I was first briefed on the Savvis Symphony VPDC (virtual private data center) back at Cloud Expo NYC in April of this year and had intended to post about it back then, or at least when they went live in July… so much for good intentions… They are starting to market this more heavily now, so perhaps it’s not a bad time to get this done because VPDC has a few innovations that are worth noting.
Perhaps the most interesting aspect of VPDC is their tiered QoS model. Think about it. Today clouds come in a one-sized-fits-all model. You either have the open Amazon model with minimal SLAs and fairly opaque underlying infrastructure based on commodity gear, or you get the “enterprise cloud” model with higher SLAs (and costs) based on enterprise-grade gear. Or something in the middle. But you can’t typically get two or three SLA/QoS configurations, with commensurate pricing, from the same cloud provider.
With VPDC, that’s exactly what you get.
VPDC never goes to the Amazon level – with very low cost instances and no SLAs – but they do offer two QoS tiers today, with a third tier planned.
VPDC Essential is their starter level, with 99.9% SLA, best-effort QoS and inexpensive SATA-based storage. This is targeted at the dev/test use case.
VPDC Balanced is the mid-tier offering, with 99.99% SLA, VLANs, enterprise QoS on a 100 Mbps network, and 2-tier ILM storage. They are targeting Balanced at the Web application use case.
VPDC Premier (planned) will have 99.995% SLAs, more VLAN provisioning, 1 Gbps network, and 3-tier storage for more “mission-critical” workloads.
As you move up, you get more prioritization of bandwidth, less storage contention, fewer VMDKs per LUN, faster drives, etc.
Savvis would not give me any pricing information, but clearly you will pay more for Premier and likely even the Essentials pricing will be significantly more than Amazon or Rackspace. Lack of pricing transparency puts them a bit at odds with AT&T (Synaptic pricing here) and Terremark (vCloud pricing here). The only information I have I that pricing is hourly based on CPU, RAM and which operating system you are using (Microsoft’s SPLA fees presumably causing the difference). Interestingly they are disclosing bandwidth fees and are charging for bandwidth like a hosting provider ($50/Mbps 95th percentile model) vs. the more typical straight per GB in/out metered model.
Savvis has no current intention of allowing credit card self-signup models for new users, even with their Essentials package. This could be a mistake as so many projects start off with a very small buy and the Amex charge is easy to expense. AT&T and Terremark might get those customers that don’t want to start with a sales rep, though the buyer seriousness is certainly better formed if they are willing to go through that pain. By and large, making it easier to sign up could be in Savvis’ best interests.
What’s VPDC Made Of?
VPDC is an enterprise cloud based on VMware virtualization, Cisco UCS blades, Cisco Nexus switching, HP Opsware provisioning automation, Compellent SAN storage, and other technologies – okay, good enterprise-grade stuff. Savvis relies heavily on deep integration with UCS and Nexus to get the QoS tiering to work with VMware. They also rely heavily on the flexibility of Compellent’s “Fluid Data Storage” virtualized storage software. All images are monitored using TripWire and connectivity can include MPLS and VPNs.
What’s VPDC Mean for the Cloud Market?
Amazon, Rackspace and others have grown largely on the backs of Web developers, SMBs, and enterprise usage outside the control of corporate IT. This is but a fraction of the potential future market as the enterprise moves more and more to the cloud. Enterprise IT buyers are much more precise and demanding when it comes to infrastructure than most Web and game developers. Having tiered SLA/QoS levels, with pricing to match, might become an important consideration for even the mass-market cloud providers if they want to win in the enterprise. The market is just too big to ignore.
Alternatively, you could see the mass-market guys go the opposite route – adding the “five nines” SLAs and high QoS capability to their core commodity-priced offerings. This is a technology and scale issue that Amazon and Google are probably in a good position to leverage. After all, if you can get VPDC Balanced for the price of EC2 reserved instances, it will be pretty hard for Savvis to compete. But that’s a big if.
In any case, Savvis has done a nice job leveraging the technology now available to create a differentiated offering based on QoS and SLAs. VPDC is available through data centers in the U.S. and U.K.
VMware, Compellant and Tripwire = a Cloud? VPDC works more like a virtualized datacenter. Business model notwithstanding it seems like the focus is on SLA not dynamic scalability, which is kind of what Cloud is all about. The demos I’ve seen suggest they don’t expect to do much in the way of scaling and even their marketing ignores that feature in their product.
Admittedly they have a lot of gaps to support the type of scalable workloads that you find on Amazon and others. For the enterprise, the SLA focus is indeed important. Being able to have autoscaling to 100’s of instances is not how enterprises are thinking about cloud computing – at least not in IT. EC2 didn’t have autoscaling until May 2009, but we still called it a cloud. VPDC is more like Amazon VPC than straight EC2, which is a point I should have made above. However, they do offer deeper SLAs, auditability and the tiered service approach that seemed to resonate with a lot of the attendees to a Savvis seminar I attended in Boston a couple weeks ago.
John, Agree with your observations on the enterprise mindset. However, I also have to point out companies like Google (I am an alumni) and Salesforce.com are changing the buying perception. They are putting more emphasis on self-service, ease of use and configurability. Those features are becoming equally critical. Disclosure: I work for Skytap.com
Sundar – agreed. In the long run the volume players will have the enterprise capabilities while many of the enterprise players won’t have the usability or low friction of the volume guys. Today, however, mature process-oriented IT shops will find the Savvis-style offerings more comfortable.
John, if Enterprise It is NOT thinking about auto-scaling 100s of instances what are they thinking about? At the end of the day IT manages thousands of physical servers and 10s of thousands of VM. They can’t be doing that on a one off basis can they?