So, now that we’ve discussed who the major players are, let’s take a closer look at why it’s really such a big deal that we can now deploy our Unified Communications products as part of our virtual infrastructure, using the Cisco UCS.
Virtualization Simplifies the Way We Do Business
Once again, datacenters can be very interesting and challenging environments. Due to the drastic decrease in server prices over the years, we have gone from large monolithic mainframe servers that process many applications to one that utilizes much cheaper and smaller servers used for individual applications.
This change in datacenter design means that we have drastically increased the number of physical servers from a few large ones to hundreds and even thousands of very small ones. This created a problem known as “server sprawl.” The result of this was many, many servers added to our datacenter that each ran a single application. These servers were extremely underutilized and collectively created excessive amounts of heat. Since this heat had to be dissipated with appropriate cooling measures, this resulted in equally excessive power and cooling costs.
In modern datacenters, we are now able to virtualize these many servers onto a relatively few physical servers. Although it depends on the type and utilization of a particular application, a normal rule of thumb has shown that we can effectively virtualize 10-15 servers per physical host. This represents a significant reduction in a datacenter’s footprint and heat generation. For our Unified Communications servers, we are currently able to virtualize up to four servers per physical host. However, this capability will certainly increase with each new version.
Lower CAPEX and OPEX
Given that datacenters are incredibly expensive to run (they require expensive real estate, use incredibly large amounts of electrical power, and have high operational costs), virtualization of our datacenters represent a reduction in a company’s Capital Expenditures (CAPEX), but more significantly, their Operational Expenditures (OPEX).
One of the highest operational costs within a datacenter is the servers themselves. Modern single-purpose servers have been shown to only have 5-10% utilization, on average. Conversely, if your servers are using a large percentage of your datacenter’s power, like 75%, then we have a very inefficient scenario. By virtualizing these servers, we can now combine many servers (10-15) onto a single physical server, which raises the overall utilization significantly and, therefore, lowers the power requirements, relative to running 10 to 15 separate servers. This alone represents a savings of millions of dollars in a year’s time, according to feedback from many companies.
Additional benefits can be seen beyond the reduced power requirements. In a traditional datacenter, many smaller servers generate a LOT of heat, which must be dissipated via cooling efforts. The yearly cooling costs are extremely expensive; but just as bad, strict formulas dictate that we can place only so many servers within so many of square feet in the datacenter. Through virtualization, we are able to place a greater number of servers per square foot of the datacenter with reduced cooling requirements. This equates in a much more efficient datacenter that is more productive and profitable.
Likewise with these other operational costs, cabling represents a HUGE investment in a typical datacenter. For every server that is virtualized, we are able to reduce the physical cabling required to connect that server. Additionally, when implemented within the Cisco UCS, we are able to take advantage of Unified Fabric and further reduce cabling requirements, since we can send both LAN and Storage signaling over the same cables.
In a traditional datacenter, if one of your Unified Communications servers crashes, you normally must restart that device manually. This represents a potentially significant outage for that UC device.
When we virtualize our UC servers, we can take advantage of VMware’s great tools like High Availability (HA) and Site Recovery Manager (SRM).
- With HA, the failed virtual server will be automatically restarted, which saves a significant amount of downtime, resulting in greater productivity and profits from that server.
- With SRM, we can provide Disaster Recovery by quickly failing over a virtual machine from a main production site to a secondary site while ensuring the VM remains active.
One of the great benefits that we’ll see when virtualizing our UC servers on the Cisco UCS platform is that we can achieve actual platform mobility. On traditional Cisco Media Convergence Servers (MCS), the UC license is tied to the physical server’s MAC address of the primary network interface card (NIC). This means that the instance of the UC application is always tied to that physical device. But what if that server hardware fails? This means that we are forced to endure some amount of outage until we can build a new server.
With the Cisco UCS, we are able to create Service Profiles that represent the identity of the actual server. We can then “associate” this service profile with any of the blade servers within the UCS 5108 chassis. By doing this, we can simply associate our UC server with a new blade server, if needed.
Of course, you may ask how this is possible when the UC license is tied to the physical NIC of the server. Within Cisco UCS, we are able to build what’s called a “MAC license,” which is based on several configuration components of the UC server:
- Time zone
- NTP server
- NIC speed
- IP Address
- IP Mask
- Gateway Address
- Primary DNS
- SMTP server
- Certificate Information (Organization, Unit, Location, State, Country)
Once the MAC license has been configured, it can simply be associated with the Service Profile and applied to whichever physical blade server is needed.
Ease of Installation in UCS Using OVF Templates
One of the barriers to entry, when it comes to UC servers, is the general complexity required to build and configure these applications properly. There is a considerable learning curve required, which requires the presence or availability of administrators or engineers with these specific skillsets.
With Cisco UCS, configuration templates are available that allow administrators to build and configure complex UC servers, often with little to no knowledge of datacenter requirements to install UC applications. To make things even easier, these templates are freely downloadable from Cisco.
The templates conform to an industry-recognized virtualization format called the OVF, or Open Virtualization Format. The OVF is an open standard for describing a virtual machine template. These templates will actually come with an .ova extension. The Open Virtualization Archive (OVA) is an open standard to package and distribute these templates. For most supported UC applications, a preconfigured OVA file is provided by Cisco and must be used. Otherwise the customer must manually build OVA files that meet the indicated requirements.
This post is reprinted and used with permission from Virtualizing Your Cisco Unified Communications with Cisco UCS