DreamCompute Comes Out of Beta And Undercuts Hourly Competitors By 40% On Price

[DomainBop]

DreamHost's OpenStack powered DreamCompute has come out of its 2 year  beta and entered general availability.  This is actually a nice full featured OpenStack offering compared to GoDaddy's recent OpenStack Cloud launch (which still felt like a beta to me), or OVH's Public Cloud, both of which are still missing important features that many of their OpenStack competitors offer (OVH: real private networking and IPv6 are still "coming soon", GoDaddy" no IPv6, no Object Storage, no ISO's, no extra disks, etc).  While Bob Parsons at GoDaddy was chasing elephants, DreamHost was getting heavily involved with the development of OpenStack which may explain why their offering is more feature complete than GoDaddy's...


My main reason for taking two minutes to post this however is because with this launch DreamHost is shaking up the US hourly billing pricing market with 1GB for $6  in an industry where $10=1GB had become the standard across hourly cloud competitors like DO, Vultr, GoDaddy, Atlantic.net, Cordero, and many others.   (OVH/Online.net/Aruba and others already shook things up in the European public cloud market with their pricing)
 

Flavor Name


Memory


vCPU


Hourly Price


Max Monthly Price


gp1.semisonic


512 MB


1


$0.0075


$4.50


gp1.subsonic


1 GB


1


$0.01


$6.00


gp1.supersonic


2 GB


1


$0.02


$12.00


gp1.lightspeed


4 GB


2


$0.04


$24.00


gp1.warpseed


8 GB


4


$0.08


$48.00


gp1.hyperspeed


16 GB


8


$0.16


$96.00

Their DreamObjects object storage prepaid plans also have attractive pricing:  200 GB for $4.50 monthly .  Usage based pricing plans are also available priced at 2.5¢/GB of storage plus 5¢/GB of download.

PrePaid DreamObject Plans


Storage Included Monthly Price Effective Price/GB


40 GB  $0.95  2.38¢


200 GB $4.50  2.25¢


1,024 GB  $19.95  1.95¢


2,048 GB  $34.95  1.71¢

press release: http://www.marketwired.com/press-release/dreamhost-announces-general-availability-of-dreamcompute-2112097.htm


documentation: https://help.dreamhost.com/hc/en-us/categories/202115418-DreamCompute-Cloud-Servers


site: https://www.dreamhost.com/cloud/computing/


TL;DR more downward pricing pressure in the ultracompetitive public cloud market.   It will be interesting to see who buys who and who is left standing 2-3 years from now...

 

[drmike]

But most mature at this point are QEMU and KVM. What's interesting about OpenStack is you aren't stuck with single hypervisor.  My understanding is that it supports a slew of them. But most mature at this point are QEMU and KVM.  Can one run different hypervisors on the same box or are guys busting that out to different boxes?


Less mature but supported include docker, LXC and even baremetal.


I take it Dreamhost and other majors are running KVM only?  


Price competitiveness here of $6/GB is doable all day long.  I wouldn't go doing it, but they want to try to grab marketshare.  Remember what we do when nothing else to compete on?  Go to price slashing.  Difference here is this is front side new product promo.  If they stick at $6/GB they are going to regret it and only reason to do such is price uptick causes sales not to happen.

 

[willie]

Meh this just sounds like a natural pricing evolution, those $10/GB plans have been around for years and memory has gotten a lot cheaper recently (desktop DDR4 dimms at $3/GB!).  The object storage costs .05/GB for outbound transfer!  Do they think they are Amazon or something?  OVH outbound bandwidth is .011/GB and Scaleway's is free (though not currently offered to new customers).  Similarly OVH's fancier "cloud" vps starts at $9/mo for 2GB and their 7gb/2 vcore dedicated resource vm is around $30/month.  And I thought the OVH VRack stuff now worked with its VPS's. IPv6 is still a work in progress though.


I don't think this cloud market is particularly price competitive since it's far more expensive than vpsboard-style hosting.  It's more about marketing and to some extent, walled garden services.

 

[fm7]

... hourly cloud competitors like DO, Vultr, GoDaddy, Atlantic.net, Cordero, and many others.

To be fair I think Quadranet (InfraCloud) deserves mention -- probably on the top of any list of inexpensive "true" cloud providers.

 

[DomainBop]

To be fair I think Quadranet (InfraCloud) deserves mention -- probably on the top of any list of inexpensive "true" cloud providers.

Quadranet's InfraCloud is just another OnApp install.  I'm not a fan of the proprietary OnApp (probably due to the many bugs I've encountered with providers using it over the years,).  I really don't see OnApp being that competitive with other cloud platforms going forward because it really hasn't made much of an inroad into the enterprise market, and the bulk of its customers are still small to medium sized hosting companies.  Its main attraction is it fits a market need by allowing smaller companies (like those using OnApp's SolusVM platform) who lack the resources  to develop something in house to quickly launch a cloud platform at minimal cost but that's about it.


Quadranet / AS8100 does however deserve a special mention for announcing a bogon that is full of spammers (http://bgp.he.net/net/45.127.206.0/23#_netinfo ) but that's another story

 The object storage costs .05/GB for outbound transfer!  Do they think they are Amazon or something?

The 0.05/GB isn't charged if you choose one of DreamCompute's prepaid monthly plans.

I thought the OVH VRack stuff now worked with its VPS's.

Private Network in the OVH Public Cloud OpenStack panel still says "soon"  It's expected to launch within the next few months.  It's on the roadmap for "Phase 3". 


Other Phase 3 planned new features: Monitoring / vRack private networks / Movement between datacentres / Floating IP addresses, Load Balancing as a Service and Firewall as a Service made in OpenStack / Complete infrastructure templates / Autoscaling / Automated backup / IPv6

What's interesting about OpenStack is you aren't stuck with single hypervisor.  My understanding is that it supports a slew of them. But most mature at this point are QEMU and KVM.

 http://docs.openstack.org/developer/nova/support-matrix.html 

 

[fm7]

Quadranet's InfraCloud is just another OnApp install.  I'm not a fan of the proprietary OnApp (probably due to the many bugs I've encountered with providers using it over the years,).

I don't trust OnApp but if "many bugs" is the criteria OVH and others beloved providers using OpenStack or CloudStack would be banned from the lists as well. Too bad the remaining VPS companies are Rus Foster's copycats, some posing as innovators.

 

[drmike]

Quadranet's InfraCloud is just another OnApp install.  I'm not a fan of the proprietary OnApp (probably due to the many bugs I've encountered with providers using it over the years,).  I really don't see OnApp being that competitive with other cloud platforms going forward because it really hasn't made much of an inroad into the enterprise market, and the bulk of its customers are still small to medium sized hosting companies.  Its main attraction is it fits a market need by allowing smaller companies (like those using OnApp's SolusVM platform) who lack the resources  to develop something in house to quickly launch a cloud platform at minimal cost but that's about it.

I've seen a lot of small shops who went loco and spent far too much on OnApp installations.  The stuff is NOT selling for them.  Most turn back and do what they can to recoup money, like participating in the CDN offering.  Not a cheap deployment for such an experiment.  Nothing for a big company or one spending VC money, but for those that actually work for their take, dropping I don't know?  $20k low side for unproven thing, meh.

http://docs.openstack.org/developer/nova/support-matrix.html 

Awesome reference!

Rus Foster's copycats

I cringe every time I see that blokes name.

Private Network in the OVH Public Cloud OpenStack panel still says "soon"

OVH just had other panel for customers do a 4 day outage.  They launch lab experiments then fold the stuff.  They keep expanding like a plague.  Lots of right gripes about support lacking.  Prior vanilla VPS sucked horribly.   I don't know why I have no confidence in OVH.  I view them as a sovereign fund sponsored trouble maker (which has to be proven yet).  I've spent $0 directly with them and plan on keeping it that way.   Cloud there is going to go into multiple distributed disk / SAN fails.  Already has I think.

 

[willie]

The 0.05/GB isn't charged if you choose one of DreamCompute's prepaid monthly plans.

Are you sure of that?  It wasn't explicit in the doc page whether there was a bw charge or not, only that the storage price was a bit lower.  How much transit is included with the monthly plan?


Regarding OVH, they had two VPS lines, some expensive-ish VMware ones that were supposed to be good and some cheap ones (OpenVZ slabbed in VMware) that sucked and everyone knew it.  Now they have replaced both lines (both now are KVM/Openstack) and both are supposed to be good, with the expensive ones having more resources and some HA features.  I had two OVH dedis in the past and one of the sucky cheap VPS.  The dedis were perfectly good and the cheap VPS was reasonably performant but kept getting rebooted.  I still might try one of the new cheap ones. 


As for the lab experiments getting cancelled, yes that's the point of a lab experiment--to see if the service is worth continuing to develop.  They're on a separate label now (runabove.com) and people who sign up for them know that they are enrolling in experiments.  That's a lot better than launching something as a production service and then pulling it later, as we see from some other places.

 

[fm7]

I cringe every time I see that blokes name.

I see the guy as the Father of VPS, pioneer of cheap VPS and fancy sites.

 I don't know why I have no confidence in OVH.

I have been a customer since 2011: I know exactly why I have no confidence in OVH


The fact is users may specify performance and reliability objectives in fine detail and any hosting company or data center conforming to the necessary characteristics set in those requirements and specifications should be fine to accomplish the project. In this context, I don't see OVH that bad.


Unfortunately, providers of low end boxes/VPS/VM attract mostly low end applications and/or clueless users and many times companies are blamed by what are in fact customer shortcomings: lack of project, loose requirements, poor specifications, unrealistic expectations, magical thinking. However more and more providers of unmanaged servers are luring non-techies with promises of greater performance and 1-click instant gratification but concealing the fact it is required a lot of knowledge to run a server connected to the Internet. Those providers can't blame their clueless customers saying they don't know jack about sysadmin, security, networks, and they deserve the bad-mouth because the prospect buys the service thinking it is the good old shared hosting plan, improved.

 

[DomainBop]

I don't trust OnApp but if "many bugs" is the criteria OVH and others beloved providers using OpenStack or CloudStack would be banned from the lists as well. 

My dedicated servers have better uptime than any of the "High Availability Cloud" providers I've tried which is to be expected because none of the platforms (with the exception of possibly VMware based clouds) are mature and so you're going to have bugs and growing pains which is bad news for anyone who believes the marketing hype about the reliability and safety of the cloud and then gets hit by one of those bugs.


Speaking of OnApp: yet more reports of disk corruption on storage arrays and and snapshots not working.  The provider in question offers a "1000% SLA" so I imagine they'll be issuing some credits...

1000% SLA


We deliver 100% availability of hardware and network or pay you 10x Pro-Rata Credits

 

[RosenHost]

I always had a sympathy for DreamHost. Looks like they are bringing some new competition

 

[drmike]

clueless users and many times companies are blamed by what are in fact customer shortcomings:

There is no doubt the cheapies attract more disproportionate and in public loud group of knowledge lacking folks than the major brands.  Lack of price barrier is part of it and communities that seem to trivialize the complexity by having other 'yuts allegedly doing the same make the daydream doable in those fertile and idealistic minds.  When it's time to get it done, hell even my ancient wrinkly self gets frustrated and expends too much energy and time often trying to self cobble even from docs.


The idea that many are shared hosting people lured by novelty of price and a new non supported app on shared, well I think that really has merit.  Definitely applies to a big enough subset not to just dismiss.


Now often there is this loco-motion of blame the company and blame the consumer in public spats.  I fall all over on sides there.  I just want the damn facts  Facts are rarer than knowledge on both sides in typical spat.

I see the guy as the Father of VPS, pioneer of cheap VPS and fancy sites.

He had more scams and fails than.... and shills....  I view him as 2011 and earlier king of shade and part of the move that gave virtualization products in cheap world an ugly black eye.


Same disregard and dropping of customers continued under his name all the way through UK2 merger/sale/whatever it truly is.  Hopefully he's gone and stays gone from hosting.

My dedicated servers have better uptime than any of the "High Availability Cloud" providers I've tried which is to be expected because none of the platforms (with the exception of possibly VMware based clouds) are mature and so you're going to have bugs and growing pains which is bad news for anyone who believes the marketing hype about the reliability and safety of the cloud and then gets hit by one of those bugs.

It's really hard to claim any durability and reliability until something has been deployed under your company for years. It also properly requires you have the actual know how, familiar with your deployment, perhaps created your own stack and can do the plumbing and daily operations.  Plus have an actual sysadmin involved, plus additional competencies.


Bare metal typically will outperform and have better uptime than cloud.  There are less moving piece, way less to go wrong.  No contention issues unless their upstream is undersubscribed on bandwidth.  Mind you, this statement banks on your the owner/renter being competent enough with a dedi.


There is a reason why cloud pisses me off and that's because straight up it means a whole lot more invest and commitment.  I entirely lose it when some pimple faced austistic 16 year old runs brand out saying cloud this and that.  Term abuse central right here:


https://www.chicagovps.net/services/cloud-vps


Cloud VPS



Show me what is cloud there... That's a drop-in replacement tranny outfit dress wearer disguising nothing more OpenVZ + Solus.  


ChicagoVPS's Cloud VPS Servers are the industry standard in reliable and cost effective OpenVZ based VPS servers.


^^^ --- see what I said...


What can I use a Cloud VPS for?
Host email, game servers, web servers, or anything you might want to with a Cloud VPS from ChicagoVPS.
Who knows, you could start the next big web unicorn startup and become a bazillionaire. Sky is the limit!


^^^ email and game servers now are cloud-ready just cause can run on OVZ commodity shitware stack with Solus?   What is this unicorn thing about?...


Sky is the limit?  No vapor is what this product is.  Manufactured marketing nonsense wrote by someone mentally retarded and intending on duping customers of hard earned money.  Everyone here knows you can't and shouldn't run an email server on CC's network.  It won't work, blacklisting yo'.

pay you 10x Pro-Rata Credits

I read SLA stuff all the time.  Good damn luck to customers getting such.  Most shops make it a PITA to get what you should get automatically.  The company knows it failed, they can easily deduce which customers were impacted and slide money into their account as non-refundable / service only credits.  No, you have to go jump through hoops, submit the stuff, get ignored, go to church and pray for it, then at random, maybe they'll get around it.  It's brutal how people market such and almost never live up to such.

 

[tonyg]

For anyone that has one of these instances running...how is the CPU performance?


Any overall benchmarks?

 

[tonyg]

I just did a bit of research and apparently the CPU is a 2.2GHz offering.


So CPU performance will be similar to Digital Ocean and Vultr, which in my book is uninspiring.


It's shame that most hosts are either moving away from 3GHz+ CPUs or just plain not offering them.

 

[drmike]

I just did a bit of research and apparently the CPU is a 2.2GHz offering.


So CPU performance will be similar to Digital Ocean and Vultr, which in my book is uninspiring.


It's shame that most hosts are either moving away from 3GHz+ CPUs or just plain not offering them.

I take it this clock matter is because they try it with old chips = electric burn.  They try it with desktop chips = cobbling misfit gear and might get crap for it from customers and maybe thermals are hell.  They try it with E5's but build costs are too high vs. income per node.  So they go the slower many more core approach. In theory 2.2Ghz clock and lots of old chips you could shimmy into mix and do so with gear laying around unused in many DCs...  pretty much have it gear if you want to buy the space plus a bit more for the boxes.


Vultr pumped 3GHz+ specs earlier and turned a 180 and hid their tail.  I see a lot of these slow core plays out there with single threaded performance that is meh.  Still lots depends on single thread to most people.

 

[willie]

The lower clocked chips have better performance per watt than the higher clocked ones, so multi-tenant hosts like them, though sometimes you really want the e3's single threaded speed. 


Is there a backstory with DreamCompute that didn't come up before?  I.e. the operator now seems to be someone with a history here?  I don't otherwise understand why DreamCompute is interesting.  With EC2, DO etc. you pay a lot more than with budget VPS, but you know you're dealing with a well capitalized company with lots of staffing and hardware, which matters to some people.  Can DreamCompute say the same?


Meanwhile, one of those cheap E5 servers was advertised for sale on LET earlier today--not in the cards for me right now, but quite affordable, and presumably more of them will show up.  It was cheap enough to be attractive even compared to Hetzner budget dedis (what I have right now).

 

[DomainBop]

It's shame that most hosts are either moving away from 3GHz+ CPUs or just plain not offering them.

I think you'll see more hosts taking the route that OVH did with its VPS Cloud and Public Cloud CPU instances and charging a premium price to users who need faster 3GHz+ CPU's (which at OVH is a 3.1 GHz E5-2687Wv3).  The slower 2.x GHz E5's (OVH iuses 2.4GHz  E5-2620v3) that most cloud providers are deploying now are probably more than enough for the needs of the average user.

Is there a backstory with DreamCompute that didn't come up before?  I.e. the operator now seems to be someone with a history here?  I don't otherwise understand why DreamCompute is interesting.  With EC2, DO etc. you pay a lot more than with budget VPS, but you know you're dealing with a well capitalized company with lots of staffing and hardware, which matters to some people.  Can DreamCompute say the same?

DreamCompute is interesting due to DreamHost's size and heavy involvement in OpenStack development.  They've been around 20 years and besides being one of the largest hosting companies (400,000+ customers) they've also contributed very heavily to OpenStack development over the years (Dreamhost's co-founder Sage Weil was the creator of the Ceph storage system and after DreamHost spun off its Inktank Ceph services unit a few years ago RedHat purchased Inktank  for $175 million in 2014.  The OpenStack Akanda networking system was also developed inhouse by Dreamhost and they later spun it off into a separate unit)


edit:

I see the guy as the Father of VPS, pioneer of cheap VPS and fancy sites.

I see RoseHosting as the father of VPS (or at least the first provider back in 2001).  Rus on the other hand can best be summed up by this quote from an old WHT thread "his password was "f0ster" ... lol"  To be fair, after the ThrustVPS/damnVPS/a2b2.com/cheapvps/gigevps/fsckvps bad old days he did seem to turn over a new leaf after joining VPS.net/UK2...it's just that many of us remember all of the pre-2011 crap
 

 

[drmike]

I see RoseHosting as the father of VPS (or at least the first provider back in 2001).  Rus on the other hand can best be summed up by this quote from an old WHT thread "his password was "f0ster" ... lol"  To be fair, after the ThrustVPS/damnVPS/a2b2.com/cheapvps/gigevps/fsckvps bad old days he did seem to turn over a new leaf after joining VPS.net/UK2...it's just that many of us remember all of the pre-2011 crap
 

My dislike for Foster comes around his finger wagging post-hack event blaming it on panel software.   That contributed to the author of the software taking his own life.

 

[fm7]

I see RoseHosting as the father of VPS (or at least the first provider back in 2001).

"I see the guy [Rus] as the Father of VPS, pioneer of cheap VPS and fancy sites" = irony

 

[fm7]

DreamCompute US-East 1 Cluster Service Disruption


Posted (October 11th, 2016 at 5:03 am PST) by Justin@DH
Our beta DreamCompute cluster named US-East 1 is currently having issues with its storage system. We have our entire team up and working on the issue, and hope to have service restored shortly. The storage cluster is vital to many functions of the cluster, and so is affecting networking, routers and instances. We apologize for the issues and are working hard to restore service.


Update October 11, 2016 6:30 AM: Work continues to restore service for networking and storage. We have all needed team members working on the issue and are making some progress, as some services are working normally for customers. We are working to restore service as soon as possible.


Update October 11, 2016 7:15 AM: More expertise has been pulled in to help with the situation. We are working on restoring the networking and Ceph cluster, and with it service for DreamCompute US-East 1 cluster.


Update October 11, 2016 11:45 AM: Work continues on the Ceph cluster for the DreamCompute US-East 1 cluster. We still have lots of expertise working on restoring service, although there are come issues still to be addressed. We hope to have more info soon and to be able to have more details on service restoration soon.


Update October 11, 2016 12:30 PM: The networking situation was improved, and the original cause of the system issues was caused from this, however the Ceph service using the network and hosting the storage subsystem is still having issues.


[FYI: Mar, 16 2016 -  Dreamhost replaces VMware SDN with open source for big savings]


Update October 11, 2016 7:30 PM: We apologize for the delay in getting services in this cluster restored. We are still working to safely re-establishing the storage backend for US-East 1 cluster and so there is likely still issues with access to instances. We don’t have an ETA on service restoration at this time but are understand customers needs and are working to make it as soon as possible.


Update October 12, 2016 6:00 AM: We apologize for the delay in getting services in this cluster restored. We still have our entire team on the case, working to restore service for the Ceph cluster that is the storage backend for the DreamCompute US-East 1 cluster. We are also working on some hardware upgrades (RAM) to speed the process along as well. There aren’t any more details at this time, but should we have more news we will update all customers.

Update October 12, 2016 9:00 AM: After some memory upgrades, the Ceph cluster is making very good progress. We have some cleanup to do for a few machines, and then we can look into getting OpenStack services working shortly after. We hope to have this issue resolved soon, and apologize for the delay and issues.


Update October 12, 2016 12:30 PM: The memory upgrades have helped, and we have nearly all the hardware for the Ceph cluster available and added in. Once that is complete, we can fit up any OpenStack services that are not working due to the Ceph cluster issues. Some instances that are not stuck in an error state may host data slowly in the meantime, and those in an error state will need additional cleanup. We hope to have service fully restored soon.


Update October 12, 2016 3:45 PM: The Ceph cluster is back up, and so are OpenStack services. The next thing to tackle are any routers that are broken from the situation. If your instances are not in an error state, they should be working unless there is a router issue. We will start cleaning up any broken instances now to get everyone fixed up as soon as possible.


Update October 12, 2016 7:45 PM: The Ceph cluster is stable and we are working on the remaining issue which are some routers that are not working properly. Most customers should have service restored, and the remaining should be working again soon as the routers are rebuilt by the management system. We apologize again for the delay, and hope to have all issues resolved shortly.


Update October 13, 2016 2:00 PM: We have repaired the network switch, the Ceph system, and we believe all routers that were having issues with the US-East 1 cluster. There are possibly some individual customer instances that are having issues and we will be reviewing them to fully clean up the cluster. If you notice something still not working, please contact support so we can get to your specific issue faster.


http://www.dreamhoststatus.com/2016/10/11/dreamcompute-us-east-1-cluster-service-disruption/