Not logged in » Login
Oct 12 2016

Storage for the 21st Century: FUJITSU ETERNUS All-Flash Arrays (Part 1 of 2)


An old adage says IT departments only know three problems with storage – throughput, latency and capacity, or TLC for short. Fujitsu's brand-new ETERNUS AF line of arrays will cure the ills in all of these areas. In part 1 of our 2-part feature, we discuss the challenges flash storage systems are faced with today and compare them to the benefits they bring.

Today's data-driven economy tends to create a scarcity of its own: there never seems to be enough information in the right place at the right time to let you do the job you want to finish. Occasionally, that's because required data hasn't been recorded. However, more often the problem is that the info exists, but isn't accessible within a reasonable time span – meaning the time span required by business applications or OLTP systems to support a constant flow of input and transactions. The reason is that regular, disk-based storage arrays are technically unable to deliver the necessary performance.

Over the past 15 years, customers and ICT vendors have applied various strategies to resolve this issue, from using high-speed systems with disks that spin at up to 15k RPM to implementing application-friendly data management. Unfortunately, none of these could really get to the root cause of the problem until a more radical shift occurred in 2008. Back then, the first enterprise-grade solid-state drives (SSDs) appeared on the market – a new class of devices that had enough oomph to serve as fast caches within high-end disk arrays. The effects were as tremendous as they were immediate; reports of applications running twice as fast or at even more blistering speeds soon became the norm. But since SSDs were still of limited capacity and comparatively expensive, they would only be used in dedicated arrays meant to power important key applications, such as databases or ERP systems, but not standard web or file servers. Still with mobile and cloud computing ever expanding, such restrictions couldn't last for a very long time, and soon the first all-flash arrays appeared.

While the broader adoption of flash technology was generally hailed as a step in the right direction, it brought about issues of its own. One is that although SSD prices have dropped, they still carry a relatively heavy price tag when compared to HDDs. Another is that early SSD-based systems were complicated to handle, because they did not support central storage management techniques and features and thus required separate, usually manual efforts. Storage vendors will have to tackle both problems with their next generation of all-flash arrays – and that's precisely why Fujitsu built the ETERNUS AF.

Fresh Challenges for Flash
Simplifying flash management is also going to help with several other difficulties early adopters have encountered over the past few years, most of which had to do with tuning SSD-equipped storage systems and ensuring non-stop, automated operation. Earlier this year, researchers from the UK-based analyst firm Freeform Dynamics came up with a comprehensive list of capabilities that an all-flash array (AFA) must possess if it has to shoulder both routine and mission-critical enterprise workloads. Below is the abbreviated version:

  • An AFA must be fit for general-purpose shared storage instead of being "specific to or optimized for any single application or group of applications." As a consequence, it shall not require manual fine-tuning.
  • It should be "designed as a complete system, ideally with a non-blocking architecture and without internal bottlenecks." Only then will the system be able to achieve "fast and consistent response times" and offer reliable quality of service (QoS).
  • In order to accommodate varying workloads, the system should be easy to administer even for less experienced staff – or better still, be able to "tune itself."
  • Prioritizing workloads must be simple and straightforward, so that IT departments can meet strict SLAs and deliver the best possible QoS. Again, an automated process is preferable.
  • Management capabilities must cover the following areas: data protection (DP), high availability and disaster recovery (HA/DR), and data reduction (deduplication, compression).
  • An AFA should have an above-average lifespan and be easily scalable in order to justify and protect the initial investment.

Flash Benefits
These challenges notwithstanding, AFAs can provide IT departments with huge technical advantages over the competition. We already mentioned the massive performance/speed gains that result from a combination of higher throughput and lower latencies than you will find in any disk-based array. Other benefits however, are less obvious – though equally important:

  • Low latencies mean that AFAs have enough time – or rather: instruction cycles – to spare to allow for performing data reduction operations "in-line," i.e. before data is written to an SSD. As a result, users can avoid over-provisioning and delays in (or worse: manual triggering of) data deduplication and compression.
  • HDD performance deteriorates over time as a medium fills up – the more files users store on a spinning platter, the more time it takes for the read/write heads to snap into the right place and do their job. This inevitably causes delays in data processing that slow down applications; in addition, mechanical components are more likely to produce read/write errors or fail entirely as they age. SSDs have no mechanical components, so they offer more reliable, even predictable performance and are less susceptible to wear. Former reservations about flash's limited lifespan are no longer justified, since both the chip designs and the management software improved substantially in recent years.
  • Thanks to the greater reliability of SSDs, administrators will have to spend less time on adjusting arrays for better performance. And because deterioration is no longer an issue, the usable capacity of a medium is much closer to its raw capacity. In short, AFAs require less tuning efforts and support better storage utilization.
  • Due to the capacity and performance limitations of HDDs, IT departments often resorted to overprovisioning – that is, they purchased arrays with more disk capacity than they needed to achieve the performance gains required to run demanding applications. As pointed out above, AFAs do not suffer from these limitations. Moreover, SSDs in general offer higher storage density and eat up considerably less power than HDDs; so IT departments can work with smaller arrays and still achieve the same results at lower costs and with a smaller ecological footprint.

All of these benefits help to put the higher initial investment into a more realistic perspective. We will explain how they play out in our brand-new ETERNUS AF platform in part 2 of this article, which is due for release in an hour – so stay tuned!

René Hübel


About the Author:

René Hübel

Senior Product Marketing Manager Storage Solutions 


Comments on this article

No comments yet.

Please Login to leave a comment.


Please login

Please log in with your Fujitsu Partner Account.


» Forgot password

Register now

If you do not have a Fujitsu Partner Account, please register for a new account.

» Register now