How modern infrastructure becomes part of a data center business plan
A comprehensive collection of articles, videos and more, hand-picked by our editors
Remember the good old data center? It's on the way out thanks to the rise of converged systems. Gone are the twisted...
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
cable masses that connected many servers to one another and the SAN, while an overburdened cooling unit tried desperately to keep the temperature below 80 degrees Fahrenheit.
Converged infrastructure puts multiple servers, networking components -- and sometimes storage -- together in one box and includes a management platform to tie it all together. In these converged systems, hardware is managed as a service and delivered by an internal Platform as a Service (PaaS) cloud infrastructure. The goal here is to enable IT shops to be agile and rapidly deliver assets to the business's users.
A converged infrastructure system is one way to get a handle on hardware costs. Often, IT equipment consumes more than 70% of the IT budget. With a traditional approach, where IT is made available in silos, many resources are available and paid for, but sometimes aren't used at all -- the phenomenon of IT sprawl. Based on the current business demands on IT, this model isn't cost-effective anymore.
Customer demand for IT services in the office and on the road has increased dramatically. In those situations, it's not an option to order new hardware and wait days -- or weeks -- before it can be delivered. In a converged infrastructure system the hardware is already there, it just needs to be deployed or provisioned.
To some extent, a converged infrastructure system is not just a hardware platform, but an overall approach to IT. In such a system, resources like storage, servers, networking, power and cooling and security are all brought together in a pool, from which resources can be allocated when needed. An essential part of the solution is the management platform used to bring all these parts together and make them available on demand.
In a converged system, everything is virtualized. That means that server virtualization, storage virtualization, network virtualization and I/O virtualization are used together and made available from one integrated management interface to deliver resources on demand. This tight integration between virtualization and hardware makes it more difficult to see components as individual pieces instead of one integrated entity.
To a certain level, an IT shop can create a converged infrastructure system by standardizing and consolidating its technologies. What currently makes it difficult to rapidly deploy PaaS in a cloud environment is the lack of integration and standard components. Once blade servers replace rack servers, it is much easier to develop a management interface to allocate those servers on demand, together with the associated storage and networking. Currently, industry-wide standards are still lacking; that is where different vendors come in with varying technologies.
Current players in the converged infrastructure game
Right now, vendors like Hewlett Packard, IBM, Dell and Cisco are pulling their hardware together and offering it as a converged infrastructure system. At this early stage of development, most vendors are looking inward only; the focus is on integrating their own parts, not making them compatible with other vendor offerings. It's a new market and each vendor wants to be the market leader -- there's a lot of money to be earned.
As it stands, there are a few major players, but that number is increasing rapidly. Here's a short list of product descriptions for today's top offerings.
HP Converged Infrastructure
HP's Converged Infrastructure is not really a single product, but a system that brings different components together. It is comprised of four parts. The first is the HP Virtual Resource Pool, where server, storage, network and other key components are virtualized. Second, there is HP Data Center Smart Grid. This helps customers manage and optimize power in their data center by, for example, making it possible to switch off parts of the data center when not in use. Third is HP's Flex Fabric, which integrates the Ethernet and storage networks into one fabric for better management. Finally, HP has placed the HP Matrix Operating Environment on top of the hardware elements. This allows the allocation of resources from a single location.
Cisco's Unified Computing System (UCS) is as modular as the customer wants it to be and has several interchangeable options. The most important components in a Cisco UCS are the rack, fabric interconnects, proprietary switches and blade servers. All this hardware is managed from Cisco UCS Manager or UCS Express. UCS Manager is the main platform, tightly integrated with firmware on the different hardware components to make deployment of resources easy. UCS Express, on the other hand, is designed for remote management of IT infrastructure in branch offices; it integrates an original equipment manufacturer's version of VMware vSphere and connects to Cisco SRE 700 and 900 servers. These low-end servers have a maximum of 8 GB of RAM and two 500 GB hard drives, and are typically designed for use in branch offices.
Vblock is a system from VCE, a company that doesn't build its own hardware but focuses on integrating hardware built by others. In the Vblock model, the customer purchases routers and blades from Cisco that are connected to EMC Storage and joined with VMware Virtualization to create one integrated solution. VCE builds the solution at their premises and delivers it to the customer site, which means that one or more racks filled with hardware are placed in the customer data center. The advantage of this approach is the work can be finished at the customer site in a minimal amount of time, which means there is little interruption for the customer.
IBM is another interesting player in the converged systems game. IBM PureFlex offers systems that are relatively cheap compared with the competition, targeting both large enterprises and medium-sized companies. What is so interesting about the IBM offering is how similar PureFlex systems are to the tried-and-true mainframe. In other words, the company has started doing again what it was already doing decades ago but in different markets.
Smaller converged infrastructure vendors making a splash
There are many other vendors in the market of converged infrastructure systems such as Hitachi, Dell, Oracle and NetApp. Their products align with the basics outlined in the systems above, where integrated hardware is sold with management interface software that makes it possible to manage all parts of the system.
The skeptic's view of converged infrastructure
Converged infrastructure systems have drawn the attention of some companies, but they aren't cheap. Some of the least costly systems start around $100,000, with prices easily rising to more than $1,000,000 per rack. The main argument vendors are using to convince their customers is converged infrastructure systems can cut down costs of operating the IT environment. But how useful is it to save a million dollars on people if you have to spend the same amount of money on hardware every refresh cycle?
Apart from price, on the current market there seems to be a concern that old technologies are just rebranded to be sold in a box as converged infrastructure systems -- routers, fabrics, switches, blades and management software have all been around for a long time. Maybe a converged infrastructure system is just a rack filled with hardware that is organized in a better way and comes with a better management interface. And if that is the case, the customer might have other choices than to buy it all from one vendor.
Converged infrastructure systems are an important step forward in the configuration of the data center, especially since they allow the data center to provide PaaS services. Currently, we're just at the beginning of this game -- characterized by huge investments by the different vendors to develop a platform that works. As a result, interconnectivity is not on the list of developers' concerns, and as the systems are sold as an entity, it is difficult to integrate parts of other vendors. In the current market, there's a huge risk of vendor lock-in at high prices.
Developing converged systems for the future
The most important development that needs to happen for converged infrastructure systems is making them open. It is only a matter of time before customers start asking how to integrate products from other vendors, and someone comes along to make interoperability a unique selling point.
VCE's Vblock already makes small moves in the interoperability realm, but it is still tied down to a few specific vendors. Once the first vendor has started to really open its product, other vendors will follow and cheaper parts will enter the market -- with prices dipping to more reasonable levels as a result.
About the author:
Sander van Vugt is an independent trainer and consultant based in the Netherlands. He is an expert in Linux high availability, virtualization and performance. He has authored many books on Linux topics, including Beginning the Linux Command Line, Beginning Ubuntu LTS Server Administration and Pro Ubuntu Server Administration.