If your workload requires you to have a small percentage of attributes in a table then you can achieve a considerably efficient I/O with exemplary ease. For this, you may have to learn how to take full advantage of the column oriented storage model and make the most of it. Just so you know if you use the column store then only the required data shall be read from the disk and you will be able to achieve the desired results without fail. Interestingly, a column store may not be your best bet if a large number of attributes are involved.
In such a situation, the column store is unlikely to stand the test of the time and may not prove to be as good as the row store. In fact, if you continue to use the column store for more attributes you’d realize that it is less competitive when compared with the row store. This is because if you use the column store then you’d also have to deal with an overhead. The overhead shall come into the picture when you try and combine the separate attribute values into complete records. The worst part is that if you happen to come across queries which require accessing several attributes of a table then you may notice a drastic dip in the performance as far as the column store is concerned.
However, there are situations where it is better not to use the row store. The bottom line is that if you are not sure whether you’ll be requiring a row store or a column oriented storage model then you should think about using a hybrid data store. With the hybrid store, you won’t have anything to lose because you’d have the privilege of choosing the optimal storage model. Of course, while making the choice you may want to consider the given query workload to be sure of your decision. The good news is that whether you choose the row store or the column oriented approach you will be to maximize the performance without any difficulty.
Meanwhile, it is worth mentioning that if you are dealing with simple reporting queries then you can most certainly benefit from the hybrid data store. What’s even more surprising is that the aforesaid store can also come in handy when you are dealing with analytical workloads. Nevertheless, once you choose the appropriate storage model you may consider using the data for analytics and this is where the data appliance shall come into the picture.