Pricing out some new HDD; Toshiba wins

dvl@

Developer
I've been pricing out some hardware lately. The Toshiba 3TB drives have won. They have a good reputation for using in storage units (i.e. ZFS) and seem to have implemented TLER rather well.

If I was using someone else's money, I've had bought faster drives, but since I'm not....
 
(The link needs trimming before it points to a url). I see no Samsung on that list... I awhile back had bought a 1T which showed only 26 or 33 mb to the bios, ghostbsd.org, etc. It took an obscure on-the-web-somewhere[1] floppy-based interactive modification of the drive's firmware to actually report one terabyte. So that is something to remember if one ever runs across that issue on new large drives...
[1] One person reporting a fix out of hundreds of links...
 
I've decided to order both a TOSHIBA DT01ACA300 and a WD Red WD30EFRX (both about $142 each) and compare them.
 
The only tests I can think of for those two HDD: dd. bonnie++. And maybe: load the FreshPorts database and do some queries.
 
The diskinfo(8) benchmark is easy to use and gives more information than a simple dd(1) from the drive. It's kind of like EPA gas mileage in that you will never see those numbers in real life. It does tell the maximum read speeds at the outer, middle, and inner parts of the disk. It can be run on individual partitions, too.

% diskinfo -tv ada0
 
FYI, I have good things to say about https://www.serversupply.com/. I ordered the Toshiba drive from them last night. This morning I noticed Newegg had the same drive for $15 less. I thought: oh well, I've ordered already, too bad.

This morning, ServerSupply called me to clarify the order. At the end of the call, I mentioned the price on Newegg. He apologized, said he could not meet that price, and offered to refund me. I asked for the refund, which they did. Then I ordered from NewEgg.
 
Tonight I tried the first benchmark on my new server. It's more of a test of the Seagate ST2000DM001 HDD, and LSI card than anything. I'm getting about 170MB/s on a sequential write. I also ran a bonnie++ test.
 
That's ridiculously fast. (You have the dd(1) to a file repeated in the dd(1) to raw device section.)

A 32K block size may be a little small. I use 64K or 128K when testing hard drives, or more recently, 1M for both hard drives and SSDs. Hard drives don't go any faster with 1M buffers, but makes the math easier.
 
I've moved the errant dd to the correct section. This means MB/s when writing to the raw devices is now 157 MB/s. And when writing to the filesystem, it's between 160-179 MB/s.

I will run some tests with a larger blocksize.
 
Back
Top