We haven't been shy about acknowledging our shared hosting performance issues this past year. It all goes back to our core storage architecture, a SAN-attached NFS server. The short version is that there's too much reading-and writing happening too fast for this architecture to provide exceptional performance 24/7/365. Backups, which need to inspect every file on the storage system for changes, are taking more than 12 hours to complete now, degrading performance along the way.
While there can be many reasons for perceived slowness, we know our storage system performance is less than perfect from time to time, and we have seen the solution, in our mind's eye (and engineering whiteboards): Clustered storage. We'll be able to add machines and add storage and add read/write capacity simultaneously, in a single, seamless, migration-free step. We've been working on it for about a year now. We have a ways to go still, and I'll post an update in June on the progress.
In the meantime, we have known that we really need to do something about this -- and today we have. We've deployed additional storage hardware -- not as a permanent fix, and not even as a step towards the new system, but as an 'overflow' system to which we can move a handful of busy accounts.
One of the first we've moved is the webmail system inside OnSite. If you use it regularly, I think you'll find it considerably faster today than it was a few days ago. In fact, all of OnSite should feel faster now due to software upgrades we deployed last week.
We are still ironing out some quirks with this 'overflow' hardware, but next week we should be able to migrate a small number of additional sites to this system. If you've previously contacted us about performance issues with complex, template-driven sites, let us know by emailing feedback at modwest dot com that you're interested in having your files moved to the new system. We'll handle everything, and you won't have to do anything different, but the move process can cause a site to be unavailable for up to an hour or so, depending on the amount of data and number of files.
This isn't a permanent fix, but it should help mitigate our recent less-than-stellar filesystem access times occasionally experienced. And again, the longer term fix will solve the NFS-performance problem once and for all.
So, short term, we sent our request for more power to the engine room, and we've managed to eek out just a bit more. If only we had a dilithium crystal... more news soon.