Cache-style ATSs copy active data from traditional mechanical storage to a high speed memory based cache (RAM or FLASH SSD). In this copy mode they can act as a large read cache, rarely actually owning the sole copy of data. Even when they are used as a write accelerator by caching inbound writes the amount of time in which they maintain the only copy of that data can be measured most often in minutes. In both modes these systems can help cloud storage technologies service more mainstream storage requirements.


Cloud storage systems also come in several forms. One of the more common is a configuration as a 'value tier' class of NAS storage that's extremely cost effective and highly scalable. But that economy and scalability often comes with a performance penalty, which can keep a cloud based storage system from being deployed as a mainstream, day-to-day storage resource. Many customers and providers would like to be able to more broadly deploy cloud storage and are looking to ATS to bridge that gap.


The challenge that cloud storage systems present when attempting to use for more mainstream purposes, however, is that they are often a software-only solution and sometimes leverage user-provided commodity hardware and disk drives. This keeps cost down as it spreads the data set across multiple types (and brands) of storage hardware. The major storage manufacturers are focused on delivering ATS for their tier 1 storage platforms to drive up performance while keeping costs contained. However tier 1 storage is not typically used in cloud storage deployments and the major storage manufacturers are resistant to offering ATS technology on their lower tier storage systems for fear that these performance enhanced systems would cannibalize their tier 1 markets.


Negotiating BOTH the performance gap and the disparate storage gap will likely require an ATS solution from a storage supplier that is focused on delivering ATS on an external, storage agnostic appliance. This style of ATS solution can then be placed logically in front of the cloud storage network and, with just a few changes to how the users or applications access the data.


With the ATS solution in place, all network traffic goes through that system. That traffic is analyzed by the ATS appliance and then, depending on its access attributes, the active blocks of data are stored on high speed storage, typically RAM or SSD and potentially high speed SAS. Subsequent access to that data comes from that high speed storage area and, as a result, is delivered to the user or the application very quickly.


In NAS based cloud storage there are several use cases that demand more performance from the system. The first is a NAS-based cloud storage system set up as a less expensive, more scalable alternative to traditional NAS; an internal private cloud storage system. Inevitably, in this use case there's a need for greater performance than what the cloud storage system itself was designed to deliver. Adding ATS to the front end of the cloud storage can often eliminate most performance issues.


The second use case is that of the more classic 'cloud storage provider' model. If the provider has clients who have data that suddenly becomes extremely active, that data may be a candidate to be promoted to the ATS. Even though most of the individual requesters of that data may be accessing it from a relatively slow connection, the combined access of 1,000s of users creates a bottleneck on the storage.


This is compounded on the typical NAS cloud storage system. As we discussed in our article "What To Look For in Cloud Storage", most cloud storage systems are 'loosely clustered', which means the performance of the individual nodes becomes the bottleneck, since data is not striped across nodes as it is with tightly coupled clusters. The result is that if an individual file becomes 'hot' it can only be served up from one node at a time. The workaround is to copy that file to multiple nodes in the cluster and then change the application to know where else to look for that file. Additionally, the extra copies of that file need to be found and deleted as the file becomes 'less hot'. In most cases this last step rarely happens, leading to wasted space in a highly competitive, capacity charged market. In both cases this requires extra management time on the part of the storage administrators. 


A potentially easier and more efficient solution may be to add ATS. The ATS system will move the hot file, or sections of the file to a RAM or SSD-based cache area. Then, while the file is hot the system will serve it from that high speed storage area. With no (or limited) changes to the environment these hot files can be identified as they become hot and moved into high speed storage. Then, as the files become less hot they can be automatically de-staged from cache. The storage becomes self-managing and self-tuning.


Traditionally, ATS solutions are implemented to accelerate high end NAS. These systems already have high speed disk sub-systems and multiple high-speed network connections. High end NASs are commonly tuned to deliver the best performance that mechanical drives can offer. ATS is often the last resort before replacing the entire storage array.


NAS cloud storage systems, on the other hand, do not as often have the same performance capabilities as traditional NAS. As stated earlier, the focus is often on cost reduction and scalability, at the expense of performance. As a cloud storage environment grows, or as it's being considered for a more mainstream use, the lack of raw storage performance forces storage managers into more traditional solutions. An option these managers should consider are storage agnostic ATS Appliances. Combining the two technologies provides top-end performance, while maintaining much of the cost and scalability advantages.

George Crump, Senior Analyst