The pace of innovation in data center design has been rapidly accelerating over the last five years, driven by the mega-service operators. I believe we have seen more infrastructure innovation in the last five years than we did in the previous fifteen. Most very large service operators have teams of experts focused on server design, data center power distribution and redundancy, mechanical designs, real estate acquisition, and network hardware and protocols. At low scale, with only a data or center or two, it would be crazy to have all these full time engineers and specialist focused on infrastructural improvements and expansion. But, at high scale with tens of data centers, it would be crazy not to invest deeply in advancing the state of the art. Looking specifically at cloud services, the cost of the infrastructure is the difference between an unsuccessful cloud service and a profitable, self-sustaining business. With continued innovation driving down infrastructure costs, investment capital is available, services can be added and improved, and value can be passed on to customers through price reductions. Amazon Web Services, for example, has had eleven price reductions in four years. I don't recall that happening in my first twenty years working on enterprise software. It really is an exciting time in our industry. I started working on database systems twenty years ago during a period of incredibly rapid change. We improved DB2 performance measured using TPC-A by a factor of ten in a single release. The next release, we made a further four-fold improvement. It's rare to be able to improve a product by forty fold in three years but, admittedly, one of the secrets is to begin from a position where work is truly needed. Back then, the database industry was in its infancy. Customers loved the products and were using them heavily, but we were not anywhere close to delivering on the full promise of the technology. That's exactly where cloud computing is today--just where the database world was twenty years ago. Customers are getting great value from cloud computing but, at the same time, we have much more to do and many of the most interesting problems are yet to be solved. I could easily imagine tenfold improvement across several dimensions in over the next five years. What ties these two problems from different decades together is that some of the biggest problems in cloud computing are problems in persistent state management. What's different is that we now have to tackle these problems in a multi-tenant, high-scale, multi-datacenter environment. It's a new vista for database and storage problems. In this talk, we'll analyze an internet-scale data center looking at the cost of power distribution, servers, storage, networking, and cooling on the belief that understanding what drives cost helps us focus on the most valuable research directions. We'll look at some of the fundamental technology limits approached in cloud database and storage solutions on the belief that, at scale, these limits will constrain practical solutions. And we'll consider existing cloud services since they form the foundation on which future solutions might be built.

High-scale cloud services provide economies of scale of five to ten over small-scale deployments, and are becoming a large part of both enterprise information processing and consumer services. Even very large enterprise IT deployments have quite different cost drivers and optimizations points from internet-scale services. The former are people-dominated from a cost perspective whereas internet-scale service costs are driven by server hardware and infrastructure with people costs fading into the noise at less than 10%.

To be agile and cost effective, data centers should allow dynamic resource allocation across large server pools. In particular, the data center network should enable any server to be assigned to any service. To meet these goals, we present VL2, a practical network architecture that scales to support huge data centers with uniform high capacity between servers, performance isolation between services, and Ethernet layer-2 semantics. VL2 uses (1) flat addressing to allow service instances to be placed anywhere in the network, (2) Valiant Load Balancing to spread traffic uniformly across network paths, and (3) end-system based address resolution to scale to large server pools, without introducing complexity to the network control plane. VL2's design is driven by detailed measurements of traffic and fault data from a large operational cloud service provider. VL2's implementation leverages proven network technologies, already available at low cost in high-speed hardware implementations, to build a scalable and reliable network architecture. As a result, VL2 networks can be deployed today, and we have built a working prototype. We evaluate the merits of the VL2 design using measurement, analysis, and experiments. Our VL2 prototype shuffles 2.7 TB of data among 75 servers in 395 seconds - sustaining a rate that is 94% of the maximum possible.