Compared with the traditional hard drive, SSD has many attractive characteristics, which makes them more and more competitive. The energy consumption of the SSD more small, random access to read mode faster, with standard hard disk interface standards (such as SATA). Can say the emergence of hard disk SSD makes traditional almost to the end of fate. Standing in the DBA's point of view, SSD read speed is high speed of its biggest advantage, because it to solve the I/O bottleneck problems played a crucial role.
But on the other hand, SSD is not perfect, especially on the SQL Server database, the following several reasons for it will usually allow DBA flinch. First think must be cost, they brought about by the data throughput speed upgrade whether can be worth it? A enterprise managers in the processing has more than one hard drive storage system, the concern is not only pure performance issues, they are more interested in each spend a penny performance can improve how much. If you use the cheap ordinary hard disk and at the same time and can solve the bandwidth, get performance can't bad arrive where go to, why choose to SSD? In the use of SSD, you may want to spend 10 times as much money to buy, then you need to ask themselves, whether performance but also improve 10 times? The answer, of course, is often can't, then I suggest you use normal hard disk.
In 2009 by Microsoft research institutions release called 'server storage to switch to SSD: analysis and balancing' series, analysts think that in nearly a period of time, for any test server scene, SSD drive all is not the best choice. 'Only SSD 3-3000 times the price promotion after, it will be possible to become a real traditional hard drive replacement,' analysts wrote: 'SSD as the value of the middle layer cache is very limited, according to various work load tests, and less than 10% of the scene that use the SSD get be rewarded.' SQL Server database is not within the scope of this 10%, also we a 5000 users of Microsoft Exchange Server on the SSD test (using the embedded database), and the results proved not suitable for investment solid-state hard disk.
Another SSD for discussion of the problem is that it is more reliability, namely can use for a long time repeatedly: flash unit can withstand repeated the writing? The same discussion in the U dish also has appeared when the markets, but we are now focused on is the enterprise applications, with individual consumer electronics is different, the enterprise application of I/O total than individual consumers, especially considering the database such I/O intensive applications. Data for the importance of an enterprise without obvious now, stability is the top priority. Therefore no one hope the new technology is the price of the data in the time bombs bound.
Of course if we will reality and theory you can find that in fact contrast SSD long-term use of problems and not so serious, besides a good design can largely ease the problem. SSD market analysts Zsolt Kerekes himself on the issue to a research and came to the conclusion that the: in a design good flash SSD, you may need to fill the entire hard drive would appear said above problems. So even if the database such application, of which contains a large number of write operation, but to repair the SSD would not cause for use threat.
In view of the above situation, caused by long-term use of write operation problems actually does not cause too much trouble, it within the service life of yes and will have a new SSD technology, the update, faster and higher, more energy saving capacity model will occupy the market.
Of course, at present the SSD market is in constant innovation, although the price in the short time won't cut too much, so if you want to be in the database system present spending in the tens of thousands of funds for the SSD drive, it would be better to spend the same money to buy other database hardware. Such as increased memory reduce the load of the I/O, in contrast, such SSD drive more cost-effective than to buy. If real scene have too much I/O not to reduce, then decided to purchase SSD is not late.
From Microsoft engineer James Hamilton published a series of formula, can help users buy calculation SSD is worthwhile, to determine replacement equipment storage return on investment. In the formula (reference link), with it to a database server as test cases, in his discovery, we can see in the back and forth between random disk I/O is dragged other I/O after the main reason, so he decided to use SSD to replace the original storage devices. But still as before, the return on investment is to become the largest problem of his formula given that use of the scene is not suitable for replacement SSD.
Although SSD development is very fast with the trend of replacing the traditional hard drive, but in the enterprise applications, especially the database environment (such as SQL Server), the high cost and the low return on investment are the two serious problems for SSD. So only when work loads or SSD price is lower, it just may be qualified substitute.