News Stay informed about the latest enterprise technology news and product updates.

SD-WAN a tool for combining networks, engineer says

When your company acquires another, combining networks can be a chore. But SD-WAN could ease the pains associated with integrating disparate topologies.

It's a good thing to work at a growing company. But it can be a bad thing when that growing company acquires another,...

and you're the one charged with combining networks into a cohesive whole.

Ethan Banks, writing in PacketPushers, said the pressure to integrate applications and systems can be intense, leading engineers to cobble together quick-and-dirty options to keep the data flowing. But those options -- say, a quick-and-dirty IPsec tunnel -- can cause headaches later on.

Yet, there might be another approach to ease the pain associated with combining networks: SD-WAN. Software-defined WAN can be the glue engineers are looking for, Banks said. Among the technology's advantages, it's easily managed, offers redundant connectivity and it supports the Interior Gateway Protocol, including the use of a dynamic multipoint virtual private network. In addition, Banks said SD-WAN permits network segmentation and service chaining. Banks also listed some caveats, among them cost and complexity.

Still, he said, "I see SD-WAN as a way to onboard an acquired network permanently, while retaining the fast time to connect that an IPsec tunnel offers. For organizations who already have an SD-WAN solution in place, there's not much to think about. For organizations who haven't invested in SD-WAN yet, this might be an additional driver to do so."

Read what else Banks has to say about using SD-WAN as a tool for combining networks.

Juniper's embrace of automation and what to expect

Dan Conde, an analyst with Enterprise Strategy Group in Milford, Mass., said he expects Juniper Networks to use its annual conference to shed more light on its Self-Driving Networks initiative. The company last week released details about a trio of bot apps engineered to automate telemetry, auditing and peer monitoring, which will be released early next year.

Conde said he believes this is just the start. "Juniper has been an advocate of automation for a while," he said, citing first-generation devices that relied on APIs instead of command-line interfaces to program them. "Automation is nothing new to them."

What is new is layering intelligence to automation, giving the software the ability to adjust network performance as needed. Conde said it's immaterial what role the intelligence is used for -- whether it's to check configuration or status. What is important is automating as many processes as possible.

"I look forward to a day when even more items get automated, and when IT pros will someday leave behind their skepticism and conservatism on automation and embrace timesavers that make their lives easier," Conde said.

Dig deeper into Conde's thoughts about Juniper's strategy.

What Google's nifty chip may say about AI

So, Google's nifty AlphaZero computer algorithm not only learned how to play chess in four hours, it went on to demolish Stockfish -- known as the highest-rated chess computer -- in a match of the gadgets.

After 100 games, it was AlphaZero 28-0 over Stockfish, with 72 draws, said Brad Shimmin, research director at Current Analysis Inc. in Sterling, Va., in a recent post. But it wasn't about the game. Instead, it was about the chips.

Google engineered AlphaZero with 5,000 AI-specific TensorFlow processing units (TPUs), which the machine used to "learn" how to play chess. The machine also had 64 second-generation TPUs that provided the necessary neural training. Once the games began, Google stripped AlphaZero down to only 4 TPUs, which is all the machine needed to defeat Stockfish.

"AlphaZero's mastery of chess stemmed from the sheer, brute force of Google's AI-specific TRUs," Shimmin said, adding that each TRU can deliver up to 225,000 predictions per second. A conventional CPU, by contrast, can only churn out 5,000 predictions per second. "It is this hardware-driven ability to iteratively learn at speed that unlocks the door to AI's potential," Shimmin said. "That's where we'll see the most innovation and competition over the coming year as vendors speed up AI through purpose-build hardware."

Check in with Shimmin to read more about what Google is trying to do.

Dig Deeper on Network Security Monitoring and Analysis

Join the conversation

1 comment

Send me notifications when other members comment.

By submitting you agree to receive email from TechTarget and its partners. If you reside outside of the United States, you consent to having your personal data transferred to and processed in the United States. Privacy

Please create a username to comment.

How have you attempted to integrate disparate networks?
Cancel

-ADS BY GOOGLE

SearchSDN

SearchEnterpriseWAN

SearchUnifiedCommunications

SearchMobileComputing

SearchDataCenter

SearchITChannel

Close