Another major cloud service is angling today to join the pantheon of public cloud providers such as Amazon Web Services, Google Cloud, Rackspace and Windows Azure. Verizon Terremark, a unit formed by Verizon’s $1.4 billion purchase of the Miami-based cloud vendor Terremark in 2011, is launching two new public infrastructure services today: Verizon Cloud Compute and Verizon Cloud Storage.
But what will this new public cloud offering give the world that it already doesn’t have?
Verizon Terremark’s infrastructure-as-a-service Cloud Compute and object storage Cloud Storage offerings are most analogous to Amazon Web Service’s EC2 and S3 services, respectively. But VT claims it offers some key technological and business advantages. Let’s have a look.
The Technology Argument
For starters, VT claims its networking architecture will allow for multi-tier networking, with multiple network interfaces available to any virtual machine. Applications running on the VT cloud service can have any number of tiers available, from the client-server two-tier architecture, to the more stable three-tier presentation-business-data architecture that is common for many web apps and beyond.
Using multi-tier networking can introduce a lot of stability and fault-tolerance into an application; the more parts of an application’s process that are handled by different machines on the network, the less chance there is for the application to fail. And, if you want to swap out a software component of the application, it can be done without rewriting the whole app.
But, there’s a catch. Any time an application process or data has to jump from one machine to another, it slows the process down. The more tiers introduced, the slower the process gets, as information has to potentially hop multiple times from one server to the next.
To adjust for this problematic latency, Verizon Terremark CTO John Considine told me that the topography of the network will also be laid out as an “ultra-flat network,” which means that inside each of the new public cloud’s seven regions, virtual-machine-to-virtual-machine communication will be done in one network hop, even if there are hundreds of thousands of virtual machines within a region’s virtual network.
This is a sharp departure from what Considine called the typical leaf-and-spine topography seen in many places today. That topography, he claims, leaves many network links oversubscribed and can introduce latency into the system—especially with multi-tier network architecture. “We’ve removed the restrictions on where in a region we put a customer’s servers,” he said. “We can do this by rewriting network packets on the network layer as the packets move through the network.”
Besides the improvement in multi-tier processes, Considine argues, Verizon Cloud customers shouldn’t have to suffer the so-called “bad neighbor” effect that can adversely affect one or more virtual machines if a nearby virtual machine is hogging too many network resources. Considine says that should help increase reliability of the VMs on the network.
This kind of network flexibility is pretty much what you would expect a network provider would feature: after managing a network of its own for so long, one would expect VT to have some decent tricks up its sleeve. Such network flexibility should translate into faster multi-tier applications, if developers want to build them.
The Business Argument
From a business perspective, Verizon claims its Cloud Compute customers will have far more control over the resources they use on the cloud.
For example, in AWS’s EC2 service, the m1.small server instance offers 1.7GB of RAM. So, Considine said, a customer with a 2GB app can either try to squeeze the app into the smaller 1.7GB space, or upgrade to Amazon’s next instance, the 3.75GB m1.medium, and pay more.
By contrast, Verizon prices its services are priced by the gigabyte and the gigahertz. Servers, then, can be better sized to a customer’s actual need, Considine said.
Verizon has also simplified pricing by eliminating, for instance, charges per input/output of memory transferred.
Simplified pricing would be a cool part of this new cloud service; given that some customers actually attend seminars to learn just how to read a bill from AWS, simplicity would be a welcome breath of fresh air.
The Bottom Line
U.S. customers will get the most access right now, with four available cloud regions. The EMEA will have two available cloud regions, and another cloud region in Brazil. The APEC region will see availability of Verizon Terremark cloud services in 2014, Considine said.
If these cloud services deliver what they promise, Verizon Terremark could shake up the public cloud market. The company is targeting enterprise customers (naturally), but it is also driving towards medium-sized and government clients as well—the latter being a direct square-off with AWS’s recent claim staking of the government sector this summer.
Lots of companies have tried to knock AWS off the top of the public-cloud heap, but Verizon Terremark might have a better chance than most. Unlike AWS and all of the other cloud competitors, VT controls a network, which may give it an advantage until the other services can catch up.
Image courtesy of Wikimapia