A very simple feature with complex inside.
I want when a user input tag, I give suggestions. For example, when type a, I suggest ASP, ASP.net, Apache etc etc.
I have (or will save) user inputted tags. I have two options(places) to save them.
- SQL Azure
- Azure Table Storage
- or Both
About Data Structure, I am not sure yet. But I guess <tagText> <tagUsedTimes> should be essential. For the <tagUsedTimes>, no need very accurate (since its just a suggestion). And for <tagText> I don’t know if I should save it literary, but if I do, I have to compare text by .left(length)=, will not perform well.
The scale. You can get an idea from how many tags have been used in stackoverflow from millions of posts. And not like stackoverflow, user can input any tags, not limited to the existing tags (I found I can only create new tags after I have 1500 reputations here).
Please point me a way to correctly construct this.
[EDIT]
To be clear, I am not providing this feature now. But I plan to save all data in a correct structure for my next step. So, I like to know any reasonable way to do it
UPDATE
for <tagUsedTimes>, I have a way since its deployed from start up. I can +1 for exist tags when a new entry( or article etc) saved, and create unexist tags and set <tagUsedTimes> to 1. For editing old entry, just -1 for removed tags…. May not accurate for long run, since may have error here and there within times. But I guess the number is enough for tag popularity.
Table storage works fast if you know exactly what partition are you going to use. If you are going to use the first letter to define the partition could work, but sooner or later you will end with a huge partitions and the performance will be affected.
Sqlazure working with shards could be a effective solution for this case and gives to you the hability to easily move the shards providing a very scalable solution.
In addition, Table storage takes few seconds to setup and have more latency than Sqlazure.