As our resident half-programmer mentioned, it's been brought up before and at our scale would be an awful lot of work for the server. It's already a lot of work for the server that it's having to do this at the board level. At the post level? Fuhgedaboudit.
Especially since we'd have to make it most *unique* reads to circumvent most of the manipulation that already occurs with board reads.
However, as someone else pointed out, we do have a bigger/better server now. It'd be a comparatively easy (but still low priority) thing to add so we can quantify just how big the workload would be, and set it up so it can easily be manually or automatically shut off if server utilization gets too high.
(still thinking out loud)
The new server is running at about 15% of its capacity during market hours where the old one was more like 60% with too-frequent spikes much higher. So, the horsepower definitely exists. Today. I have zero doubt it could handle a workload like that. But would make it sooner that we'd need something more powerful again.
After we've stabilized the current server (there are concurrency, locking, and isolation issues still to be addressed) and gotten some other important projects done, we'll give it a try and see what happens.
On a side note, though Moore's law has been dead for a while, a separate law whose name escapes me is still in full effect.
Less than 2 weeks after buying this server, I got an email from Dell that a machine that was way outside the range I'm willing to pay for a computer is on sale.
Had we waited, we could've gotten roughly 250% more power for 50% more money. AARRGGHH!! We paid around $10k for a box with a couple of dual-core Xeon 3.0's. For just over $15k, we can now get one with FOUR dual-core 3.4's! Previously, the price jump from 2 to 4 processors had been enormous!
In the not-too-distant future, I'll be replacing SI's box, though it's seriously underworked. It's old and the case is ridiculously large. We're planning to convert to exclusively rack-mountable machines (in a Dell cabinet and using proper rails instead of shelves), so all the tower machines will get removed/replaced and brought to the office as development boxes (I already dib the adserver as my personal desktop machine, so back off, Matt and Dave).
I can probably make a good case for getting the quad-processor box, moving iHub's db to it, and moving iHub's new db box to the SI db role. Then we'd have the future very out-gunned on both sites, as well as having enough power on each db box to step in to db-serve BOTH sites if one of the boxes breaks.
So, (end of essay-length thinking out loud) we'll try what you suggest when we get a chance and when we get more important things taken care of, and see how it works out. The cost of trying it would be very low in terms of geek time. And we'll be able to quantify the cost in computing horsepower to see if it's worth it.
But the fact that you see just such a feature (among other features) on small sites running off-the-shelf software is one of the reasons they can't ever become big sites. Some things simply aren't do-able at our scale at a reasonable economic and horsepower cost.
AI
