IBM FlashSystem

IBM FlashSystem

Find answers and share expertise on IBM FlashSystem

 View Only

Apples and Oranges

By Tony Pearson posted Fri November 12, 2010 11:31 AM

  

Originally posted by: TonyPearson


On Wikibon, David Floyer has an article titled [SAS Drives Tier 1 to New Levels of Green] that focuses on the energy efficiency benefits of newer Serial-Attach SCSI (SAS) drives over older Fibre Channel (FC) drives. This makes sense, as R&D budgets have been spent on making newer technologies more "green".

Fellow blogger Hu Yoshida (HDS) encourages people to [Invest in the Future with SAS, SATA and SFF], referring to Figure 1.

Of course, people might consider this an [apples-to-oranges] comparison. Not only are we changing from FC to SAS technology, we are also changing from 3.5-inch drives to small form factor (SFF) 2.5-inch drives. It seems odd to specify 2000 drives, when only two of the five scale up to that level. Few systems in production, from any vendor, have more than 1000 drives, so it would have seemed that would have been a fairer comparison.

However, Hu's conclusion that the combination of SAS and SFF provides better performance and energy efficiency for both IBM DS8800 and HDS VSP than FC-based alternatives from any vendor seems reasonably supported by the data.

Meanwhile, fellow blogger David Merrill (HDS) pokes fun at IBM DS8800 in Figure 2 in his post [Winner o’ the green]. This second comparison was for 4PB of raw capacity, which 4 of the 5 can handle easily using 2TB SATA drives, but the DS8800 is based on SAS technology and does not support 2TB SATA drives. A performance-oriented configuration with four distinct DS8800 boxes employing 600GB SAS drives is used instead, causing the data for the DS8800 to stick out like a sore thumb, or perhaps more intentionally as a middle finger.

The main take-away here is that IBM offers both the DS8700 for capacity-optimized workloads, and the DS8800 for performance-optimized workloads. Some competitors may have been spreading FUD that the DS8700 was withdrawn last month, it wasn't. As you can see from the data presented, there are times where a DS8700 might be more preferable than a DS8800, depending on the type of workloads you plan to deploy. IBM offers both, and will continue to support existing DS8700 and DS8800 units in the field for many years to come.

technorati tags: , , , , , ,

2 comments
13 views

Permalink

https://community.ibm.com/community/user/blogs/tony-pearson1/2010/11/12/apples-and-oranges

Comments

Archive User
Tue November 16, 2010 10:18 AM

Originally posted by: TonyPearson


David, Thanks for the time and effort you put in to making this data available. I am not disputing the results of these data points, and certainly we are in agreement that any information that can help provide comparison for purchase decision can be useful. If the focus was comparing 2.5" versus 3.5 inch drives, it might have been better to keep all other attributes as equal as possible. Having two boxes of some units (DS8000,HDS USP,HP 24000) compared to one box of others raises doubts on whether the non-disk portion of the energy consumption might be involved. On some of these disk arrays, the non-disk portion includes power supplies, processors, cache, fans, and so on. If your results were similar having 1000-drive configurations for all of them, please publish those results. I would like to see them. -- Tony

Archive User
Mon November 15, 2010 11:42 PM

Originally posted by: dfloyer


Hi Tony Thanks for your interest in the Wikibon green comparisons for Tier 1 storage arrays. If an “apples to oranges” comparison is comparing arrays with 2.5'' SAS technology with arrays with 3.5'' FC/SATA technology, guilty as charged; our Wikibon members are asking just that question. The choice of 2,000 disks was dictated by three of the four Tier 1 storage array vendors supporting 2,000 disks. If only a 1,000 disk array is offered (rounding the numbers), then two 1,000 drive arrays were configured in the comparison. This is a fair, but not the only comparison that could be made. However, the results would not change significantly if 1,000 drives had been chosen as the unit of measurement. The point of the performance and capacity configuration enveloping was to show that there is significant difference between the 2.5'' SAS and 3.5'' FC/SATA technologies – the 2.5'' SAS technology is less mature and is not suitable for capacity workloads. IBM has chosen to have two separate boxes for Tier 1 3.5'' FC/SATA and 2.5'' SAS, EMC does not yet have 2.5'' SAS for Tier 1, and Hitachi allows both to be mixed within the same array. The purpose of the Wikibon green comparison is to help customers decide the best fit for their environment. David Floyer CTO & Co-founder, Wikibon david.floyer@wikibon.org