At this point, the spread of COVID-19 seems inevitable in much of the world. Working from home, if possible, is more than just a good idea. It’s a mandate from employers and governments around the globe.
While many folks are considering the social and business implications of a potentially permanent increase in remote workers, at TeleGeography we’re of course thinking about what this means for WAN and IT infrastructure managers worldwide.
Remote work is most feasible for knowledge workers who spend most working hours in front of a computer. But even within industries such as retail or manufacturing, there are back office jobs that can be done remotely.
TeleGeography and Capacity Conferences recently hosted a discussion on this topic as part of our Virtual WAN Summit, which replaced our COVID-19-postponed WAN Summit NYC.
Ciaran Roche of Coevolve moderated the discussion with three WAN managers: Leonard Bernstein from C&S Wholesale Grocers, Michael Wynston from Fiserv Technology Services, and Michael Martin from McKinsey and Company.
While COVID 19 won’t last forever, the lessons from this time can extend far into the future. As Bernstein said, “[We] must embrace a concept of remote work, not just as BCP strategy, but to ensure workers can be productive at all times.”
In this post I’m going to bring together thoughts from this session with TeleGeography’s market data, within the context of these elements of the remote work surge:
Perhaps the first thing that comes to mind in home or remote work is the availability of a broadband connection. Broadband penetration rates in most developed countries are around 90%.
Who Actually Has Broadband?
This number is certainly higher when dealing with urban and suburban zones around major employment centers. In fact, Micheal Wynston, Director of Network Design and Architecture at Fiserv, said that they “are finding some of our end users have more bandwidth than they even have in the office. Typically they have 100 Mbps to GigE connectivity.”
Micheal Wynston, Director of Network Design and Architecture at Fiserv, said that they “are finding some of our end users have more bandwidth than they even have in the office. Typically they have 100 Mbps to GigE connectivity.”
By comparison, our WAN Manager Survey indicates that the most common office connections capacities–whether MPLS, DIA, or broadband–are in the 11-50 Mbps range.
What Are the Typical Capacities for MPLS, DIA, and Broadband?
Compare that to Akamai’s 2017 State of the Internet Report, which lists seven states and the District of Columbia—along with locations like Singapore, Hong Kong, and South Korea—as having average peak connection speeds of ≥100 Mbps.
But what about backups?
Well, 54 million households in the U.S. have access to at least two wireline broadband providers of at least 25 Mbps. Further, most wireless providers offer some kind of unlimited or high-utilization LTE or even 5G data plan that can be used as a hotspot failsafe. As a personal example, I live in Northern Virginia where I have access to 100+ Mbps plans from Verizon Fios (FTTh) and Comcast Xfinity (DOCSIS), as well as an unlimited LTE data plan from T-Mobile.
But that is all about consumer broadband. What about business broadband?
Our Business Broadband pricing app covers 7,000 plans in 170 countries; many of these plans are likely available to home workers. Business broadband is often just consumer broadband with a few extra guarantees such as annual availability or features like static IP addresses.
While it may cost more than consumer plans, business broadband prices still pale in comparison to MPLS or DIA.
While it may cost more than consumer plans, business broadband prices still pale in comparison to MPLS or DIA. So there may be a bright side to the future of remote work. I would wager that, in most cases, the cost of a home broadband connection is passed onto the employee anyway. It is hard to see corporations take responsibility for sourcing thousands more internet connections themselves.
How Do Business Broadband Prices Compare to MPLS and DIA?
One very possible downside that we’ll have to look out for is local DSLAMs and the like getting slammed (pun intended) with unusual amounts of daytime traffic. It’s possible that even local wifi routers would, at times, be unable to handle all the additional traffic/devices (especially with the kids home from school at the same time).
Perhaps it's counterintuitive, but it seems the bottlenecks are often not at the remote or home site, but rather the aggregation points. While remote users often have access to high bandwidth broadband, the pressure points can often be the connection point from the public internet back into the corporate network.
While remote users often have access to high bandwidth broadband, the pressure points can often be the connection point from the public internet back into the corporate network.
A surge of users outside the network trying to get back in can put stress on the static parts of the network, said Roche. He explained that the “WAN business has historically not been able to deal with big upswing[s] and then downswings [of bandwidth]."
Many workers are likely connecting over an IP-sec/VPN connection, and suddenly increasing this traffic may put strains on the gateways to those services. As Martin explained, “We already have an extensive remote access platform. The challenge is making sure that pathways into the global network have enough capacity to meet the increased usage.”
“We already have an extensive remote access platform. The challenge is making sure that pathways into the global network have enough capacity to meet the increased usage.”
Bernstein brought up two additional issues with bandwidth constraints: licensing and hardware capabilities. “The challenge for us was capacity and we were looking for temporary licenses for VPN users,” he said. “We determined we could not go that route as we were midstream on edge data center points, and even if we could get temporary licenses, the devices were incapable of taking on more users.”
Finally, Roche brought up a key point in all of this: imperfectly aligned views of the security and networks teams. “[There has] always been some tension between network and security teams. The network team might want to avoid backhauling all traffic through an internal data center or firewall and use more SIP tunnelling and local breakout.”
Security teams however, particularly if they don’t have a non-centralized authentication system in place, may disagree.
Bernstein responded that they were transitioning to Z-scaler, which was allowing them to secure locally broken-out traffic rather than having it hit central firewalls.
“Our end users don’t have to come all the way back to the data center to hit a data center-based internet connection, which makes it a much more efficient process.”
Wynston agreed on avoiding the local breakouts even though they are not using Z-scaler. They reconfigured their actual devices so the authentication goes through the VPN then to local breakouts. They could do this after building out edge nodes using NFV and Versa. He explained, “Our end users don’t have to come all the way back to the data center to hit a data center-based internet connection, which makes it a much more efficient process.”
We’ve been tracking SD-WAN adoption through live polls at the WAN Summit (along with our WAN Manager Survey) for years.
At the 2019 WAN Summit London, 22% of end-user poll respondents had already implemented SD-WAN on at least part of their network. In 2016, only 3% of end-user attendees had. However, many enterprises are still evaluating SD-WAN options. About 40% of 2019 Summit attendees indicated they were in the process of selecting a supplier. For that 40%, this might be a teachable moment on the utility of adopting SD-WAN now.
About 40% of 2019 Summit attendees indicated they were in the process of selecting a supplier. For that 40%, this might be a teachable moment on the utility of adopting SD-WAN now.
While “social distancing” might lead to a much greater reliance on UC and various web-based tools at the expense of in-person meetings and more traditional communications technologies—like SIP trunking and ISDN that are not easily extended to remote workers—the real key is that WAN managers want those tools to function and feel the same at home as they do in the office.
SD-WAN has already shown it can help with this. As Roche said during our webinar, “Small SD-WAN endpoints allow you to have it live on top of whatever home infrastructure that individual might, but get the look and feel down to IP addressing and business policies that you would have in an actual office. This is a great difference to how this would all look five years ago.”
First, it’s important to note that SD-WAN can be useful in facilitating VPN connections into the WAN, but it can also just extend the WAN everywhere.
Wynton said they have actually sent employees home with small “almost NOC-like” SD-WAN devices and many SD-WAN providers actually have devices aimed at the small/home office in addition to those set up mainly for larger corporate uses. This can essentially extend WAN services to remote workers and seamlessly extend policy changes, security updates and other centralized activity directly to remote users–assuming of course the SD-WAN controllers can handle lots of additional devices.
On the network side, Martin pointed out that the key for them is the flexibility of SD-WAN. Particularly in terms of bandwidth. As he explained, “If you are leveraging the large cost savings in terms of ratio increases in bandwidth … [SD-WAN] makes it possible to almost disaster-proof your environment.”
Indeed, I very much agree, and we have lots of data to show why!
When we compare the cost of traditional MPLS networks (with backups) to the more flexible hybrid or all internet networks, a couple of things stand out:
This chart shows our latest data filtered through our hypothetical network. The MPLS network had an average of 125 Mbps of active bandwidth per site. The other networks have between about 420 Mbps and 500 Mbps.
Can You Get More For Less?
Tons of additional (and flexible) bandwidth can make a lot more sense in disaster situations than expensive MPLS backups.
Even if that additional bandwidth is used—or spun up on demand—mostly at the data center or cloud connections, SD-WAN can facilitate that flexibility in both cost and functionality.
Our panelists tended to agree that SD-WAN made office closures much easier to deal with from a WAN perspective and elements of dealing with this exigency are perfectly applicable once we return to normalcy.
I certainly agree that the low cost and high availability of high-bandwidth consumer broadband mean that it’s easier than ever to extend the office environment to other locations. I also agree that the additional cost of SD-WAN can be easily erased both by the ability to leave behind more expensive network services, amounting to a small portion of most network costs.
Read more of our COVID-19 analysis here.