New data center appliance houses network, server, and storage resources as virtual services.
Cisco Systems last week introduced several products under the name Data Center 3.0. Though all designed for the data center, they have little else in common, but that doesn't mean that Cisco lacks an overall plan. For the first time, the networking vendor has spelled out the endgame of its virtualization strategy, which looks increasingly like the end of physical servers as we know them.
There are three main new products. The most ambitious, VFrame Data Center, is an appliance based on VFrame, the virtualization management software Cisco acquired with Topspin in 2005. Whereas Topspin was focused on InfiniBand and storage, the new VFrame DC appliance is IP-based and can orchestrate other resources besides storage. Cisco describes it as a platform for directing network, server, and storage resources into new virtual services, offering an API that developers can build on. Cisco says the API is an interface to program the network itself, not just the VFrame DC.
John Chambers' new reality: virtual services
In practical terms, this means Cisco is competing with many of its software and server partners, though Cisco denies that. It claims that a network device is far more powerful than mere software, offering deeper visibility into the application stack. Whereas management software could only see the virtual resources, a network appliance can look inside packets, understanding what hardware a virtual service depends on.
There may be some truth to this. Cisco is certainly right about the problem of mapping the virtual to the physical, but there's no reason that the same can't be done in software. VFrame DC is essentially middleware in a box, abstracting multiple resources into a single API, so its success will depend on the API's adoption by developers. Cisco says it's working with developer partners, and the API is open to customers and ISVs.
A second new product, ACE-XML, shares a brand name with Cisco's Application Control Engine, a giant load balancer blade, but little else. It's essentially the XML firewall that Cisco acquired with Reactivity earlier this year, repackaged as a blade for the Catalyst 6500 data center switch. Naming the Reactivity box after ACE hints at closer integration to come, though for now it just means that they're aimed at the same kind of customers and share the 6500 chassis. Both represent a hardware option for many service-oriented architecture functions, competing with Citrix Systems and F5 Networks.
Cisco also introduced Trusted WAN optimization, which encrypts the enormous caches of data on Wide Area Application Services boxes.
Cisco plans to release more data center products over the next 24 months. Last week, it also agreed to invest $150 million in VMware, the dominant server virtualization vendor that's a unit of EMC. In Cisco's view, the long-term trend in server hardware is replacement of local devices by shared network resources. It began with peripherals such as printers and has moved to internal hardware, including disk drives. The next step is networked CPUs and networked memory. Cisco admits this is unlikely to happen within 24 months but believes it will occur within the life cycle of Cisco products, around 10 years.
If memory and CPUs move out into the network, what will be left for servers to do? Cisco doesn't really have an answer but claims it isn't trying to compete with servers. Server vendors would have to adapt to the new network-centric architecture, just as they are embracing virtualization.
Cisco's motivation is clear: It wants the network to be the most important thing in the data center. To fulfill its vision, it must persuade customers that there's a compelling reason for the network to be more than just a pipe, and that is less clear.
2014 Next-Gen WAN SurveyWhile 68% say demand for WAN bandwidth will increase, just 15% are in the process of bringing new services or more capacity online now. For 26%, cost is the problem. Enter vendors from Aryaka to Cisco to Pertino, all looking to use cloud to transform how IT delivers wide-area connectivity.
The UC Infrastructure TrapWorries about subpar networks tanking unified communications programs could be valid: Thirty-one percent of respondents have rolled capabilities out to less than 10% of users vs. 21% delivering UC to 76% or more. Is low uptake a result of strained infrastructures delivering poor performance?