Your hard drive is too big(stefankendall.com)
stefankendall.com
Your hard drive is too big
http://www.stefankendall.com/2011/12/17/your-hard-drive-is-too-big/
12 comments
Price per GB is a crap metric; price per used GB would be more relevant. If your test data is terabytes, you externalize to servers or lock yourself into only being able to work off a desktop. I also question whether or not a subset of that data could be run locally, and the CI system could run a full stack.
This just sounds like a justification of a bad test environment.
I also can't think of any project that compiles "in a second or two". Compilation and tests usually run build time into the minutes quickly, and time grows linearly, or possibly superlinearly with a project assuming even mild test coverage.
If you're locked into a desktop, you're locked into a network, and you can stick big data on a NAS or centralized server for re-use and testing. Or do you copy and paste the terabyte of data to every new developer?
This just sounds like a justification of a bad test environment.
I also can't think of any project that compiles "in a second or two". Compilation and tests usually run build time into the minutes quickly, and time grows linearly, or possibly superlinearly with a project assuming even mild test coverage.
If you're locked into a desktop, you're locked into a network, and you can stick big data on a NAS or centralized server for re-use and testing. Or do you copy and paste the terabyte of data to every new developer?
The latency and bandwith of a hard disk is unacceptable, but then you suggest using a NAS instead?
Interesting.
Interesting.
$ iostat -kx sda sdb sdc sdd
Linux 3.1.0-1-amd64 (Zia) 12/19/2011 _x86_64_ (8 CPU)
avg-cpu: %user %nice %system %iowait %steal %idle
0.93 0.02 0.72 0.25 0.00 98.08
Device: rrqm/s wrqm/s r/s w/s rkB/s wkB/s avgrq-sz avgqu-sz await r_await w_await svctm %util
sdb 7.41 2.00 1.48 5.06 547.46 67.40 188.21 0.31 46.80 129.57 22.66 7.51 4.91
sdc 7.40 2.01 1.47 5.16 547.20 77.23 188.43 0.33 50.02 164.44 17.34 6.75 4.47
sdd 7.42 1.87 1.42 4.51 546.97 45.87 199.69 0.39 65.07 211.90 18.71 7.67 4.55
sda 7.37 1.86 1.46 4.42 546.57 36.04 198.25 0.26 43.54 122.63 17.52 7.18 4.22
Big builds, etc. pretty much are CPU bound, even on a quad-core i7.
That's true on my single 5400 RPM disk laptop, too (though it has a much slower CPU).
These both run Linux—maybe Windows has ineffective disk caching; I don't know. But please avoid calling people who have actually bothered to measure such things and make decisions on it "fucking idiots".
(Disclaimer: no ridiculously expensive graphics cards here)
The title really should be 'Your hard drive is the bottleneck/too slow, spend money here for an SSD'
My work laptop came with a 5400rpm 750gb drive, however they provide a stipend for hardware/software I need. First thing I did, 120gb SSD, optibay (removed optical drive, put 750gb drive there), and 8gb of RAM (not near as important, especially with an SSD, but still cheap).
My work laptop came with a 5400rpm 750gb drive, however they provide a stipend for hardware/software I need. First thing I did, 120gb SSD, optibay (removed optical drive, put 750gb drive there), and 8gb of RAM (not near as important, especially with an SSD, but still cheap).
Exactly. I've spent about $1600 so far on this PC, I have 6 hard drives and about 7 TB of disk space (plus 2x1TB external drives). One of those drives is an SSD, so all my programs launch fast and the documents I'm working with all open and compile quickly.
I think what the title meant is not "Your hard drive is too big. Period." but rather "Your hard drive is too big, you won't need so much space, so why don't you go for less space but much more speed, for the same price?".
Purchasers need to get their heads out of their asses. It's awesome that you got a stipend, but I've born witness to the following phenomenon.
"I need a $300 SSD." "No, here's another $1200 desktop instead. Does that work?"
Until every developer is crying out in revolt, companies and individuals will keep making this same mistake. It is unacceptable to throttle a Core i7 with 16GB+ of DDR3 RAM with a 7200RPM hard drive for costing concerns.
"I need a $300 SSD." "No, here's another $1200 desktop instead. Does that work?"
Until every developer is crying out in revolt, companies and individuals will keep making this same mistake. It is unacceptable to throttle a Core i7 with 16GB+ of DDR3 RAM with a 7200RPM hard drive for costing concerns.
Yes, I am a developer. I'm also a photographer. I shoot in raw. Therefore this argument is invalid. When Macbooks can have dual hard drives, then this will make sense.
I have burned through 8 SSDs in VM hosts. Using TRIM. I am not ready to trust them yet.
A 1TB, 7200RPM drive costs like 100 bucks nowadays. At that price being able to store large amounts of crap is hardly a giant waste. Half the folks I know have burned that much money on collecting cases for their iPhones. For that matter, it might not even be necessary. Me, I've got a couple big disks, because my test data sets alone consume nearly a terabyte.
Second, I agree that having an SSD can be really beneficial. Which is why I got one. But that doesn't imply that the presence of a 1TB drive is harmful. It's absolutely OK to have both. Even a good idea, since they have different strengths. SSDs certainly have great bandwidth and latency, but price per gig is another metric that's worth considering.
For that matter, if most the work you're doing is CPU-bound and the software you work on already compiles in a second or two, there's a chance that an SSD doesn't really provide much value anyway. Booting might take longer, but any reduction in boot time below how long it takes to get my morning coffee is meaningless.