Rackspace Ups and Downs
You pay a lot of money for Rackspace managed hosting. You pay for decent hardware in a top-notch data centre, with great Internet connectivity and of course fanatical support.
If your web service demands low latency and minimal downtime, then Rackspace is an obvious choice. For all of the benefits, however, there are downsides which take the shine off:
- Ordering new servers is tedious and error-prone. You need to negotiate a contract, on 12-month terms to avoid high set-up costs or inflated monthly charges, and you should double-check that your delivered solution is what was ordered. Both hardware and software configurations may not be as requested.
- Requesting a configuration change can cause issues, and this appears to be because of the separation of network and system teams. If a network change is made, the network team may not think to check whether the connected servers are affected by the change.
- Server monitoring is entirely your responsibility. There is no pro-active alerting for server load or excessive resource usage; for that you will need a third-party service like server-density. That's not a problem, but don't assume that your servers are being 'looked after' when you're asleep.
- Front-line support staff can actually miss critical server issues, because they are only looking at the specific alert that has been raised.
- Chronic or intermittent problems can be extremely frustrating to get resolved, as your support ticket will usually get pinged back and forth between network and server teams, and will be looked at by a variety of people (who don't read through the issue history).
Now, it's certainly not all doom and gloom. Despite the problems described above, the teams work hard and will (usually) work at it until the problem is resolved. I think that they do care and genuinely want to make the web service provider's life easier. They are typically quick to respond, and a phone call will get your problem escalated if you need.
You can specify your own instructions for support staff to follow, to ensure that you get a customised yet reliable support process; these instructions can also identify the dependencies between your servers. This is really the essence of the Rackspace managed hosting service: it can be great service, but don't assume anything. It's not a free lunch, and you still need to do your own systems monitoring and administration. In this respect it's probably no better or worse than any other dedicated host, but the ability to fine-tune your requirements must be a cut above the rest.
On balance I think that the service is great overall, which makes it all the more frustrating when something isn't done right. Perfect, no - but they're definitely trying!
Add your comment
Thank you for your comments about your experience with Rackspace (my employer). Feedback is essential for us to know how we're doing and how we can do better. While everything you mention is something that needs addressing I want to focus on one point in this brief comment (more discussion of all these points is welcome ~ my email address is robert.taylor[@]rackspace.com), specifically the monitoring issue.
The basic monitoring option for Managed servers is pretty basic: it monitors specific ports for interruptions and is not proactive in determining degrading performance above failure thresholds or resource depletion on the server itself. However, we do have more robust monitoring solutions that can be added to a dedicated account. More importantly, we recognize this issue of spotting potential problems before they cause an interruption. For this reason we recently acquired CloudKick.com, which has deep system monitoring capabilities for a wide range of deployments. CloudKick is being integrated into our systems and support to be a seamless part of our product offerings, yet it can today be used to provide robust monitoring and administration for servers you have with us, other providers or even in house.
Please feel free to continue this discussion. I appreciate your candor and treasure your perspective.
Robert J Taylor
Sr Systems Engineer
Fair point about the monitoring limitations in the standard package, but I think it is worth pointing this out; it is something a potential customer might not expect.