The Hitachi disk HDS722020ALA330 stores 2 GigaBytes. Big. Cheap. Reliable. What more could you want? Quieter and faster is a start.
Linux has a disk benchmark test built in. You would expect it to be reliable and accurate because it runs at the operating system level. In the following test the Linux benchmark appears accurate for individual disks but not for RAID arrays. I do not know if it is a fault with Linux RAID or the Linux benchmark. I will test only the hard drive, not the RAID partition.
Here is a comparison with a Samsung disk using the standard speed test built into Linux. Hitachi sell 2 GB drives spinning at 72 RPM. Samsung sell 2 GB drives spinning at 5900 RPM but not at 7200 RPM. I used a Samsung 1 GB 7200 RPM disk for comparison.
|Read average MBPS||100.4||120.3|
|Read minimum MBPS||63.2||78.2|
|Read maximum MBPS||128.6||146.8|
|Write average MBPS||93.9||76.9|
|Write minimum MBPS||60.0||17.3|
|Write maximum MBPS||117.2||97.0|
|Access time ms||13.9||14.1|
The Samsung disk has similar a seek speed (access time), a slightly lower write speed, and a higher read speed. In most cases you read more often than you write and the increased read speed will help you more often than you will be slowed down by the decreased write speed.
That seek speed (access time) is a killer compared to more expensive disks. The Samsung, at the same price per GB, has the same seek speed. The disks will be fast for continuous reads of large files and slow when the disk head has to jump about finding lots of small files or rows in a database. Linux applications often use hundreds of small configuration files while others update to using an SQLite database, making Linux slow on a disk with a slow seek speed. Windows based applications tend to use one large configuration file and survive better on slower disks.
I tested two disks in a RAID 1 array, where the read speed should be twice as fast, but the Linux speed test used only one disk in the array. If you test RAID array speed, make sure you are testing the combined array speed, not the individual disk speed. Writes should be the same as the individual disks and reads should be twice as fast for RAID 1 then faster again for RAID 5 depending on the number of disks.
As a comparison to SSD, solid State Disk, an OCZ Vertex 2 SSD produced an average read rate of 268 MBps from a minimum 207 MBps and a maximum of 276 MBps with an access time of only 0.2 ms. SSDs really speed up start times when you are reading lots of files. SSDs have slow write times and can be slower than magnetic disks, making shut down times slightly longer.
User speed feedback
The finished system is incredibly slow compared to almost identical hardware running Windows, Samsung disks, and not using RAID. The result is either Linux is slower than Windows or Linux RAID is slower than non RAID or Hitachi disks are slower than Samsung disks.
The comparisons use an SSD for the operating system partition and magnetic disks in a RAID 1 array for all user and data files. In the Windows system, many configuration files and temporary work files are incorrectly mixed with data files and end up on the fast system SSD. In Linux the files are scattered in a different way and may produce a different usage pattern. The file location depends on the applications and I run the same applications on both systems, making the file location variations small and probably not significant.
I know Linux is faster in some areas than Windows and slower in others. The Linux on Hitachi is slower than Windows on Samsung in every area but a similar Linux system on Samsung disks is often as fast as Windows on Samsung and occasionally faster. I suspect the Hitachi disks are a big contributor to the slow down.
I know RAID is faster than non RAID (real RAID, not RAID 0 or some other fake). These computers have far more power than is required for the small overheads of RAID 5 and I am using RAID 1, which has almost no overheads. The problem is not RAID.
That leaves the Hitachi speed.
User noise feedback
The Hitachi disks click click all day. They click at an annoying rate and volume. The click is typically a badly designed disk head returning to base. There are lots of complaints about Western Digital disks annoying people with the same fault.
I tested an open case with one Samsung disk, one Hitachi disk, one standard AMD processor fan, and a standard Antec case fan. The Hitachi disk produces a grinding rumble noise greater than the combined noise from all other sources. Replacing the Hitachi disks with Samsung disks is a great leap forward for any computer that will be in the same room as you.
The test room has four main noise sources, the open window, the laser printer, the CD drive, and the Hitachi disk. The laser printer makes the worst noise and is only on for an hour per week. The CD/DVD drive is used for 20 minutes per day for a backup. The open window lets in the fresh air, the noise of birds singing, and the occasional car noise. The constant rumble and clicking of the Hitachi disk is something I can remove easily.
The Hitachi disks were not the cheapest option when I started looking, then one source put them on sale for a week, making the Hitachi disks a few dollars cheaper. After saving $5 per disk, I would happily pay an extra $10 or $20 per disk to remove the noise. I am replacing the Hitachi disks with Samsung disks today.
The minor differences in cost between equivalent models of Hitachi, Samsung, Seagate, and Western Digital are too small to worry about when shopping for your own system. Nether Samsung or Seagate have a direct equivalent to the Hitachi disk HDS722020ALA330. The closest Samsung 7200 RPM disk is 1.5 GB instead of 2 GB. The cost per GB is the same. I do not need the full 2 GB. The Samsung 1.5 GB or 1 GB could be the replacement in any of my computers.
If you need the full 2 GB and cannot fit a RAID array in your computer, consider the expensive Seagate 2 GB 7200 RPM disk or the Seagate 2GB 5900 RPM disk or the Samsung 2 GB 5400 RPM disk. Seagate disks are almost as quiet as Samsung and way ahead of Hitachi. Western Digital sells many models with different characteristics and some have the same click noise problem as the Hitachi.
I am switching back to Samsung disks for a peaceful work environment. The Hitachi disks will go in the rubbish bin.