Welcome to Performance Cloud Servers – Have Some Benchmarks
developer.rackspace.com
developer.rackspace.com
> Pick a language framework, pick a web framework, heck, pick a web server or entire application we can throw apachebench at with one of the 120GB monsters we have and we’ll put it together
We've asked Rackspace to run our suite and the introduction [1] still contains the following sentence: "How does EC2 compare to, say, Rackspace Cloud? We don't have the data now, but if you have a Rackspace Cloud account and are willing to run the full test suite, we'd like to be able to render that here."
We'd love to see Rackspace Performance Cloud as another hardware tab alongside our current i7 and EC2 results.
Separately, we had a contributor who put together some Windows Azure scripts but we've not heard from him for a while and I think we need someone else to volunteer to pick that up.
The i7 hardware we test on is just our workstations, which are possibly similar to your Performance 1 8GB, but you would probably be able to map it more accurately. Specifically, they are Sandy Bridge i7-2600Ks with 8GB of memory.
Their current ("next-gen") cloud offering has had serious noisy-neighbor issues with networking performance, and the maintenance to attempt and resolve these issues architecturally caused some serious pain during two consecutive maintenance windows. ( Hour+ downtime for dozens of nodes. )
Management API availability and reliability also leaves a lot to be desired.
Something I think that Linode has done really well with is making these next-gen type upgrades available across their entire range of instance sizes. EC2 and Rackspace both created a separate (expensive) tier that you have to buy into, whereas with Linode I just woke up one morning and had 2x the RAM, 2xThe cores, better disk speed, etc.
Different companies, different strategies, but I do wish some of these new instance types would trickle down.
"We eliminated the lowest end flavor (512) and will start with the competitively priced the 1GB flavor, while drastically increasing its performance from the previous 1 GB offering (we have charts!). This means that the price per month for a 1GB instance is now $29.20/Month ($0.04/Hr). All together? The performance you get for the new prices across all of the Performance Cloud Server flavors is amazingly competitive and a great improvement to the Rackspace portfolio."
Are these charts on the previous and current performance for the 1&2GB instances visible somewhere?
Are they grandfathered at that pricing? And for how long?
Can new customers even get that low end server today?
The blog post says "eventually" but some others have pointed out that the pricing is missing today.
Today, the Performance flavors are available in IAD only (additional regions coming soon). The Standard flavors continue to be available in all regions (IAD, ORD, DFW, SYD, HKG)
See also: http://www.datacenterknowledge.com/archives/2013/11/05/racks... for a good review.
That aside, I'll have to give them a test run. I was happy with rackspace cloud servers before I switched to linode and now digital ocean. The only thing that has had me switching is pricing, but if these perform well, I may use them for certain applications. I'd love for a 3rd party to benchmark these vs digital ocean so we can compare price and performance.
Ideally I want to get the code in place to empower others to run the same set of tests on the host of their choice, but I personally won't participate - just like the fact we don't comment on, or take advantage of others' downtime[1], I don't want to get in a shooting match, especially as I represent one of the vendors, and can be accused of bias.
[1]: https://github.com/rackspace/social_media_guidelines/blob/ma...
AWS instances generally are a bit lower in performance than rackspace offers.
Disclaimer: I work at exoscale. https://exoscale.ch
The new so called performance cloud servers are in a price range, that calls for a dedicated server at Hetzner or OVH.
Sorry, no bounty: Xen should be cheaper then dedicated, and not crippled. I wont recommend Rackspace. Not even to my competition.
I'd be interested in how you test your servers and what methodology you use to make comparisons in performance and pricing. It's an issue everyone in the industry will need to address as compute moves toward becoming commodity.
About pricing: $5-$20 is a good price range for Xen. $25-$100 is a good range for dedicated servers. $100++ calls for colocation. The 'cheapest' of your performance servers compares to a cheap Hetzner or S4U dedicated server, that allows to run Linux Containers.
And most important: !NEVER! bundle hosting contract with domain name contract, so you can switch easily, if your site hits traffic.
My current server is Win 2008 1GB.
http://www.rackspace.com/knowledge_center/article/migrating-...
$2000/mo for a 60G box...
For that money you can buy an equivalent server every month.
And don't get me started on $4000/mo for 120G, that's almost two equivalent servers. Every month.
I understand why their costs might be higher (excellent support staff, always had stellar support from RS), but it doesn't really justify that much of a difference.
The advantage of a "old" hosting company is that it's less likely to go out of business in a heartbeat, it's not necessarily enough to convince me in paying more for the same services that I can get elsewhere.
Also Linode is a security nightmare and Hetzner doesn't always have instances available.
To be fair, Rackspace could also be a security nightmare just waiting to be discovered. For example, as of about a year ago, the way Rackspace generated the initial root passwords for cloud instances was by taking the hostname and appending a bunch of random characters to it. Unfortunately, they put a DES hash in /etc/shadow, so only the first 8 characters mattered, most or all of which were taken up by the predictable hostname. And yes, root logins over ssh were on by default. When this was reported to them, they initially denied it, and only acknowledged and fixed it after much wrangling and, in my opinion, an excessive amount of time.
That's clearly the v1 (Sandy Bridge) E5 2670 that came out in March 2012, with 8 cores per CPU, 16 cores per machine, and 32 logical threads.
The v2 (Ivy Bridge) E5 2670v2 came out back in September with 10x 2.5GHz cores per CPU, or 20 cores per machine, and 40 logical threads. As well as having slightly improved performance/clock due to the new architecture.
Not that it should really matter; once you get up to the prices of the Performance 1 4GB or 8GB, bare metal makes more sense. An E3 1240v3 has 74% the performance of an E5 2670v1, and would obviously run circles around a 12.5% or 25% share of the E5, without even taking the I/O hit of running virtualization into account.