KML_FLASHEMBED_PROCESS_SCRIPT_CALLS
 

Analysis of Recent Cloud Announcements (Rackspace & Amazon)

October 23rd, 2008 by - 13,250 views

There have been a flurry of announcements in the Cloud Computing space in the past two days, most notably coming from Rackspace and Amazon. I have been trying to digest these quickly and wanted to post a recap and my analysis of this news. It’s pretty obvious to those of us within the Cloud Computing space that this is not a trend, but actually a logic progression of technology and services. The benefits of the Cloud are clear: pay for what you use, use only what you need, internet infrastructure provisioned using a web browser or API. But once again as the space becomes cluttered with new providers or features, the confusion starts bubbling up.

Rackspace Announcements

mosso_rackspace_logos So let’s take a look at what Rackspace announced on Oct 22nd. Billed as expansion of their “Cloud Hosting Portfolio”, Rackspace’s Cloud announcement provides some insight into their vision and roadmap. I listened to their “Cloud Event” which seemed to get a lot of hype, however, nothing truly jumped out at me as being earth shattering. A phrase comes to mind after all of the dust settled from the event “innovation through acquisition.” Don’t get me wrong, Rackspace’s achievements within the hosting business are definitely impressive and the company did manage to pull off one of the only IPO’s of 2008 within the Technology Sector. But this event seemed to be more of a requirement stemming from stockholders and their Board to “show something noteworthy.” They did do a great job at getting everyone’s attention though.

From the Cloud Event, I wrote down some points that they mentioned:

  • They broke down the Cloud into only 2 segments: Cloud Hosting & Cloud Applications
  • Their Cloud Hosting Division now consists of 3 products: Cloud Sites, Cloud Files & Cloud Servers
    • Cloud Sites – this is the current Mosso offering, rebranded. I view this to fall under the Cloud Platform part of my Cloud Pyramid. You are free to do what you want within it, but with some limitations (predefined application frameworks, no SSH or RDC access currently, no API access, month-to-month billing). This is a good option for people who want slightly more than what Google App Engine offers Python users for free.
    • Cloud Files – one can view this as a CDN meets an online storage solution (e.g., an Enterprise DropBox). For this solution, Rackspace acquired JungleDisk (which interestingly uses Amazon’s S3 service for their cloud storage solution – this is expected to change to CloudFS, Rackspace’s own product, at a later date). I think of this offering as falling within the Cloud Extender’s portion of the Cloud Pyramid, detailed within this presentation.
    • Cloud Servers – with the acquisition of Slicehost, an innovator within the Xen virtualization, low-cost VPS hosting arena, Rackspace adds to their product line “cloud servers.” Similar features here compared to EC2 and GoGrid with some feature omissions and I view this to be Rackspace’s Cloud Infrastructure entrance.

My Competitive Analysis of Rackspace’s Announcements

While many within and outside the industry may view Rackspace’s recent announcements as a “unified cloud strategy,” my, albeit biased, analysis of the news is a bit different, not as unified and lacking in some areas. Some points to ponder:

  • Why does Rackspace view the Cloud as only having two segments (Hosting and Applications) as stated by their CEO, Lanham Napier, in his keynote? It is much broader and richer than that, with Cloud Applications, Cloud Platforms, Cloud Infrastructure, Cloud Extenders and Cloud Aggregators. Perhaps Napier’s statements were specific to their strategy, and that would explain how Infrastructure and Platforms seem to fall under the umbrella of Cloud Hosting. Over-simplifying does make sense if Rackspace is trying to go mainstream with their announcements. However, if you try to be everything within the space, you could be spread a bit thin. This is one of the reasons why GoGrid chose to be at the most fundamental layer of the Cloud and provide Infrastructure as a foundation to build other Clouds or Services on top of.
  • Do their new services seem to be a cobbling together of technologies? Will they have integration problems as they move toward a more cohesive offering? This isn’t necessarily an issue as they are trying to accomplish something on many parts of the Cloud Spectrum. But this could potentially present logistical and business issues as they attempt to wrap the offerings together.
  • Cloud Storage as as stand-alone solution is good. It will be interesting to see if and how their Cloud Files offering can integrate with Mosso and Slicehost.
  • Billing madness on the horizon? Will users of many of their products get unique bills for each one? How will monthly pricing (SliceHost and Mosso) work with Cloud Files (usage based)?
  • What about Windows? Slicehost offers no Windows Cloud Servers. Mosso offers limited platform support for Windows environments. Remember, GoGrid is the largest Windows 2008/2003 Cloud Infrastructure provider on the planet!
  • Where is the rest of the infrastructure? If Rackspace is going to bill this as Cloud Hosting, what about hardware-based load balancing?
  • SLA’s – At the event, Rackspace claimed to have a 99.99% uptime guarantee, but I believe this was only related to their Cloud Files offering. I’m actually a bit confused by Mosso’s. It states that for each 60 minutes of unscheduled downtime, they will credit you 1 day of hosting. Does that mean you have to have an outage lasting more than 60 minutes for the SLA to kick in? Can six 10 minute outages be lumped together for that 1 hour? Does it span months? There is no mention of an uptime guarantee. Similarly (and coincidentally), Slicehost thinks that “SLA agreements are just plain silly”. Try telling that to an Enterprise user or corporation who has been hit with an outage. No guarantees there. I expect both of these to change soon. Note, if a GoGrid user is down for 7 hours, they get a month for free.
  • Play Cloud Computing Product Announcement Bingo for yourself to see other areas where GoGrid stacks up better.

Amazon’s Announcements

AWS_logo Today, Amazon rattled its saber with some Cloud announcements of their own. Amazon has done a tremendous job in the Cloud Computing space and analysts and the press consider GoGrid to be Amazon’s #1 competitor. Two of their three announcements today seem to be more of a “catch-up” play than anything else. More about that in the analysis section. Today marked three important items, specifically:

  • The removal of the Beta tag from their EC2 product
  • The availability of Windows Servers in EC2
  • The availability of a Service Level Agreement (SLA) for EC2

Amazon had announced that they would be offering Windows several weeks ago and most likely this “availability” announcement was brought on by the fact that the Microsoft Professional Developers Conference (PDC) is starting next week. Note: GoGrid is an exhibitor at the PDC and will have some exciting Windows Cloud Server announcements at the event.

My Competitive Analysis of Amazon’s Announcements

It’s nice to have some competition in the Cloud. It drives innovations, prices and surrounding services to the benefit of the end-user. So let’s take a quick look at what these three items mean (and once again, I put on my GoGrid-green colored glasses for this):

  • Removal of Beta – the beta tag is over-used and has lost a lot of its value. However, it does provide a shelter for some companies and forces other companies to live to higher standards. EC2 has been out in the market for a long time so this does make sense. But if you couple this with the details contained with Amazon’s new SLA, you can see how the removal of the tag may have forced their uptime guarantee to be lower (99.95%). Note: GoGrid has a beta tag, however we treat all aspects of our product to be that of a production environment. “Beta’s” are not technically covered under our SLA credits, yet we have been honoring SLA Credits fully since GoGrid’s launch in March.
  • Windows availability – Windows in the cloud is huge. Over 2/3′s of GoGrid’s cloud servers deployed are Windows. It is only logical that Amazon release Windows then. The important thing here to note though, is that they are ONLY launching Windows Server 2003 and not Windows Server 2008 (already available on GoGrid). Amazon also offers only Microsoft SQL Server Express and Standard Editions; GoGrid offers Express, Workgroup and Standard editions. While EC2 now does offer higher-end Windows Server 2003 with larger RAM allocations, GoGrid already has plans for larger RAM allocations and other methods for utilizing higher-end SQL Server environments. Also, just figuring out your Administrator password for new Windows instances on EC2 seems a bit more complex then necessary (GoGrid does this in an easy way, you simply look it up within the GoGrid portal…no command line interface required – you could, if you want, get your passwords via the GoGrid API if you like command line interfaces). Since we are on the topic of command lines vs. GUIs, our Windows in the Cloud experience indicates that many Windows users are more comfortable using a point-and-click environment (GoGrid’s Web Portal) vs. command line. Users trying Windows on EC2 may be greeted with an unwelcome surprise at the level of “code” they will need to employ with Windows on EC2.
  • SLA details – SLA’s are good and there for the protection of the end-user (Mosso and Slicehost should listen and follow our and Amazon’s lead). It makes business sense to me that as EC2 removes the Beta tag, they also add an SLA. But let’s read a bit between the lines here:
    • 99.95% availability during the service year – that’s about 5 hours of downtime a year (not bad)
    • 10% SLA – for every $1.00 you spend during an outage, you will only get $0.10 back; GoGrid takes a different approach by offering 100 times the fees back, so for every $1.00 you spend during an outage, you will get $100 back in Service Credits. To put it differently, if you are down on EC2 for a month, you only get credit for 3 days.
    • No mention of data loss or coverage/recovery; at least with GoGrid, your storage is persistent we will make all reasonable efforts to recover your data.

While I realize that this is not an exhaustive analysis, it did tire me attempting to read between the lines. And do understand, this post is not intended to criticize any of the companies mentioned, it is more designed to provide another Cloud vendor’s perspective on the announcements and hopefully open a critical eye to these announcements. I always recommend “shopping around” as you look for a Cloud Solution as some vendors will have offerings that better meet your business or technical needs. Don’t get caught up in the hype of it all. It’s good to see the space starting to mature…I think I’m getting a few more grey hairs in the process.

The following two tabs change content below.

Michael Sheehan

Michael Sheehan, formerly the Technology Evangelist for GoGrid, is a recognized technology, social media, and cloud computing pundit and blogger who writes regularly about technology news and trends.

8 Responses to “Analysis of Recent Cloud Announcements (Rackspace & Amazon)”

  1. Paul says:

    Michael -

    No doubt that GoGrid takes the prize in terms of usability (you have a great interface). But, Amazon’s model of “machine images” is most compelling for our uses. What are GoGrid’s plans in this regard?

    Thanks,
    -Paul

  2. Michael Sheehan says:

    @Paul,

    Thanks for the hat tip for usability. If you are talking about the ability to save AMI’s (e.g., server cloning), that is something that is under development. Instrumental to this cloning feature is first to have Cloud Storage (which is coming soon). I’m sure you will find many similarities to the AWS offerings within the near future.
    Thanks,
    Michael

  3. Martijn says:

    @Paul: nice to see that you are proud of your product but GoGrid has some things making it a lesser offer in my opinion.

    - No possibility to store your own images.
    - The need to pay for a machine even when you do not use it.
    - No shared storage
    - No ability to attach a storage

    just to name a few.

    But you sure have the best UI. Compliments for that. I hope you quickly match the above options in you UI.

  4. Michael Sheehan says:

    @ Martijn,

    You probably meant your post to be directed to me (grin). Thanks for your list of requested features. These are all on our feature road map. Cloud Storage will be the most likely feature that you will see first and it will address points 3 and 4 above. Also, do remember that all GoGrid Cloud Servers do have persistent storage built in to them.

    Thanks,
    Michael

  5. Steve says:

    The thing that kinda gets me about the Amazon SLA is how they require you to lay an SLA outage. You’d need to produce the LOG files from the dead server showing the failure time as proof. What if the server node is lost… I heard that was an issue. Where are the log files?

  6. Mark says:

    @Michael,
    thanks for the analysis/overview. I was intrigued by the SLA comparison, then disappointed upon digging a little.

    If one wants to make a spend based comparison of SLA’s, which I’d like to do (rather than just time based), then I really should compare using a common basis/unit of measurement.

    Specifically If I spend $1 with AWS does that really translate to the same resources I would get by spending $1 with GoGrid?

    - I did just look into this briefly, but the following was my impression….

    It’d be a brave soul who tried to map the AWS to GoGrid. AWS breaks their machine specs down much more clearly. I have a real problem progressing my analysis using your web site’s data. Seriously, what does buying one ram hour buy me in terms of cpu cycles and HDD space?
    I think I’m correct in complaining here since AWS give more data.

    This has made me suspect that the AWS-GoGrid target markets are so different in their use-cases requirements that perhaps such comparisons are not of interest to their two target markets?

    I accept that different use cases might make this type of comparison irrelevant, but really if you’re going to suggest a SLA is 100 times better, it is less appealing if I have to use 200 times the resources/time to ‘recover’ from the interruption. think here of one MPI 1.x node dying during a task.

    Otherwise thanks for the interesting blog – it’s still in my reader :)

    Cheers

  7. Brian says:

    Michael….

    Is the feature roadmap published anywhere for us to see ??

    These have been mentioned before but it would be great to have one spot to be able to review what’s upcoming & when for GoGrid:

    - Custom Server Images & Snapshots

    - Larger RAM/CPU options

    - Ability to upsize an existing server (increase RAM/CPU)

    - Roll out of the RightScale MySQL support functions

    - Tweak the Dash
    > key server stats and config details via click
    > internal IP assignment via the Dash (like what is available now for external IP)
    > ability to RD or SSH directly from Dash to server by clicking on server icon
    > in IP tree show which are assigned to what server
    > Remember Me login option – so username/pass doesn’t have to be typed each time
    > Fix forum login so it happens automatically

    - Ability to “turn off” a server (and not get billed) without deleting it.

    - An edge firewall controlled in the interface for basic port management of all your external IPs rather than having to setup individual software firewalls on each server.

    - Increased security of the grid management dashboard….for deleting/changing existing servers. Maybe require re-entering of password.

  8. [...] Interesting fact: GoGrid accounced that over 2/3’s of GoGrid’s cloud servers deployed are running Windows. [...]

Leave a reply