known as FlashMirror and FlashMove respectively and were introduced in AvereOS 3.0.

AvereOS 4.0

Version 4.0 of the Avere operating system introduces the idea of Cloud NAS.  At it simplest level, the storage provided by SSPs becomes another back-end storage pool, just like and interchangeable with dedicated NAS equipment.  All the existing benefits of single name space, data migration and data replication work the same as before.  The only difference is that data is in an object-based cloud.

There are a number of obvious benefits in using cloud within the Avere architecture.  Data can be replicated or moved seamlessly to a lower cost storage medium, without compromising on performance.  Avere quote a typical use case saving of 73% TCO, however every user’s figures will vary and should be examined as part of any cloud migration process.  Apart from cost, the Avere solution could be used for scale up/down on demand, for data centre migrations and hardware refresh.  There are lots of use case scenarios out there.

Enterprise Class

Figure 2: Introducing Cloud Storage

So the concept of introducing cloud into the Avere architecture seems quite simple from the user’s perspective.  However implementing this technology isn’t so simple.  Firstly cloud storage is all object based, so there is a requirement to translate files into objects, which means keeping metadata to cover things like ACLs.  Then there’s a question of performance; most object stores allow the storage and retention of whole objects only, however it’s inefficient to store and replace an entire file across the Interweb when only a small portion of it has changed.  In fact, that’s part of the technology Avere already rely on for traditional filers as they divide data into 16KB blocks to make use of tiering and to optimise the amount of data traversing the network.  So storage on the cloud has to be implemented to take into consideration these requirements.

We should also talk about locking.  Accelerating traditional NAS means extending the file locking process, so within a global namespace where a core filer is accessed from multiple locations, the Avere appliances manage and honour the core filer locking.  With object, no such process exists, so Avere had to implement locking within their appliance for cloud objects.

One final point to ponder is one of data integrity.  Some cloud storage solutions use snapshots to replicate data into the cloud.  The snapshot represents the data integrity point and retaining multiple snapshots gives the ability to roll back to a previous point in time.  The speed of the network determines the frequency of snapshot replication and so a slow network and infrequent snapshots means lower levels of granularity on updates.  In Avere’s case, as the edge appliance itself is resilient and data is cached (rather than permanently stored), data can be flushed from cache asynchronously and in parallel, potentially improving write-back performance.

All of the above highlight the thinking that is required when implementing an enterprise-quality solution.  The key here is to understand and honour data protection and integrity requirements that any large business would insist on.

Warm To Cold

Avere’s initial cloud support will include Amazon S3 and Cleversafe, using the customer’s own S3 account, not one from Avere.  This means the customer controls encryption keys.  In addition, S3 support will extend to Glacier with the ability for administrators to move subdirectories in the global name space over to Glacier storage.  Presumably more cloud vendors will be assimilated over time, allowing for protection against SSPs going out of business.

Performance

As part of the briefing on Cloud NAS last week in London with Ron Bianchini (CEO) and Rebecca Thompson (VP for Marketing), we discussed the subject of performance.  Readers may remember, I touched on this subject in a recent blog post where I mention Avere’s 32-node cluster.  What came up in conversation and what I missed at the time was that the Avere back-end storage in this SpecSFS test was in fact located 150ms away from the caching appliances to simulate a cloud configuration, specifically in advance of this release.  Check it out here and you can see what I mean.

The Architect’s View

Avere have continued to evolve their technology, moving from pure optimisation to a distributed storage model.  Introducing cloud was a logical next step and makes the solution even more attractive.  For enterprise adoption, there needs to be a focus on features and functionality, including the detail of how those are implemented to ensure no risk of data loss.  Chatting with the Avere team it’s clear that they have thought through the requirements of the enterprise and understand exactly what makes them tick.  I look forward to reviewing the product in the lab and hearing about what will be developed for releases 5.0 and up.

Related Links

 

Comments are always welcome; please read our Comments Policy.  If you have any related links of interest, please feel free to add them as a comment for consideration.  

Copyright (c) 2009-2017 – Post #8BBF – Chris M Evans, first published on https://blog.architecting.it, do not reproduce without permission.

Written by Chris Evans

With 30+ years in IT, Chris has worked on everything from mainframe to open platforms, Windows and more. During that time, he has focused on storage, developed software and even co-founded a music company in the late 1990s. These days it's all about analysis, advice and consultancy.