Last Friday, we introduced a $125 container that includes 1 GiB RAM and 10 GiB SAN storage
Like all of the containers, adding more bandwidth, RAM and space ($1/GB/month for distributed storage) is a simple utility.
When we introduced the containers in July of 2006, we started with containers in single setups and at higher price points for a few reasons: we had some large customers wanting such a solution and that’s the size and prices that they wanted; we also wanted to keep the number of them down a bit as we refined a few things.
Namely, how we were going to manage 1000+ pieces of third party software and libraries across many systems, how to do distributed, resilient block storage over standard gigabit networks and cat6 cables, and how to take advantage of the fact that we’re now in multiple datacenter locations (sorry, all are still in the US). All of it geared towards the ability to scale up and scale down as needed.
What are some of the things you can expect in the very near future?
- For a number of services on our private network to be made available (just imagine what I can think of) to customer applications on containers
- For intelligent, pre-defined clusters to appear.
The clusters are particularly exciting because they come from the growing experiences with our own applications, are closer to things from my own background and are the setups that’ll impact people from shared hosting on up.
9 Comments
I used to be with textdrive on bidwell, when it was having tons and tons of problems. If I get a Container and the machine it lives on turns out to be a lemon, is it pretty easy to be moved to a different machine without downtime, the same way Xen guests can be migrated between hosts?
Yes evariste, that’s correct.
Great…thanks, Jason!
Another question, and this might be a slight pain in the neck: can you guys run unixbench on a $125 Container so I have an idea what kind of performance to expect compared to my current Xen-based host? It would be useful to have some benchmarks to help decide what to do. I really like what you guys are doing in general and I want back in.
One more question: any chance of you guys offering a Parallels image of the Solaris build you’re installing on the containers, so I can get familiar with administering it from the comfort of my MacBook Pro before making the leap? I’m more used to administering Debian servers, and I know Solaris is quite different.
Yes we’re thinking about a parallels image.
Sweet.
evariste: At least the last time I looked (last year) UnixBench was an awful benchmark. Many benchmarks it does wind up mapping to “how fast can we fork?” since, on a modern system, it takes vastly longer to fork than to take the square root of a big number (which is one of its tests). As such, in my experience, the benchmark skews towards measuring workloads that don’t exist in the real world.
Solaris Containers generally speaking induce no performance overhead, so you should go roughly as fast as you would on the native Solaris box.
dp-thanks for the insight! What would you say is a better benchmark, in that case?
Consider this another vote for the parallels image. (It’ll be the final straw that makes me go buy a Mac Book Pro!) I’d really love to use one of these containers for some of my web apps, but I’m wary about the learning curve of switching back to Solaris—last time I used it in anger was Sun OS 4.1.4…
Unless there was a way to have a wee play around with a container before committing to it…?