Software and SSDs are the Lethal Combination that may turn out to be the Real FC HDD Killers
A year ago at least two analysts forecast that 2010 would see the accelerated adoption of solid state drives (SSDs) with an accompanying drop-off in the purchase of high end FC hard disk drives (HDDs). Well, here we are 12 months later and while the adoption of SSDs is picking up, at least one SSD provider is reporting a short term oversupply of SSDs even as FC HDDs continue to sell. So what gives?
The value proposition of SSDs seemed so clear a year ago. Sure, the price of SSDs on a per gigabyte basis was high when compared to FC HDDs (roughly 10x) but it was thought that SSD's performance edge and lower power consumption would be more than enough to compensate for the price discrepancy.
Now we are getting a reality check. The truth is that while enterprise users want SSD, most are NOT willing to pay any price to obtain them. Maybe a few high performance applications justify an investment in SSD to accelerate their performance but for other corporate applications, SSD is on hold for now.
The difficulty in deploying SSDs is as much about SSD's cost as it is getting the right application data on the SSDs. Even the most performance intensive applications that could benefit from SSDs typically only need their most active data (typically 10% or less) on SSD. So to put all of an application's data on SSD is not only waste of SSD resources, it can become cost prohibitive, especially for those production applications that require hundreds of GBs or even terabytes of storage capacity.
So in the last six months, we have seen SSD start to be positioned as a form of cache. Rather than trying to put all of an application's data on SSD, new solutions are emerging that are only placing the application's most active data on SSD.
To do this effectively, SSDs and software are needed. A prime example of this is 3PAR's recent introduction of its Adaptive Optimization software along with SSDs in its InServ Storage Servers. A previous blog I wrote details these how these two work together but, in short, the SSDs provide the performance boost while the Adaptive Optimization analyzes the application's data and makes sure that only the most active application data is placed on the SSDs at the appropriate times.
This is also where it starts to get interesting in terms of the future of FC HDDs. Assuming the statistic is correct that 10% or less of all application data is active (and feedback from the field suggests that it is), it begs the question - why put any data on FC disks at all once SSDs are deployed in this configuration? By dynamically and intelligently place the most active application data on SSD and the rest on lower cost but adequately performing SATA, FC HDDs may not be needed at all.
In the short term (next 6 - 24 months), do not expect organizations to drop FC HDDs entirely. As they deploy Adaptive Optimization and SSDs, they will likely still want to keep both FC and SATA HDDs in the mix for the simple reason that they will want to verify both how well Adaptive Optimization actually works and what percentage of the data it places on FC and SATA HDDs.
Craig Nunes, 3PAR's VP of Marketing, expects that 3PAR's customers will initially deploy SSD, FC and SATA drives on its InServ Storage Servers in the following ratios: 5% or less of the storage capacity will generally be SSD; roughly 10-15% will be FC HDDs and the remaining 80-85+% of storage capacity will be SATA based upon workloads that 3PAR expects.
Because of how Adaptive Optimization works, it will then place data on the appropriate tier such that the majority of application I/Os should originate from SSD and most of the remaining from the FC HDDs with the SATA HDDs contributing close to 0% of the application's total I/O.
But if you notice the breakdown of the storage capacity percentages of SSDs, FC and SATA on 3PAR's storage servers, it is not unrealistic to think that the combined storage capacity percentages of SSD and FC might range from as low as 7% to as high as 15%. This would suggest that as organizations get a better understanding of their application workloads, become comfortable with how Adaptive Optimization works and the price per GB of SSDs continues to drop, 3PAR customers will abandon FC HDDs altogether in favor of SSDs.
SSDs were initially thought to be the technology that kills off FC HDDs but as the last year has shown, that was not the correct assumption. Instead what is playing out is that software like 3PAR's Adaptive Optimization is needed to cost-effectively deploy SSDs in enterprise environments. It is this lethal combination of software and SSDs that may ultimately kill off high performance FC HDDs in the not too distant future.
The value proposition of SSDs seemed so clear a year ago. Sure, the price of SSDs on a per gigabyte basis was high when compared to FC HDDs (roughly 10x) but it was thought that SSD's performance edge and lower power consumption would be more than enough to compensate for the price discrepancy.
Now we are getting a reality check. The truth is that while enterprise users want SSD, most are NOT willing to pay any price to obtain them. Maybe a few high performance applications justify an investment in SSD to accelerate their performance but for other corporate applications, SSD is on hold for now.
The difficulty in deploying SSDs is as much about SSD's cost as it is getting the right application data on the SSDs. Even the most performance intensive applications that could benefit from SSDs typically only need their most active data (typically 10% or less) on SSD. So to put all of an application's data on SSD is not only waste of SSD resources, it can become cost prohibitive, especially for those production applications that require hundreds of GBs or even terabytes of storage capacity.
So in the last six months, we have seen SSD start to be positioned as a form of cache. Rather than trying to put all of an application's data on SSD, new solutions are emerging that are only placing the application's most active data on SSD.
To do this effectively, SSDs and software are needed. A prime example of this is 3PAR's recent introduction of its Adaptive Optimization software along with SSDs in its InServ Storage Servers. A previous blog I wrote details these how these two work together but, in short, the SSDs provide the performance boost while the Adaptive Optimization analyzes the application's data and makes sure that only the most active application data is placed on the SSDs at the appropriate times.
This is also where it starts to get interesting in terms of the future of FC HDDs. Assuming the statistic is correct that 10% or less of all application data is active (and feedback from the field suggests that it is), it begs the question - why put any data on FC disks at all once SSDs are deployed in this configuration? By dynamically and intelligently place the most active application data on SSD and the rest on lower cost but adequately performing SATA, FC HDDs may not be needed at all.
In the short term (next 6 - 24 months), do not expect organizations to drop FC HDDs entirely. As they deploy Adaptive Optimization and SSDs, they will likely still want to keep both FC and SATA HDDs in the mix for the simple reason that they will want to verify both how well Adaptive Optimization actually works and what percentage of the data it places on FC and SATA HDDs.
Craig Nunes, 3PAR's VP of Marketing, expects that 3PAR's customers will initially deploy SSD, FC and SATA drives on its InServ Storage Servers in the following ratios: 5% or less of the storage capacity will generally be SSD; roughly 10-15% will be FC HDDs and the remaining 80-85+% of storage capacity will be SATA based upon workloads that 3PAR expects.
Because of how Adaptive Optimization works, it will then place data on the appropriate tier such that the majority of application I/Os should originate from SSD and most of the remaining from the FC HDDs with the SATA HDDs contributing close to 0% of the application's total I/O.
But if you notice the breakdown of the storage capacity percentages of SSDs, FC and SATA on 3PAR's storage servers, it is not unrealistic to think that the combined storage capacity percentages of SSD and FC might range from as low as 7% to as high as 15%. This would suggest that as organizations get a better understanding of their application workloads, become comfortable with how Adaptive Optimization works and the price per GB of SSDs continues to drop, 3PAR customers will abandon FC HDDs altogether in favor of SSDs.
SSDs were initially thought to be the technology that kills off FC HDDs but as the last year has shown, that was not the correct assumption. Instead what is playing out is that software like 3PAR's Adaptive Optimization is needed to cost-effectively deploy SSDs in enterprise environments. It is this lethal combination of software and SSDs that may ultimately kill off high performance FC HDDs in the not too distant future.
Newsletter Signup
Leave a comment