Top game server hosting, 3rd party security, free SSL/TLS certificates, cloud-based VPS plans, cheap dedicated servers
We will review those customers very carefully to identify the real hard drive manufacturers, not to mention the honesty and transparency of the “Do you work for us?” requests from customers for warranty support. The shady-sounding boutique companies and unscrupulous affiliates might attract many consumers, but these business practices are unlikely to attract long-term success.
Second, we will ensure that the drive we use is completely reliable and supported with no hidden or unconfirmed issues. Finally, we will compare drive prices before making a purchase decision, including everything but the specific unit that we are talking about. But to avoid any confusion, I will rephrase the above last sentence as follows: We will use drives that we test thoroughly and that we feel safe supporting for the long-term. We do not make assumptions about drive pricing when talking about desktops, notebooks, and other physical machines. OK, back to your original question: Which size of SSD do you use in your own system?
I’ve been using a small SATA SSD in my testing systems, mostly because I have trouble managing capacity beyond 4GB. My testing of all other types of drives has been with 240GB drives that are permanently filled with data, without any kind of active or background stress.
Therefore, my focus for reviewing drives has been to compare performance with or without any files and folders being written or written and cleaned up on a regular basis. That has been a somewhat hard standard to get right; we test drives for a set time, and we end up with an “I’m done, I’m done” report, while the drives we tested before this time would be left behind, with some of them being permanently left behind on storage management (just like a RAID array).
We use a set of SMART (Sensormarking Methodology for Relative Accuracy) reports to stress the drives during the tests, but we do not stop the tests after they reach the expected SMART limitation and when we end up with consistent final report results. We go through and test the drives as thoroughly as possible in our testing, but we do not keep our sights on the absolute end of endurance in the MLC drive. Instead, we aim to stress and stress hard with as much workload as we can in our testing, so that we are confident that the drives will handle a significant amount of stress going forward.
That stress does not, however, come free with any MLC drive. So, what we aim to do is create a product review that shows both the maximum endurance we’ve seen on the drive and how we went about testing that endurance. The objective is to make sure that there is a good conversation going on between the review, the vendor, and the purchaser.
Sure, when we begin writing reviews about drives we look at the specs. We look at benchmark numbers to assess performance and capacities. We use that information to discuss what we feel comfortable putting in our hands. But as we get into our in-depth reviews, we learn what works and what doesn’t in terms of writing and end-of-life capabilities, and we incorporate those aspects into our reviews.
For example, one of the MLC drives we have tested has a Z- endurance of 128GB and we know that will stress the drive pretty hard. I wanted to know if it would write its contents at the maximum SMART metric of 100 MB/s, or if that SMART value is due to something else entirely. So, I scheduled several stress tests to see what would happen.
The results were very interesting. One of the drives would write its contents at close to the maximum SMART metric over time, but would show a sequential write rate well below 100 MB/s, which is the speed we are looking for. That drive is the Kingston SSD600 Pro, which we reviewed just over a year ago, but we are confident that the same issue exists on other drives. A more recent drive, the Samsung 830, has the same issue.
We also understand that the SMART standard is not a perfect metric. It may take some time, but a drive that is SMART qualified could end up having an SMART report that shows a read and write speed that is slower than the SMART specification allows. The fault, however, would not lie with the drive. While the extreme SFF power consumption is expected from a MLC drive, I wanted to see how many MB/s it would write in sequential write test, since the traditional benchmarks did not have a number of writes to the Western Digital 830 (8 drives for power consumption testing were necessary). The results I received for sequential write were inconsistent with the other drives in my testing.
The Real Story
I ran a few tests to determine what was going on, and these tests all reflected and confirmed the data that I had seen in the sequential write tests. The 830’s sequential read and write speeds were consistent with what I had seen during SMART testing.
Post a Comment