A tale of two clouds
Over the past few weeks I’ve been kicking the tyres on two new(ish) entrants to the IaaS space. Both services are still in beta.
Savvis Virtual Private Datacenter
I first came across this back in June at the Cloud Computing World Forum, and I signed up straight away for a trial. Sadly there was some kind of issue with the first token that they sent out, and it wasn’t until a few weeks ago that this got fixed.
To be fair to Savvis it’s worth pointing out that VPDC isn’t meant to compete head on with the likes of Amazon Web Services (AWS) etc. It’s clearly a bridge product between dedicated hosting and public cloud for those who still have concerns around security and privacy. From a contractual and delivery perspective this makes the whole thing less ‘on demand’ than I’ve become accustomed to.
The main point of contact for the user of VPDC is a Flash application that’s used to design and deploy hardware architectures. I’m a Flash hater, so this was a bad start for me, but as Flash apps go it wasn’t too grim. The experience was somewhat reminiscent of early Sun N1 demos – drag and drop machines, click to configure, wire them together.
There are a few things regarding the firewall config that I could whine about, but at least it has one.
It was when I first hit the deploy button that disappointment really set in. AWS has made me expect (near) instant gratification, and VPDC doesn’t do that. I didn’t time the first deployment (of a single machine ‘essential’ configuration), but I had plenty of time to get on with other things. I did time the second deployment (of a three tier, three machine ‘balanced’ config) and it took over an hour. I guess an hour shouldn’t matter so much when you’re buying by the month, but it just felt wrong.
The next disappointment was the sign in process. Something else that AWS has spoiled me with is key based SSH that just works. VPDC is a bit old school, and emails out machine/usernames and passwords. Quite annoyingly it doesn’t bother to put the FQDN or IP into those emails, so you have to go back to the Flash app to figure those out from machine properties. The emails also say that a password change will be forced, but that isn’t the case.
Once I got this far I had a bare machine running Linux (Windows is also on offer). There isn’t (yet) some menu of middleware and persistence tiers that you can browse. Once again I suppose that the build of such machines from scratch might be a small issue if you’re keeping them for a long time, but Savvis desperately need to offer some choices and automation in this area (and something like CohesiveFT’s ElasticServer would be a great place to start).
Overall the trial served it purpose of illustrating to me the capabilities (and limitations) of the platform. I don’t think I’ll be rushing back with my company chequebook in hand, but I think I get what’s going on here.
I saw some cynical reactions on Twitter when Brightbox first announced their cloud platform as being a UK first (along with some mistaken impressions that AWS had UK based services), but I was intrigued, and signed up for the beta.
Brightbox have not taken the path of least resistance, and just thrown up something based on Eucalyptus (and the AWS APIs). This could be a problem in the long term as APIs settle down into standards, but in the short term it’s quite exciting to be able to try something new.
Brightbox’s background seems to be firmly rooted in the world of Ruby hosting, so it should be no surprise that their command line tools to access their API come as Ruby Gems. I ran into some trouble trying to get things going on my old Ubuntu 9.04 test VM, so I eventually caved and brought up a new 10.10 machine on which the CLI tools installation went without a hitch.
With the CLI in hand the getting started guide had me up and running in no time. It could be just a subjective opinion, but my sense is that the Ruby CLI runs faster than the Java based EC2 equivalents – it probably comes down to the startup times of the various language interpreters. The machine that I got felt much like an EBS backed small instance on EC2, though the storage usage pattern is quite different (no S3 here). I liked the snapshot facilities a lot (though it’s not a feature I’ve played with on AWS, so I can’t compare), though I did hit some trouble when I tried to use a snapshot that wasn’t completed. The exciting news on this front is that Hybrid Logic will be putting their beta within the Brightbox beta, offering fine grained snapshots to provide fault tollerance and scalability (check out the webcast demos to see how cool this is).
My one gripe about Brightbox at this stage is the lack of a firewall (besides IPtables on a machine). I feel like things are just a little too vulnerable if left out there on the Internet without something to filter out the lumps – it’s all too easy to configure extra services on a machine that increase the (unintended) vulnerability surface area. On the plus side they use public key authentication for SSH into machines.
Overall I must say that I liked my (so far) brief experience of Brightbox, and I look forward to seeing what pricing looks like and how the service develops.
Filed under: cloud | 1 Comment
Tags: API, aws, Brightbox, cloud, Hybrid Cluster, Hybrid Logic, iaas, Ruby, Savvis, snapshot, VPDC