Posted on January 26, 2006 by steven

When did SATA become full duplex?

I am often asked to comment on the performance of SATA drives as compared to FC in NetApp gear. Although I can’t seem to find any reliable perfromance figures comparing the two in a filer application I can’t see how a SATA drive can perform as well as a FC drive since the FC drive is Full Duplex and the SATA drive is not.

Jon Toigo has a comment on all of this which I found simple & easy to understand:

January 25th, 2006

I keep getting pinged by folks regarding the differences between SATA and SAS. Is SATA full duplex? Is it enterprise-ready (whatever that means)? Do the drives really shake so much that over the course of the evening the drives will shake themselves out of the array cabinet and end up in a pile on the floor? The answers are, in order, no, who knows?, and bullshit.

A good debate between Fujitsu and Western Digital on the relative merits of the technologies can be found here.

He also pointed me to this article:“A first glance of paper specifications may make SATA look similar in performance to SAS but do not be fooled by first impressions as when it comes to enterprise class server performance SAS drives will considerably outperform any SATA drives due to the fact that the interface is Full Duplex rather than Half Duplex (as is the case for SATA), the SAS drives have lower latency, lower command overheads, faster access times, deeper command queue depths and also offer the possibility of configuring dual-ported solutions which enables a further doubling of data rates.”

“Even if the SAS drives are only used in single-ported mode the Full Duplex nature of SAS could bring about a doubling in performance from the interface alone. The important thing is to measure transactions rather than data rates as it is transactions per second performance that normally matters in most server applications. If data streaming only is required then SATA interfaced drives may prove adequate and the user may as well choose the lowest cost per GB in these instances provided they have a secure backup of their data”

I wrote this note to Jon Toigo in reference to NetApp performance :

Hi Jon:

We constantly get asked about the relative performance of FC to SATA drives, because NetApp is pushing the SATA option to provide a lower cost solution to their customers on the FAS3000 series. Our customers would like to see valid, verifiable and repeatable data on the performance between a NetApp Filer with FC drives and one with SATA drives. In the NetApp case, I wonder if the NVRAM card takes up the slack caused by the half duplex SATA drives? Maybe it doesn’t matter on a lightly loaded filer, with few users. However, the FAS3000 is marketed & sold as a high performance unit, not as a disk to disk back up device.

Mike

This entry was posted in Uncategorized. Bookmark the permalink.