JP, ...If this upgrade was to be sufficient for the 250k to 500k registered user capacity, the current performance would have to be stellar, ...almost instantaneous. That didn't happen. It's better, and the performance is acceptable. But as some have mentioned, it's not great. That’s unfortunate because it means that if the website popularity begins to grow as it was growing a month ago, the engine will bog down again under the load as it did before. That’s all I know. I can’t give you a date or a number of users. It’s based entirely upon the workload, relative to the quantity of information contained in the database.
Last night, Lexey pointed out another site, which is quite popular, and she suggested that it is slower than Hypster. I was curious and investigated that site. The performance is horrible. And when I went to Alexa to look at their statistics, I discovered that they are dropping pretty dramatically. That site has lost 29% of their traffic in the past three months. The lesson there is that when your performance becomes unbearable to your Internet audience, your site dies. The site that Lexey pointed out is related to sewing and fabrics. That audience would tend (in my opinion) to be much more tolerant than the Hypster audience.
If I were the CEO of Hypster, I’d have someone working on a new design of the internal workings of Hypster now. Perhaps a server with four Xeon parallel processors and as much RAM as is physically allowed by the newest motherboard out there. Serial hard drives configured under RAID 0, to split the data access workload between two drives for increased performance. I’d investigate radical possibilities like a huge RAM storage device to act like a hard drive and cause it to contain the database data for dynamic access, with regular hard drive storage operating under a RAID background backup configuration. I’d talk to the industry leaders about which database manager would be the most efficient for the type of data that’s being stored. A few simple things like pouring the images that people upload through a filter that limits the optical resolution and physical dimensions, might help a lot. These things take time and cost money, …lots of money. But the issue is not going to go away by itself and it would be my guess that Michael is thinking about this very thing this morning already.
When we access information at Google, the response is almost instantanious, and the volume of data that's accessed is almost inconcievable. But they are utilizing countless thousands of servers, running in parallel. Performance costs money.
The question is only whether Hypster is worth that much investment. It appeared a month ago that it was, as the number of Registered Users was beginning to grow with a daily, almost an hourly upsurge in the percentage of increase. Will that begin to happen again? Probably.
