Embedded Network Applications: Friend or Foe?

M. Spanbauer

Summary Bullets:

• Many vendors offer embedded application platforms within either WAN or LAN equipment (or both), touting performance benefits.
• Customer adoption remains tepid, however, and many often opt for appliances or servers/virtual machines due to convenience or familiarity.

Nearly every major networking vendor provides an application platform with which either their partners or customers themselves may embed applications.  These platforms can come in several forms, such as HP’s ONE module, which resides in a switch; Cisco’s UCS Express, a router/switch application services device; or Arista’s new 7124SX switch, to name just a few.  Potential benefits include, for example, improved packet processing performance, faster application response times, and deployment simplicity.  Whether it is a lightweight application such as a DNS or DHCP server, or something more robust such as Exchange or a call management suite, these emerging application platforms appear to be gaining steam in the market.  Vendors say their customers find ease of use, tight integration, and performance/responsiveness top the list of benefits, though operational simplicity and (perhaps more important) network team control help.  This last element is one of the most notable, as it demonstrates the divide that remains and inhibits enterprise growth into a more aggressive cloud adoption curve.  The storage, server, application, and network teams often remain separate functions; therefore, appropriation of resources to their peer groups can oftentimes be slow.  However, these quasi ‘network appliances’ give the network team back the keys to a server resource, yet administration and control remain within their domain.

With Network Dependence Critical, Is Downtime Acceptable?

M. Spanbauer

Summary Bullets:

• For years, enterprises invested in ‘good enough’ networks merely to make sure the plumbing connected everything together functionally.
• With cloud adoption rapidly increasing, fewer applications residing on-premises, and business continuity depending on 24×7 network access, enterprises need to re-think the network design and approach.

Enterprise networks were designed for years (and, to a large degree, still are) for three application areas: campus (or access), core, and data center.  With cloud and ‘anywhere’ access of mission-critical applications, users must have quality access to resources no matter the connection point.  Whether wired Ethernet, WLAN, wireless 3G/4G, or other means, downtime is unacceptable.  Yet, as RFQs go out, access resilience is missing or getting surprisingly low priority.  I contend that enterprises must raise the stakes and invest in redundant power, resilient management (whether in-box or in-stack), resilient protocols, and ultimately solid management interfaces (assurance, monitoring, orchestration, etc.).  Now, it is true that redundant links have become more prominent with the availability of commercial cable and DSL at aggressive prices (relative to fractional T and frame a decade ago), yet within the campus, surprisingly few switches or WLAN have RPS or resilient, distributed uplinks.