Phoronix: SilverStone HDDBOOST
While SilverStone has long been revered for their range of uniquely high-end computer enclosures like the Fortress FT02, Temjin TJ10, and Sugo SG04 all with original designs, occasionally they have dabbled with other products outside of their computer enclosure and power supply expertise. We previously have reviewed such products like the SilverStone Raven mouse and an RFID-secured SSD/HDD enclosure, but their newest peripheral in this area is by far the most unique product that we have encountered from SilverStone. The SST-HDDBOOST product allows you to connect a solid-state drive and a traditional hard-drive via their custom PCB to experience the benefits of both types of storage.
Actually I think the reviewer in http://www.phoronix.com/vr.php?view=14834 took some assumptions that led to testing the device in the wrong way.
I will not go into many details, since the FAQ on silverstone site has most explanations.
Baseline is that the HDDBoost serve to speed-up a rather slow, high capacity hard disk, with the help of a cheap and small SSD.
To achieve the speedup the SSD is used as a cache for the HDD and the HDD itself is used for writing and for fetching cache misses.
Caching is done in a rather dumb way. The device barely mirrors the first part of the HDD onto the SDD.
Consequently all of the burden is laid onto the user who is responsible for maintaining the disk layout appropriately.
It is indeed a cheap hack meant to breathe a bit more livelihood into an otherwise not-so-snappy hard disk.
It is not a booster for already fast devices as the Phoronix review implied. In fact the device can (and will) act as a bottleneck in such environments.
But let me tell you that under ideal conditions it may indeed give some much needed speed-up and by that I mean that it can improve reads over the HDD alone, by reading cached data from the SSD (which is usually faster than the hdd), and it can also "improve" on the "quality of writes" by writing data to the HDD alone vs also to the SSD, which is deemed to potentially be faster and should also save the user from the limits and hassles of the el'cheapo SSD.
For windows this may indeed provide perceptible improvements, but for linux it can't work well.
They provide no utilities for linux and Linux has no "defrag" Tools, so say hallo to moving data to the cached zone (aka the front of the disk).
Ultimately Phoronix is right. This device is mostly useless on linux. But for completely different reasons.
Shit, is this a joke? I guess they don't know a thing about modern file systems, that's why this thing performs worse than either a standalone SSD or an HDD
blaming Phoronix to not get the idea of this device is stupid, because itīs a shit idea. It mainly combinies the disadvantages of both world. you have to pay much money for a ssd (even if its a cheap ssd you have to buy harddisk, this device and a ssd) you gain no more file-safety, maybe it gets even worse.
It is maybe an small performance advantage in some rare cases with one of the many filesystems on one operating system with 1 of 100 harddrives which is really slow, but so itīs a very special device who is not for normal consumers because they cannot check if the hardware they use will gain some speed from it.
My advice buy a Harddisk and a SSD invest the 50 bucks for this device in a bigger ssd, install the operating system(s) and the programm on the ssd the big stuff (games, mp3s, movies...) on the hd and use the advantages of both worlds and use the devices where they are best at.
I think Michael did a good review this time. The performance penalties are real and they must be pointed out. This is a piece of hardware where PTS works well, although some added real-life measures (boot time, open office startup time etc) wouldn't hurt.
If the device has to be used in very specific ways to gain performance, and if michael (with above average technical knowledge) cannot figure it out, how is the average end-user expected to profit from this thing?
I agree that the product might not be that interesting for Linux (see end of my post), but...
I think the test is basically unfair, what's the point of using such a device for postgresSQL or AIO Write. The main advantage of the product seems to be reduced boot times (as your system filesystem will be on the SSD).
anyway, one could install / on the ssd have /var on the hard disk /tmp as tmpfs.
I WOULD NEVER-EVER BUY THIS DEVICE FOR MYSELF
TO ME IT IS WORTHLESS, IN FACT I EVEN CONSIDER
IT A DUMB IMPLEMENTATION OF A DUBIOUS IDEA
STILL I THINK IT DESERVES TU BE JUDGED FOR WHAT IT IS
***NOT*** FOR SOMETHING ELSE
you are completely RIGHT!
No one is blaming phoronics. Still the review was tainted by misunderstandings.
The device may indeed be crap, but this assumption does not necessarily call for a wrong review.
PLUS, this device **CAN** be useful to someone and if tested right, it may show! Thus helping those few (windows users) in their choices.
@rohcQaH and the others.
The reviewer did not get it right because he did not RTFM.
The device does what its documentation says, not what you wish in your wildest dreams.
You just read the website and you'll see that the scope of the device is LIMITED, MODEST:
1) make a ***slow*** HDD "faster" by caching part of
it on a ***CHEAP*** or old model ssd.
BUT IT MUST BE A SLOW HDD TO START WITH.
Do NOT even bother with fast HDDs or high quality fast SSDs
IT IS NOT ITS PURPOSE AND IT WILL ***SLOW*** THEM DOWN
2) It tries to "protect" the CHEAP SSD from write wear.
So writes go only to the HDD
(THIS IS USEFUL ONLY IF THE SSD IS *EXTREMELY* *CHEAP* or *OLD*.
Good modern SSDs can withstand enough write cycles to last as
much as hard disks and they are FAST even in writing)
3) IT NEEDS TO MAKE A MIRROR BEFORE IT ACTUALLY DOES ANYTHING.
THE *FIRST* SYNC STARTS AT BOOT AND IT MAY TAKE **HOURS**.
YOU NEED TO ALLOW IT TO COMPLETE ***BEFORE*** YOU CAN GET
ANY ADVANTAGE AT ALL - ACTUALLY YOU HAVE A MUCH SLOWER
SYSTEM IN THE MEANTIME
YOU CAN CONTROL AND FORCE MIRRORING ONLY UNDER WINDOWS
NO TOOLS FOR LINUX
4) AS I SAID, IT MAY **WORK** WITH LINUX, BUT THERE, IT IS
MOSTLY USELESS, BECAUSE OF HOW EXT2/3/4 WORK.