What are the four ways to mitigate latency and packet loss?

By Graham Jarvis, Freelance Business and Technology Journalist, Lead Journalist – Business and Technology, Trudy Darwin Communications.

  • Thursday, 16th May 2024 Posted 3 weeks ago in by Phil Alsop

Latency and packet loss worsen Wide Area Network (WAN) performance, causing network jitter. To solve this, organisations often feel that the solution is to buy bigger and bigger pipes, but this doesn’t necessarily improve bandwidth utilisation and network performance. Equally, John Burke, CTO of Nemertes is right to suggest that the “performance of cloud-based apps that create a lot of network traffic can be hurt by network loss and latency.” At least, that’s what the standfirst says. 

In his November 2022 article for Network World, ‘3 Ways to Reach the Cloud and Keep Loss and Latency Low’, he states that there are ways to mitigate, rather than completely resolve issues such as latency and packet loss. This also applies to the adoption of public cloud IaaS platforms, such as AWS and Azure, and PaaS and SaaS solutions. They have been promoted on the basis of the simplicity of consuming services, whereby it’s possible to connect securely over the public internet to “start spinning up resources.” However, he says private communications with those resources still leaves some challenges and choices that have to be made. 

His answer is to use the internet with a VPN to connect to virtual private clouds (VPCs) “or their equivalent from company data centres, branches or other clouds.” However, again, latency and packet loss can become troublesome – even more so over significantly long distances.  

In his view, the answer to resolving these issues is threefold: a dedicated connection to the cloud via an organisation’s private network; the utilisation of an exchange simplifies the process of connecting to multiple cloud providers or connecting more flexibly to any provider; and internet access to an exchange to maintain direction connections. One means of mitigating latency and packet loss is missed out, though. That is WAN Acceleration, which shouldn’t be confused with WAN Optimisation.

WAN Optimisation v WAN Acceleration

The differences between these are explained by Sudip Saha, Managing Director and Co-Founder at Future Market Insights, in his article, ‘Global WAN Optimisation Market Poised for Robust Growth, According to New Analysis Report.’ WAN Optimisation is a good technology but, unlike true WAN Acceleration, it can’t send and receive encrypted data natively. Therefore, to optimise the encrypted data, it has to have the keys to decrypt the data. WANop then optimises it before encrypting the data back again and sending it. In contrast WAN Acceleration can send and received encrypted data without this lengthy process. It can also achieve higher data throughputs than WAN Optimisation often can.  

WAN Acceleration also uses artificial intelligence, machine learning and data parallelisation to better control and accelerate data flows, while mitigating the effects of latency and packet loss. WAN Acceleration also boosts bandwidth utilisation by at least 90%. Unfortunately, in comparison, WAN Optimisation often fails to live up to its vendors’ promises. 

Saha, nevertheless, comments in his article, published on Marketresearchblog.org on 21st March 2024: “According to the detailed report by Future Market Insights, the global WAN optimisation market recorded sales of US$ 935.1 million in 2018. The market is anticipated to achieve revenue of US$ 1,150.9 million in 2023. Over the forecast period, the global market is projected to expand at a CAGR of 6.2%, resulting in a market size of US$ 2,106.9 million by the end of 2033.”

That aside, Stephen Pritchard reports for ComputerWeekly in his article, ‘Cloud-to-Cloud Backup: What It Is and Why You (Probably) Need It’, that organisations often use cloud IT or cloud-based applications in “the belief they can spend less time managing IT and more time running the business.” He says this is because people believe that this technology is both robust and reliant. Yet, when it comes to data protection, the absolute opposite can often be true.  

He explains: “In fact, the basic protections for files and applications provided to customers by cloud and software-as-a-service (SaaS) suppliers can be very limited. This means CIOs need to make their own arrangements to back up data in the cloud.” Quite often, this is done with WAN Optimisation or SD-WANs, but even the latter benefits from true WAN Acceleration solutions, such as PORTrockIT. 

Network Wild West

David Trossell, CEO and CTO of Bridgeworks, reminds us that as soon as data is let loose outside of a data centre, control over the data transfer is lost. In other words, “We can no longer manage congestion, packet loss or latency – the killer of TCP/IP performance and lastly bandwidth.” He adds that organisations are, from the point, in the hands of the carriers and cloud companies. “This purely a money issue – the more we spend, the more control on the key factors that affect performance,” he says.  

With regard to the internet – the value end of connectivity to the cloud is for organisations that don’t have massive data transfer requirements. He explains: “It depends where you are located, city versus rural. However, you take your chances with this method, as you are fighting with others with line contention and wildly varying latency and packet loss.” 

“Another issue with the Internet is when the packets you send over the internet take different paths, packets can arrive out of order! We have to wait for the missing packets to arrive or request them again.   If you are going to use the internet – please use a VPN. One word of caution, the VPN will not solve the issue of packet loss or cloud.”

The network value chain

Moving up the value chain, there are a few more options. However, they often come at a significant cost. Those options include dedicated links to a cloud provider. Trossell says they typically include 1Gb to 10Gb WAN connections. “Here your latency and packet loss are much more stable, and under control,” he suggests, before commenting that the inherent performance killer of TCP/IP and network performance is latency. So, an organisation that wishes to egress and ingress large amounts of data may find that their WAN performance is inhibited, and that less performance than is expected comes out of having spent loads of pounds or dollars on increasing bandwidth. 

He warns: “A note of caution that many of the cloud providers use the big carriers to provide their networks. These carriers sometimes oversell the bandwidth hoping not everyone will use the full capacity of their allocated bandwidth. A bit like the airlines selling more seats than they have expecting some passengers not to turn up.  It happened to a large one of our large customers.”

Dedicated private bearer

The solution offered by most carriers is to provide organisations with a dedicated private bearer, linked to an internet exchange. This goes out from that exchange to the internet and then to the cloud companies. Trossell says carriers will happily provide you with all the bandwidth you can afford. He finds that this is ideal with those organisations that have a large presence in the cloud, and/or use the cloud as an offsite backup location.

As mentioned earlier, SD-WANs are another option for access to cloud providers and the internet. While he says SD-WANs provide more control over the access from the user to the cloud with the ability to separate traffic destined for the internet from the cloud traffic, they often need a WAN Acceleration overlay to boost their performance by mitigating latency and packet loss. SD-WANs on their own and WAN Optimisation don’t often go far enough to address latency and packet loss. 

Trossell concludes that whilst latency and packet loss may be a small issue to organisations with low data volumes, it’s an issue that starts making itself very noticeable as you move up the bandwidth ladder and large data volumes. He therefore recommends WAN Acceleration, and stresses that using a VPN makes no difference and has no impact on latency. Anyone using a VPN is still using the internet, and all that it does it protect that person’s data from others reading it. 

By AI expert Stuart Fenton.
By Mark Robertson, chief revenue officer and co-founder at Acumen.

How MSSPs can capitalise on compliance

Posted 2 days ago by Phil Alsop
By Innes Muir, Regional Manager, MSSPs, UK, EIRE and RoW, at Logpoint.
By Richard Eglon, CMO Nebula Global Services.
By Richard Connolly, Regional Director for UKI at Infinidat.
By Charles Courquin, Director, Symatrix.

Bridging the AI skills gap

Posted 1 week ago by Phil Alsop
By Tracy Stanton, SVP Client Services, EMEA at Magnit.
By Jake Madders, Co-founder and Director of Hyve Managed Hosting.