Posts Tagged HDD

DataSlide re-invents the Hard Drive

Posted by on Tuesday, 16 June, 2009

From: http://www.theregister.co.uk/2009/06/15/dataslide_berkeleydb/

UK-based data storage start-up DataSlide has announced potentially revolutionary hard drive technology, and a Partnership Network agreement with Oracle for the Berkeley Data Base to be embedded into the device.

DataSlide’s Hard Rectangular Drive (HRD) does not use read-write heads moving across the recording surface of a spinning hard disk drive (HDD). Instead an ultra-thin, 2-dimensional array of 64 read-write heads, operating in parallel, is positioned above an piezo-electric-driven oscillating rectangular recording surface, and delivers 160,000 random IOPS with a 500MB/sec transfer rate.

For comparison, a STEC ZEUSIOPS SSD, as used by EMC, IBM and others, with up to 320GB capacity, can provide 220MB/sec read bandwidth, 115MB/sec write bandwidth and 45,000 random IOPS. There is no read-write asymmetry, typically found with SSDs, with the HRD because it uses a standard hard disk drive recording medium and not flash memory.

Charles Barnes, DataSlide’s CEO, said: “DataSlide’s Massively Parallel architecture with 64 heads per surface could saturate a 32-lane PCI express bus. The Hard Rectangular Drive has the industry reliability and cost advantages of Hard Disk Drives with superior performance and lower power then Solid State Drives.

“The HRD uses over 60 per cent lower power than HDDs and during idle the media has zero power dissipation making it the green storage winner.”

The technology is also more shock-resistant than hard drives. This could be described as a solid-state drive with none of the well-known NAND flash problems, such as read-write asymmetry and write endurance.

Oracle’s Embedded Global Business Unit stated: “DataSlide provides a high bandwidth, low latency, magnetic storage device whose architecture lends itself to vastly improved database throughput and latency reduction.”

There is a description of the DataSlide technology here (PowerPoint deck pdf). Literally, it is non-revolutionary, using oscillations to move the magnetised bits to and fro underneath the read-write heads so that they can use magnetism value changes at the bit edge just as a read-write head on a spinning hard drive does, but where the recording layer passes continuously under the heads. There is no seek time access delay with the HRD.

The Embedded Global Business Unit at Oracle has an OEM charter and Data Slide meets its requirements by incorporating the Berkeley DB into the actual storage device to make what it calls a ‘smart’ storage device. It says potential applications are many and varied. Examples include TCP/IP-based systems and video applications requiring multiple concurrent streams. The company says media indexing, fast positioning, forward, back, skip, and scene/track operations will have significant performance improvements with its technology.

DataSlide is a privately-held company with locations in the United Kingdom, France and United States. It has affiliations with academia from Carnegie Mellon University/DSSC, and the Universities of Cambridge, Exeter, Sussex, Sheffield and Brighton in the UK, and Paris-Sud in France, and is backed by angel investors. It has a management team with experience from companies such as Seagate, Connor, Quantum, Maxtor and HP.

The technology is proven in a research and prototype sense, and the company emphasises that it uses standards-based mature process technologies from LCD, HDD and semiconductor manufacture. There is no need to design and tool-up a new manufacturing process.

DataSlide is currently in discussion with a number of storage and system OEMs and can provide more details under a non-disclosure agreement. It will be holding private meetings at the Santa Clara Hyatt Regency from June 22-25 during the Memcon 2009 conference.


Tracking down io problems on your Linux box

Posted by on Tuesday, 16 June, 2009

I’m sure everyone has had problem at one time or another trying to figure out why your machine is going so slow, but nothing appears to be using the RAM or CPU at all.

The first option is to top ‘top’ and look for the line which has the ‘wa’

Cpu(s): 0.0%us, 0.0%sy, 0.0%ni, 96.7%id, 3.3%wa, 0.0%hi, 0.0%si, 0.0%st

mine says 3.3%wa – this is the wait time trying to write to disk. Now from there you can install the package (under most distros) called ‘sysstat’

sysstat – sar, iostat and mpstat – system performance tools for Linux

This contains several tools for trying to track down whats using the disk to write lots.

iotop – simple top-like I/O monitor. This is installed and can show you realtime whats writing to disk at any time and using what load
iostat – Report Central Processing Unit (CPU) statistics and input/output statistics for devices, partitions and network filesystems (NFS).
sar – Collect, report, or save system activity information.

If there’s plenty of cache/buffers, and sar -W 1 0 shows lots of zeroes (and possibly the occasional blip) then the disk is getting thrashed, but it’s not swap.

Running iostat -dx 1 will show you all the partitions and how hard they’re working (look at %util). If %util is consistently at or around 100 for any partition of disk, you can definitively say that the disks are getting thrashed.

If the disk has high %util, but the actual throughput (rsec/s and wsec/s) is pretty low, then it’s possible you’ve got a hardware fault or RAID rebuild going on. A hardware error might show up on a smartctl run (smartctl -a /dev/sda or whatever), looking at things like the reallocated sector count, but SMART isn’t real, well, smart, so don’t trust it too much. A RAID rebuild should show up in your RAID management (you are monitoring your hardware RAID setup, aren’t you?). A software RAID rebuild will be shown in /proc/mdstat. (cat /proc/mdstat )