On March 10, 2021, a fire broke out in the Strasbourg data center of the French OVH, Europe’s largest cloud service provider. The DC campus had four rooms, one of which was completely destroyed by fire and the others suffered from smoke damage. OVH’s CTO Octave Klaba has set a great example of open and detailed communication on Twitter. The Status page tells you how cloud services are recovering.
Repairs were initiated quickly and 60 employees assembled thousands of new servers a week. However, there were surprises and slowdowns along the way, so the recovery has not progressed as expected. For example, the elevator could not be used and errors occurred in a hurry, which had to be corrected later. Cleaning the servers progressed three racks a day.
The official cause of the fire may still be unclear, but the UPS, which had been serviced the previous day, was suspected to be at fault. Power cables were also replaced lately. Again, 1.4. smoke was detected on the premises and a burnt filter was found in the UPS.
The case has sparked a debate about decentralization and availability of services as well as responsibility for backup. It’s good to remember that even a cloud can break down, and in the worst case of Force Majeure, the customer isn’t even reimbursed for the outages. In the cloud service, the customer is responsible for the data and its backup. Decentralizing the service to different physical locations is also the customer’s job. Forrest Brazeal reminds us with the lyrics of the song about the importance of backup.
In Finland, Nordea’s system upgrade and data center migration took two days longer than planned. The Financial Supervision Authority had to intervene for an exceptionally long break and demanded an investigation from Nordea. At the same time, it was questioned whether banks were sufficiently prepared for possible disruptions. Payment functions are an integral part of the society’s infrastructure and also almost all identification is based on banking services.
Telcos and 5G
The 5G boom surprised Finland and 99.8% population coverage would be achieved in three years. According to Traficom, The Finnish Transport and Communication Agency, high-speed 5G is now available to almost 2 million households in Finland. Traficom updated the mobile network availability table, which shows 4G / 5G coverage by the municipality. Ficom, The Finnish Federation for Communications and Teleinformatics, published its own update on broadband subscription statistics in Finland and Europe. Also globally, 5G is coming faster than 4G.
The use of mm-waves has started in Finland, although terminal support is still lacking. Telia is testing the 26 GHz frequency in Helsinki. All three operators have adopted the former 700MHz TV frequency.
With 5G, the entire technology of the mobile operator will be revolutionized. 3GPP manages to surprise. It has abandoned all exotic closed technologies and opted for a cloud-based model based on APIs and HTTP traffic for the core network. 5G core platform is now more of a third-party platform, which the manufacturers have been opting in. An open platform brings integration opportunities for online services, but can also be a nightmare for coordination. 5G’s network slicing has already revealed security threats where information can leak between private networks.
Operators need to update their operating models and skills in today’s IT world. DevOps principles are becoming familiar, but AI is also being sought for help and even business. Companies are involved in developing the ecosystem of private networks. The corporate network and the mobile network are beginning to merge into a single entity where technologies mix. Cisco or Juniper may be a viable player in private networks because they know the traditional network-side well. Telecom Infra Project TIP has also assembled a group to develop solutions for the attractive possibilities of private networks.
Compared to the Internet, an operator’s network represents a closed network and threatens network neutrality. The technology goes back to circuit switching when users are offered an SLA-defined dedicated bandwidth. The Internet’s best-effort service and openness to everyone will be forgotten. Once Internet won by avoiding unnecessary promises and providing openness to all parties. For businesses, however, a mobile network is easier than wifi, which requires “IT adjustment.” For example, connection reliability and security, remote provisioning and multi-operator capability are important to ABB. The operator offers an easy partnership where a closed and more expensive solution meets the customer’s need more easily than a do-it-yourself network. Here is a little summary and comparison between wifi and 5G. In the future, the mobile network will infiltrate the user unnoticed with many devices and there is a risk that the user will lose control over the connections and their management.
5G networks have set speed records on consumer devices: Samsung 5.23 Gbps, Nokia 4.5 Gbps, Rakuten 1.77 Gpbs. Open RAN has amassed supporters and now the battle is whether ASIC or FPGA is the right way to implement the platform. Massive MIMO is a complex antenna for 64 transmitting and receiving elements that require heavy computation. Therefore, performance and efficiency matter. An ASIC is an efficient circuit focused on a specific function, which reduces peak power, energy consumption, and device size. The FPGA offers wider reprogrammable functions at a lower cost but consumes more energy with poorer performance. 4G is mostly based on FPGA solutions, but 5G is approached with its own SoC circuits, much of which has been invested in. However, SmartNIC maker Xilinx believes the FPGA solution will suffice for 5G as well.
Nokia Bell Labs’ Mikko Uusitalo is leading the European 6G initiative Hexa-X. In the future, the mobile network will mix machines and people into one digital world with major social and economic implications. The first publication sheds light on these visions. The ultra-high terahertz frequencies allow for high data rates, but their use would not be so much for broadband, but for human-machine communication over short distances.
In satellite broadband, Arqiva, the British broadcaster infrastructure provider, has contracted with SpaceX for ground stations and a terrestrial fiber network aimed at getting the island’s sparsely populated areas with decent broadband. Saved from bankruptcy by the British Government, Oneweb continues to build a satellite network and believes in the Arctic area networking potential. The Prime Minister of Finland is overheard to asked for service. The defense sector is also seeking to benefit from commercial networks. Lockheed Martin and the Pentagon are both planning to use 5G as the basis for their communications solutions.
Nokia hosted a week-long media show that came to the fore. Network-Nokia has repeated the same management mistakes of Phone-Nokia, which put the company back in the competition. Siilasmaa and Suri made the company’s operating method rigid and complex. The Alcatel-Lucent deal rubbed through the entire five years and drove the company into two camps, internal disagreements, increased bureaucracy and eventually centralization. The goals were made too grandiose and the product design fluctuated, which frustrated the experts. With Lundmark, cleaning began and is going to last three years. That’s how far Nokia is now behind others.
The improvement program cut 11,000 jobs to save costs and adjust the organization. The reductions are mostly outside Finland. Nokia’s personnel has decreased by 11% in two years. The reductions relate to a new business model in which each unit is responsible for its own results and performance. Technology leadership is important to Nokia, and much more is now being invested in it. The straightforward, open and technology-minded Lundmark has been well received inside the house.
Nokia ran into problems with its ambitious ReefShark chipset because manufacturer Intel was unable to meet the requirements and performance and cost-effectiveness were not what it was supposed to be. Nokia was also not good at programming FPGAs. Now ReefShark is a product name under which there can be ASIC, SoC or FPGA circuits depending on the application. Nokia has also agreed on circuit development with Marvell, but will continue to work with Intel and Broadcom. This year, 70% of base stations will be SoC-based and by the end of 2022, FPGA circuits will have been eliminated in base stations. Nokia already has good knowledge of SoC circuits.
Nokia is investing heavily in the cloud and open O-RAN model, which sets it apart from other traditional competitors. Nokia sees a revolution in the industry and is investing in the Network-as-a-Service model. Simply increasing performance is no longer enough, more sophisticated software is needed. Without flexible and agile networks, 5G will not succeed. Eventually, the web becomes like a cloud service purchased as a service.
Nokia announced partnership agreements with all three big clouds. A 5G cloud platform is being developed with Google. Customer-centric solutions and a cloud RAN platform are being developed with AWS. Enterprise cloud solutions are being developed with Azure. Suri believed in selling total solutions, but now Nokia has clearly seen that customers want to choose parts of their packages separately. Nokia is now focusing on the right things and the future looks strong if things are taken care of properly.
Cloud, Edge, SASE
AWS started 15 years ago with S3. Now S3 stores 100 trillion objects (13,000 objects for every person in the world) and handles tens of millions of queries per second. Durability was designed as a feature of the service from the beginning, providing 11 nines usability.
Programmatic underlay connections are lurking in cloud services and cloud players are trying to take their share of the connection market. Their strength is the existing infrastructure and expertise in implementing a software NaaS service. There would be demand for NaaS, but it is difficult for traditional telcos to meet the need. The MSP should be able to offer multiple solutions to find the right fit for each customer’s needs. Integrating and orchestrating many different solutions, on the other hand, is very laborious. In order to respond to the modernizing ICT world, telcos will have to invest more in IT and digital departments, which was seen in Finland years ago when Elisa and Telia grabbed smaller IT houses for themselves.
Google added a Network Connectivity Center to its service, which allows traffic to be carried on Google’s backbone network. SD-WAN and VPN connections can be managed through a single view between different devices. Google also allied with Cisco to provide automatic provisioning of WAN links on the Google network. Here’s what is the Google Network and what does it cover.
Cloudflare’s hosting and CDN platform is constantly expanding. SASE services came last year, now Magic WAN and Magic Firewall have been added to the Cloudflare One package in an attempt to facilitate connections to Cloudflare services and implement a managed security policy. Integrations can be found for Velocloud and Silverpeak. Like Google, the customer can use the Cloudflare backbone for their own connections and of course, the package includes a centralized control panel. Others who trust their own backbone network are e.g. Cato and Microsoft. If this trend continues, it seems that the open internet is turning a little while into closed networks, as has happened in content services. In any case, with the cloud, concentration begins to peak, after which disintegration and the edge begin to emerge.
SASE’s journey is just beginning and the market is fragmented. Gartner’s forecast shows strong growth for the next few years. You should be careful in your choices, as the vendor dropoff and mergers are likely to start soon. The triumph of cloud platforms now seems to continue in the connectivity market as applications have moved into the cloud. Security features are important and with SASE they are consumed as a service. Fragmented pricing of services is beginning to converge.
As the name implies, SASE is on the edge, but Benu has taken SASE services to the operator’s edge in the BNG. The idea of providing service on a broadband termination point close to the user is great. Admittedly, the same idea has been used for years in CDN services. SASE and CDN are each other’s siblings. You can marvel at these lists of cloud and edge trends, inventions and companies.
Azure’s new Routing Preference feature allows you to choose whether you want to transport your traffic over the Internet or Microsoft’s backbone. Azure Route Server facilitates dynamic routing in the cloud. Ivan Pepelnjak delves deeper into the operation, implementation, and configuration of Route Server. In conclusion, cloud communications are becoming increasingly complex and the need for skilled networking experts still exists.
AWS has increased the route limits of Transit-GW and VPN 5-10 times larger to better meet the need. My colleague Markku noticed that AWS uses E-class addresses from 240.0.0.0-255.255.255.255 range in its network. The area is unallocated and reserved for future use. Probably the motive of AWS is not to use public addresses on the internal infrastructure and the E-Class seems to look like a public address, even if it is only for internal use. The E-Class is not routed publicly and on many devices, it belongs to martian addresses that are dropped automatically. My own guess, however, is that the E-Class will someday be used for internet routing. There is such a shortage of IPv4 addresses. And that’s what AWS’s E-Class use kind of proves. In the meantime, you can watch the movie “The End” about the IPv4 address range filling.
Oracle brought L2 networks to the cloud and Ivan Pepelnjak goes through the same L2 problems that have been tried to solve for decades.
The Finnish Security of Supply Center has published guidelines for taking cybersecurity into account in ICT agreements and the Police University College published a guide on cybercrime investigation. It appears that companies poorly report cybercrime to the police and carefully weigh the pros and cons. Fear may include reputational damage, the burden of criminal investigation, or a lack of belief that the whole thing would be of any use. In addition, there are several authorities in Finland that should be notified. Supo, the Finnish Security and Intelligence Service, is responsible for national threats. Traficom, The Finnish Transport and Communication Agency, creates situational awareness and assists in the investigation. The data protection officer handles personal data breaches.
Supo said the Chinese APT31 group was behind the Finnish Parliament hacking disguising the attack as an update to the security software. Intelligence services were reported to have used home routers and NAS devices in the attacks. Supo has been said to be a dog that does not bark. Now, for the first time, it named the Chinese as perpetrators of the attack, slightly twisting and curving. In Supo’s opinion, the intelligence is problematic, because the activities that threaten security are easily taken to the shelter of the home peace, to which the intelligence does not have access. At the same time, Supo is asked to provide transparency to surveillance results of which have been told almost nothing.
Ubiquiti’s hacking is worse than earlier appeared and the company has clearly downplayed what happened. The attackers had obtained root credentials from the AWS database, from which they obtained more access credential information. At least all Ubiquiti passwords should be updated, and it could also be a good practice to delete all user profiles from devices and create them with new passwords. Of course, the devices must be upgraded to the latest software and disabling remote management, at least temporarily, should be considered.
The F5 Big-IP 11.6-12.x and Big-IQ 6-7.x were four critical vulnerabilities that were requested to be updated immediately. The Trickbot vulnerability on the Supermicro motherboard affected Pulse Secure’s PSA5000 / 7000 devices. Fixes have been released for the Unix GRUB2 boot loader against a severe BootHole vulnerability in secure boot. It is recommended that updates be installed as soon as manufacturers bring them to their devices. Chrome is now also blocking browser access to TCP port 554 to prevent the use of the NAT Slipstream 2.0 vulnerability. Microsoft released very serious DNS command execution vulnerabilities in dynamic zone updates. The old NIST DNS instructions are still valid. 3653 malicious packages have been removed from Python’s PyPI package manager. The packages easily swim through it to the developers’ own projects.
According to an IBM report, the attacks target more Linux and thus cloud applications. In the cloud, devices are more likely to be available and there is a lot of capacity available in the cloud. Automatic tools are used to scan vulnerable objects and if something is found, it is broken instantly. Anyone can fall victim. New financing opportunities are also being sought from industrial automation control systems, behind which big money and possible ransoms move.
Indeed, a blackmail attack took place in the Canadian Sierra Wireless, in which personnel couldn’t access the design and manufacturing IT systems. Although there was a clear distinction between IT and production systems, factories had to be closed. According to the release, products and online services were not affected by the attack.
Fortinet and Linksys have entered into a partnership for home network security. Fortinet has invested $ 75 million in the development of Linksys routers and the goal is apparently to strengthen Linksys devices with Fortinet security expertise. Fortinet will also be appointed to the Linksys board. So far, this is a buzzword for marketers, and it remains to be seen what the collaboration will be.
In the USA, companies and manufacturers have been required to be more transparent in their awareness and management of cyber risks, although there are no legal means to address it. Corporate data is at the mercy of manufacturers and there is no visibility into practices. CISA has proposed a security rating for companies so that cyber risks can be highlighted and decided by the board. Boards see cybersecurity as an important and influential issue for the future of companies. In the case of public companies, there would be a need for more concrete information on where the risks are and how they are managed. For example, FireEye opened its operations and took a risk. By the example, it is hoped that others will voluntarily go in a more open direction.
Solarwinds reports that the investigation and recovery of its own attack have so far cost $ 3.5 million. The price is really low if you compare it to the extensive effects that apply to other companies as well. Of course, many more costs are expected and reputational damage can hardly be offset by money. On the other hand, the matter is forgotten and the business continues as before.
Very special cybersecurity is being practiced by the Britts. The new royal multi-purpose ocean vessel will begin patrolling the oceans and monitoring submarine cables by 2024. The ship will have advanced sensors and will use underwater drones to collect data. The British and US governments are concerned about Russia’s aggressive operations in the Atlantic Ocean and the growing risks to the fibers.
For cabling, the latest green OM5 multimode fiber uses Shortwave Wavelength Division Multiplexing technology. Non-standard SWDM transmission provides 50 meters more range, standard transmission has the same distance as OM4. OM5 does little more and at speeds higher than 100G, the development goes to single-mode fiber anyway.
In copper cabling, the Cat6A, which is more than ten years old has maintained its position as the prevailing cabling standard even in new installations. The latest version of the standardization, the Cat8, raises the band to 2 kHz and provides a transfer rate of 25-40 Gbps for a distance of 100 meters. With DSL technology, 1Tbps speeds are already being planned over the telephone cable. After G.fast, we talk about Waveguide and the techniques G.nlp, G.nest and G.tdsl.
On the wireless side, the Wifi7 802.11be standard should be completed by 2024, when equipment is also expected to go on sale. 6 GHz has been added to the frequencies, the bandwidth has been increased to 320 MHz, the modulation is 4096-QAM and the channelization is 16×16 MIMO to reach speeds of 30-46 Gbps. 6 GHz is already in use in Wifi6E and the frequency is being released in the EU. Saudi Arabia is making the world’s fastest wifi by opening up the entire 6 GHz frequency band to wifi use. 6 GHz solves many wifi problems, but also means reduced range, making networks more susceptible to mesh implementations. Also in preparation is the 802.11bf SENS wifi extension, which allows wifi devices to measure distance, movement, direction and location using a wifi signal. The IEEE standard is intended to coincide with 2024 Wifi7. Other similar technologies are being developed and monitoring assumes revolutionary applications, not all of which may be excited. It must be possible to leave the monitoring.
Cisco and Apple have been working together to improve the user experience since 2015. Now Fastlane + adds predictable scheduling to your Wifi6 connection, allowing Apple terminals and Cisco Catalyst access points to negotiate an optimized latency in a congested network. The magic takes place in the background without the user’s knowledge.
In 400G optics, Infinera drives its own XR optics as standard. 400G-XR optics are used in a PON network for symmetric 400G service. It is the only coherent high-speed point-to-multipoint technology that utilizes separately routable 25G subchannels. Infinera has tested its optics in the field in many locations and expects commercial deployment next year.
Next year is expected to bring 800G optics to the switches and port density will also increase. This results in problems with component integration. Co-packaged optics are still looking for their shape.
Cisco introduced new members to the Silicon One family with peak speeds of 25.6Tbps. The advantage of Silicon One is flexibility and consistency. There are nine different models in the family for different uses. However, all the models have the same architecture and features available to be deployed per use case.
SmartNICs have taken their place in cloud acceleration. They can lighten the load on the CPU and increase the efficiency of, for example, networking, storage or machine learning. SmarNICs are usually FPGA-based cards suitable for the PCIe slot but tightly integrated with hardware. The Xilinx Alveo SN1000 card offers 100Mpps with less than 3 us delay and 4 million stateful sessions with less than 75 W consumption. Netronome Agilio LX competes at the corresponding scale.
New vNOG presentations have been released. David Roy talks about Junos inline monitoring. The purpose is to stream traffic data out of the device in IPFIX format and outsource more complex data processing to the receiver. Data can be collected from a flow or mirror with the granularity of a firewall filter, either sampled or 1: 1. So what does flow collector contain and how does it work? An example is Flowhouse. With the Linux 5.13 kernel, packet sampling adds performance metrics to the flow information. Thus, Sflow can generate real-time telemetry information about packet delays and port queue lengths. Aaron Glenn introduces what is a P4 programmable forwarding plane and what can be done with it. Cisco’s Fred Cuiller explains how to merge two large ISP networks.
While the world is becoming more and more dependent on electronics, a large solar flare could destroy much of the Earth’s electronics. Nasa estimates that the probability of a major flare this decade is 12%. The flares occur in 11-year cycles and the most recent began at the end of 2019. Scientists detect the flare with an 8-minute delay and it hits the ground within 17 to 36 hours, so it’s hard to imagine that protection would be possible in today’s world.
In The Hedge podcast, Daniel Teycheney discusses open source in networking. As with commercial products, open-source developers need feedback and experience with the code. The user does not have to bear a bad conscience about free products but can provide valuable feedback that can be used to make products better. It’s the users that make products robust.
The London Internet exchange LINX has been using disaggregation with Edgecore switches and IP Infusion’s NOS for several years. Now the model expands to the Manchester point. LINX has taken a good attitude and acted professionally in bringing disaggregation into a production-critical environment. This story is nice to follow.
Typically, the switch ASIC is configured through the manufacturer’s SDK. However, the SDK limits the features and performance of the operating system. In recent years, attempts have been made to run a new way to run the network operating system NOS directly in the Linux kernel. However, switchdev and kernel APIs have not become common among manufacturers, and only one ASIC manufacturer has written a driver that works with switchdev.
Microsoft’s Sonic has garnered popularity among manufacturers and support is surprisingly wide. Sonic seems to be the future de facto standard in open networks. Even Gartner has come to the prediction that by 2025, 40% of more than 200 switch data centers would be run on Sonic. Sonic is even expected to be a standard operating system similar to Linux, which would be a welcoming backbone in a fragmented disaggregation field.
IPSec-VPN is a bloated heavy protocol family that has been accompanied by an open and free Wireguard application and protocol in recent years. This application is also running in the Linux kernel and now BSD and pfSense are included also. Operating system support is extensive from Windows to Mac OS and Android to iOS. VPN tunneling can be used to make different topologies routed and bridged. The technology is light and efficient: in comparison, the Wireguard code is only 1% of the code in IPSec and OpenVPN. Only ChaCha20 encryption is used for encryption. Wireguard simplifies and makes connection setup easier.
The Finnish Association of IT sector Employees has published the results of the labor market survey. The average salary for full-time work is € 4,414 / month, and the largest increase, an average of € 600, is obtained by changing jobs. Salary in the Helsinki Metropolitan Area is on average 19%, or € 757 / month higher than elsewhere in Finland.
Just as the platform economy workers began to be included in employment relationships, the Technology Industries of Finland, one of Finland’s largest unions, decided to withdraw from general collective agreements. The purpose of the “labor market bomb” is to increase local bargaining in the workplace. The member companies are very different and have their own needs, justifying the solution. In the future, members of the Technology Industry will agree on matters locally and a new association of Technology Industry Employers will be established for companies related to the Convention.
The electronics industry and the circuit shortage are plagued by surprising chains of influence. Circuit manufacturing requires a lot of water, but Taiwan is plagued by drought and water use will have to be limited. The same has been heard from German factories. The drought, on the other hand, is due to climate change, which no longer brings typhoons to Taiwan as often as before.
Broadcom’s order backlog this year is 90% full when the normal level would be about 25%. Delivery times have been at their worst for 8 months. Huawei has succeeded in developing its own HiSilicon circuit, but manufacturer TSMC imposes restrictions on availability. Huawei’s most successful part is the enterprise unit, which makes a difference to the telecom sector. Compensatory development has been directed at the cloud, software and other ICT sectors.
Cisco and Huawei hold the top spot in Dell’Oro’s enterprise network equipment market statistics for last year. Cisco’s 40% market share is on its own, and the next Huawei will just get a double-digit share of 10%. The statistics include everything from wired networks to wifi and from WAN to security devices. Therefore, Paloalto and Fortinet, who take places 4 and 5, are surprising on the same leg as Aruba. In general, sales of the campus and data center network decreased and router and data security were on the rise with the sale of software and licenses. Wifi remained flat.
Arista has hijacked about half of the leak from Cisco and the rest has gone to whitebox manufacturers. Microsoft and Facebook have been Arista’s biggest customers, but cut purchases last year. Arista’s customer base has grown from 1,100 to 7,200 in ten years. HPE, on the other hand, has fared better than expected, with Aruba and SD-WAN acting as drivers of digital change. Aruba Central’s cloud management has also strengthened Aruba’s position, but the integration of Silver Peak is an important area for development to take advantage of the additional features it brings. Aruba changed Central’s licensing model and raised prices as the cost increase brought about by the cloud infrastructure. Aruba Central has been running in AWS, but now it is also possible to get it from Azure. In addition, the Microsoft collaboration brought Aruba IoT Transport to Azure, which allows IoT devices to be connected to the cloud via an access point or a controller-based IoT hub.
Fortinet buys ShieldX to protect multi-cloud infrastructure. The Orca Security, founded by former Checkpoint employees, is attacking Paloalto and plans to defeat its customers for itself. The promise is truly functional and comprehensive cloud security. The relationship between the companies has embarked on its own career when Orca angered Paloalto by publishing a product comparison video. Paloalto responded with a threatening legal challenge and the soup was ready. Since then, public posts have been exchanged through blogs.
Equinix has introduced a new data center solution based on factory-built modules. Modules can be used to build capacity in smaller units and to produce energy flexibly in different ways. The model will be tested for the first time in Bordeaux, France. Equinix is the clear leader in recent Interconnection market statistics, which shows that Equinix has a total of 392,000 connections. Digital Realty and Megaport follows.
Russia has tried to incite the uncooperative Twitter. At first, Twitter traffic was slowed, but the effects hit other Internet traffic more broadly. Rostelecom had blocked all of Twitter’s abbreviated t.co domains, but someone had a bad day with the regex, and all domains containing t.co were blocked, including e.g. microsoft.com and reddit.com. Internet traffic in Russia dropped by 24%. The next step was to block the entire use of Twitter if the malicious content does not leave the service.
Blogs recall 2014 when the size of the routing table exceeded 512,000 routes and routers crashed due to lack of memory. The routing table size is now expected to exceed one million routes at approximately the end of 2023. Before that, there will be a full 128,000 routes in the IPv6 table. Now the table sizes are in the order of IPv4 860,000 and IPv6 109,000. In IPv4, the number of /24 prefixes is growing steadily and is now already 60%. For current routers, the size of the table is unlikely to be a bigger problem if the device is designed for peering and has enough RAM.
On the RPKI front, Lumen (formerly CenturyLink, formerly Level3), one of the largest tier-1 operators, has introduced RPKI validation of routes. Qrator Labs explains the difference between a BGP route leak and a BGP hijack. And what RPKI and ROA really matter.
The U.S. DoD has made a plan to sell all 13 Class A IPv4 addresses it owns. Yes, the package has a juicy market value to finance other projects and activities.
In Asia, the trade war is building a capacity wall as the U.S. has forced large producers of submarine cable capacity to put their Pacific projects on hold. Australia has done the same in the past. Facebook has planned two cables for Echo and Bitfrost from the west coast of the US to Indonesia and Singapore. However, cable capacity delays do not have a significant impact on the economy, but may feed local capacity concentrations in, for example, Japan and Taiwan.
An interesting question for fiber investors: how long does fiber last? Fiber manufacturers do not give fibers a lifespan. In practice, it can be seen that the fibers of the 1980s are starting to age. Since then, manufacturing methods, quality, and processing have evolved considerably, and newer fibers no longer have the same problems. Unofficially, researchers have stated that the fiber should last 75 years or more.
We, network engineers, are all in the same boat, so being empathetic is important. It makes no sense to joke or laugh at the problems and harms of others because one day the shit hits your own fan. Rather, you should sympathize with others and try to be helpful, understanding, and encouraging. Hugops has found a place especially in big and serious cases like the OVH fire, but why not also in every case where people need support.
We keep arguing about complexity, but for good reason. Why does the network always owns and solves everyone else’s problems and fulfill the wildest desires? Experts themselves make networks complex and cumbersome, even if less is enough for business. It’s time to say no and move features where they belong. Keep the network neat and simple. Then maybe automation can succeed.
Cisco has signed an agreement to integrate Terraform’s cloud service into its own Intersight multi-cloud management tool. Terraform’s IaC tools are available as a SaaS through Cisco Intersight.
Running lab devices in your own environment is easy as long as you can download an image from somewhere. Ethan Banks has compiled a list and instructions. For some manufacturers, it is easy (Arista, Cumulus, Aruba), for others more difficult (Cisco, Juniper).
At the level of abstraction, network validation and testing is a difficult and multifaceted topic.
Microsoft Ignite introduced a bunch of announcements. CTO Mark Russinovich gave a broad Datacenter presentation. At time 19:17, WAN will be traversed: more than 130,000 km of fiber, more than 180 edge sites, 149,000 RPKI-signed routes, to connect Azure Orbital space devices to Azure. At time 27:46 a few words about network management. At time 42:04, visions and crazy images of liquid cooling. At time 56:10 chaos testing with Chaos Studio.
A different event IETF110 was again held virtually and all sessions can be found on Youtube. Videos provide a strange experience of following the work of working groups and long sessions on all possible topics.
MWC Barcelona, scheduled for June, has suffered a drop in attendance. Of the big ones, only Verizon has confirmed participation, the others have canceled one after another. The tenacious cloud consultant planned to use the money to rent an Ericsson booth and proclaim the good news of the public cloud in his own “Cloud City” booth. The German broadband congress Angacom has also been virtualized this year and will be retried in person next year.
The Finnish branch of an international company was hit by a disaster, in which the whole service built on top of the ERP system, broke down. DNS had also been baked into the ERP system, and when it stopped working, the entire service stopped exists. The recovery seemed to be a desperate battle of a few people against the ballast of history. Let this be an example of technical debt and its consequences.