Redundancy for any UC solution, including Lync, is important. Redundancy is the provision of additional or duplicate machinery parts, systems, equipment, etc., that function in case an operating part or system fails, so that the entire system does not go down. Sounds good. However, another definition for redundancy is something that provides superfluous repetition or overlapping function. This sounds definitely less good.
When you implement more than the required redundancy, i.e. superfluous components, you both create a more complicated support environment and waste money.
When you design and deploy your Lync environment--especially if Lync is serving as your primary voice environment--there are key decisions you need to make and essential questions you need to ask yourself in order to avoid the superfluous.
The Decisions
With any UC system, including Lync, in order to architect your solution you need to make many interconnected decisions. A few of these include deciding...
1. Do we need redundant WAN connections?
This question can be asked in both the context of your central site and of branch offices.
For a single central data center, redundant WAN links can keep the data center accessible in the event of a single WAN connection failure. However, if you have duplicate data centers, perhaps a single WAN connection to each one is sufficient. Dual WAN links into duplicated data centers would actually represent redundant redundancy.
For branch sites, redundant WAN links offer the "promise" of improved voice and data redundancy...but read on for some caveats.
2. What type of PSTN connectivity model will we adopt, and what type of PSTN connectivity (SIP or PRIs) will we employ?
There are three broad architectures for Lync PSTN connectivity if you have branch sites:
a.) All centralized--in which SIP or PRIs are only connected to the centralized data center(s).
b.) Fully distributed--in which SIP or PRIs are only connected to the branch locations; i.e. there are no PSTN connections at the data center(s)
c.) Redundant--in which SIP or PRIs are connected both at the data center(s) and at the branch offices.
Option (a) is the simplest architecturally, avoids additional sites in the Lync topology, reduces the complexity of operations and may also provide more redundancy for inbound calls, as other data center components are usually more redundant than branch components and sim ring to mobile can serve as voice backup path.
Option (a), however, may pose 911 challenges (although in some cases mobile phones and analog emergency phones in key locations may address these challenges). Note that 911 in this context denotes whatever number is dialed to reach emergency services, which is really only 911 in North America. It can be many different numbers including 112, 999, 000, 110, etc.
Both options (b) and (c) provide more redundant connection points; however, depending on your use of SIP or PRIs as well as on your telcos provider's ability to automatically or manually failover inbound DIDs from one route to another, you may or may not have truly better redundant inbound call handling.
3. Should we deploy either SBAs (survivable branch appliances) or gateways at our branch offices?
This question is tied into question No. 2 in that a local PSTN connection (either PRIs or SIP) is connected to a gateway device deployed to a branch location. And a Lync server on the gateway (which makes it a SBA) provides the ability to process inbound and outbound calls should the WAN connection to the central Lync pool be unavailable.
The challenges are:
--If inbound calls are handled by a Lync response group, perhaps setup for the receptionist, the SBA will not provide redundant capability. SBAs support only basic inbound and outbound calling--calls to and from direct DIDs.
--Similarly, SBAs do not provide any branch-level voice mail or voice attendant (IVR) services.
--If you choose to connect PRIs to SBAs, any DIDs attached to these PRIs will generally represent a single point of failure. While some telcos offer number redirect services, in which a DID can be rerouted to another inbound PRI, this service is often expensive and manual as opposed to a SIP failover, which is often automatic and included in the cost of the service.
4. How long do we need voice equipment to operate in the event of a power failure?
Keeping any VoIP solution operating in the event of a power outage is certainty more challenging than it is with older TDM solutions. For a VoIP solution to operate, all of the servers, network switches and routers (both WAN and LAN) and endpoints must be powered in order for a call to be completed. Laptops, tablets and smartphones with softphone software typically have their own battery supplies, and, if charged, will provide some period of operation even when a power failure has occurred.
Data centers typically have larger, longer-lasting alternative power supplies.
This often leaves powering the office network switches and routers as the key priority. But this is not a simple task. I have been involved in several projects where providing multiple hours of uninterruptible power supply (UPS) required the floor in buildings to be reinforced in order to support the weight of the string of batteries needed. In another case, mounting a diesel generator on the top of an office tower supporting a financial trading operation required a helicopter to assist with the installation and over a $1 million investment.
Like the other decision points noted above, this may also have 911 ramifications.
Also keep in mind that in many buildings, delivering water to bathroom facilities requires power, and thus for health and safety reasons, people may be requested to evacuate the building if the power is out for more than several hours. So planning a UPS system that lasts 8 hours may be overkill.
Next page: Key Questions
Key Questions
Without asking and understanding the answers to some key business questions, even the most experienced Lync architect cannot make the right decisions related to redundancy, and this is specifically what leads to an over-engineered and overly expensive solution.
An over-engineered solution costs you more in two ways: First in the initial deployment costs and second in higher ongoing operational and maintenance costs.
You can make better redundancy decisions by asking yourself some key business questions:
1. Does work need to happen at the office? If not, in the event of a WAN failure, employees could work from home or another location (e.g. a coffee shop with Wi-Fi) or, for some professional services businesses, from a client site. If employees can work effectively from other locations, local voice redundancy and WAN redundancy may be a lower priority.
2. If physical access to the office is unavailable or branch office equipment has failed (which has happened recently in the case of some major flooding incidents), do I need employees to complete their work from an alternative location? If the answer is yes, this may point to reduced local equipment deployment.
3. If voice services are available in a branch office but no connection is available to the applications in the data center, can effective work still be done at the branch site? If the answer is no, this may point to reduced local voice resiliency requirements.
4. If local Lync voice services are available but the Lync voice services provided from the data center are not available (typically conferencing, response groups, voice mail), is this an effective work environment?
5. Would mobile phones (with sim ring for inbound calls) provide a sufficient level of voice service redundancy?
6. How can we meet desired or regulated 911 requirements? In answering this question, make sure you understand and document any legal requirements--they vary depending on geographies.
Answers to Questions Lead to Good Decisions
As with most UC architecture and design decisions, there is often no one right answer that applies to all situations. However, there is a best answer based on the specific business requirements that define your specific situation; asking the right questions to uncover these business requirements is a necessary first step. Balancing options against your specific business priorities is key to investing your UC dollars wisely. I hope that working through the above questions helps you avoid superfluous purchases.
Please use the comments section below to ask your question or comment on my answers. I will review and respond to each and every comment. Or, interact with me in real-time on twitter @kkieller.
Want to hear more about how other organizations are dealing with redundancy? Enterprise Connect 2014 and the "Living with Lync" session is a great place to hear stories from real organizations who have "real world" experience deploying and living with Lync.
Follow Kevin Kieller on Twitter and Google+!
@kkieller
Kevin Kieller on Google+
