OVHcloud Continues Making AI Simple With New NVIDIA Tensor Core GPUs and Top-of-the-Line AI Models

Posted in Commentary with tags on June 6, 2024 by itnerd

OVHcloud, a global player and the European cloud leader, has taken the stage at Adopt AI, France’s new AI-focused event, to renew its commitment to democratize AI within organizations, an effort aligned with its core values for an open, transparent and trusted Cloud. In a booming AI market, OVHcloud aims to offer businesses the optimal capacity for their every AI needs through a wide range of solutions positioned at every price point. The Group’s unique value proposition aims to bring simplicity to every level of the AI pipeline while addressing the growing concerns for security, data privacy and costs that could further delay AI adoption according to recent studies. 

To propel AI into new territories, OVHcloud now offers a broad portfolio of NVIDIA Tensor Core GPUs and the latest open-source LLMs (like Mixtral 8x22B or Llama 3) available on the shelf with unmatched simplicity in a private environment. The Group relentlessly executes on its strategy focusing on the four key items that organizations need to be successful with AI: powerful yet efficient compute resources, streamlined datasets, software tooling and cutting-edge skillsets. 

Offering NVIDIA accelerated computing for every AI workload

Strategically designed to meet the needs of AI engineers today and tomorrow, OVHcloud solutions turbocharge projects from the creation of AI models to inferencing the very latest Large Language Models. As such, the Group now offers a broad portfolio of state-of-the-art NVIDIA GPUs in its Public Cloud universe, available on demand in ecofriendly datacenters, including the NVIDIA H100 Tensor Core GPU, NVIDIA L4 Tensor Core GPU and NVIDIA L40S GPU. Regardless of whether they need to provision and manage GPU instances themselves or leverage them within OVHcloud Managed Kubernetes Services, MLOps Engineers can benefit from the full range of OVHcloud PaaS offerings, including Managed Databases (such as Vector DB like pgvector or Qdrant), Kafka, as well as high-performance Object Storage.

Further delivering on its ambitious AI roadmap, OVHcloud has introduced NVIDIA L40S GPUs on its Public Cloud in addition to the NVIDIA H100 and L4 GPUs which are already available. NVIDIA L40S GPUs combine powerful AI compute with best-in-class graphics and media acceleration to power a wide range of workloads from LLM inference and training to 3D graphics, rendering and video.

Dedicated Bare Metal servers now with NVIDIA L40S GPUs

Addressing the need for powerful dedicated AI servers, OVHcloud has also introduced new Bare Metal dedicated servers powered by NVIDIA L40S GPUs. Sitting in the High Grade range, the new HGR-AI-2 server targets use cases such as deep learning, machine learning and high performance computing. The range is equipped with AMD EPYC 9354 in a dual processor configuration, thus featuring 64 cores and 128 threads with 384 GB of DDR5 memory (up to 2.3 TB). To tackle serious AI workloads, new HGR-AI-2 features 2x NVIDIA L40S 48 GB GPUs with the ability to configure up to 4x NVIDIA L40S 48 GB GPUs per server.

Easy to use tools to gets hands on with AI workloads

To further simplify AI and democratize the way businesses use AI, OVHcloud AI solutions sit at every step of the AI value chain from ideation to production. The easy-to-use, AI Notebooks, AI Training and AI Deploy services act as a complete set of serverless tools designed to experiment with data, train models and put them into production. The high-level software solutions leverage industry standard technologies and are designed to assist data scientists and machine learning engineers without the need to manage complex infrastructure scaling, and with minute-by-minute billing.

New open-source Llama and Mistral models available through AI Endpoints

OVHcloud AI Endpoints is another serverless solution that enables access to a diverse set of AI models through API endpoints thus requiring little to no knowledge on the user side. Before placing their API calls, developers can benefit from a playground to test their Endpoints. Just like critically acclaimed OVHcloud AI Notebooks, AI Training and AI Deploy, AI Endpoints leverages the Group’s infrastructure, fully respecting data privacy (in/out). The Group has added highly sought-after open-source LLMs like Mixtral 8x22b and Llama 3 that can be deployed through the OVHcloud AI Endpoints service.

Pricing and availability

New NVIDIA GPUs options including NVIDIA H100, L4 and L40S are available now in the Public Cloud universe to execute AI driven projects.

New HGR-AI-2 servers with the NVIDIA L40S GPU are available now in France and Canada, starting at 2969,99 euros/month.

New models including Mixtral 8x22b and Llama 3 are available now through OVHcloud AI Endpoints.

Resources

Australian Mining Company Pwned By BianLian

Posted in Commentary with tags on June 6, 2024 by itnerd

Recently, Australian mining company, Northern Minerals, disclosed a breach after BianLian leaked data. Exfiltrated data includes corporate, operational, shareholder and financial information, along with detailed related to current and former personnel. The company noted that the incident did not affect mining or business operations. 

Darren Williams, CEO and Founder, BlackFog:

“The attack on Northern Minerals is an example of how ransomware gangs will proceed once their demands are not met. While this is an unfortunate reality for the shareholders and the organization itself, it reflects a common trend among ransomware gangs that are focused specifically on data exfiltration rather than encryption.  Today’s gangs are about extortion and the many ways data can be leveraged for financial gain, so the focus must revolve around data security and preventing data exfiltration to protect organizations from these ongoing threats.”

The best way to make the activities of these gangs less profitable is to make sure that they can’t get into your environment. Because you can’t be extorted if they can’t get in.

TikTok Pwned… Big Name Accounts Hacked

Posted in Commentary with tags on June 6, 2024 by itnerd

TikTok has been pwned via a zero day related to opening direct messages. That’s bad. But it gets worse:

According to the report, the bad actors behind these attacks are transmitting malware through DMs on the popular social media app. Distressingly, the malware does not require the victim to download any software or click on a malicious link. Apparently, all you have to do to potentially infect your device is open a DM with the malware present.

This is seemingly connected to a separate report from Semafor, which noted on Sunday that a hacker broke into CNN’s TikTok account last week, prompting the news organization to take the account down. While some CNN staffers told Semafor that the team had been lax with its digital safety practices, one network source added that the breach came from outside of the company. It’s entirely possible that a staffer opened one of the malicious DMs.

A CNN spokesperson told Semafor that the company is “working with TikTok on the backend on additional cybersecurity measures” to ensure that their account is secure during the presidential debates and other noteworthy events in the coming weeks. At the time of writing, the main CNN TikTok account is still unavailable on the platform.

Apparently other TikTok accounts have been pwned as well. What makes this worse is that anyone can be a victim of this attack. That’s not good. Ted Miracco, CEO of Approov had this to say:

   “The TikTok security issue serves as a crucial wake-up call about the vulnerabilities present in apps from official stores. It highlights the need for greater transparency, cooperation, and the adoption of advanced security measures to protect users. Many users believe that apps available on official app stores are inherently safe. This perception is perpetuated by the rigorous app review processes that Google and Apple claim to implement. However, the reality is that even apps from these stores are not immune to zero-day vulnerabilities and sophisticated attacks. 

   “Zero-day vulnerabilities, by definition, are previously unknown flaws that can be exploited before the developer has a chance to address them. The presence of such vulnerabilities in popular apps like TikTok is alarming. It indicates that even the most widely used and scrutinized apps can have hidden security flaws, putting millions of users at risk.

   “To build a more secure app ecosystem, transparency is key, and independence from Google and Apple controlling the release process. Developers need to control the process, and users need to be informed about the security measures taken by app developers and the potential risks associated with using certain apps. Furthermore, the industry must move towards open standards and greater cooperation among stakeholders. By addressing these challenges, we can work towards a safer and more secure app ecosystem.”

This is likely the last thing that TikTok needs. Hopefully they are completely transparent about what’s going on here and how they plan on addressing this. Otherwise this will add to their problems.

A Follow Up To London Hospitals Being Pwned In A Ransomware Attack

Posted in Commentary with tags on June 5, 2024 by itnerd

Following up on this story about London hospitals getting pwned by a ransomware attack, I have more details on this. On the surface, this appears to be yet another third party attack.

The attack targeted Synnovis, a third-party provider responsible for pathology services such as blood tests, swabs, and bowel tests for the affected hospitals. As a result, crucial services like blood transfusions and test results have been disrupted.

In response to the attack, both Guy’s and St Thomas’ and King’s College Hospital have declared a “critical incident.” The Independent reports that general practitioners have been instructed to cancel all non-emergency pathology appointments, and hospital staff have been directed to request emergency blood samples only for patients requiring transfusions. 

A spokesperson for NHS England’s London region stated, “This is having a significant impact on the delivery of services at Guy’s and St Thomas’, King’s College Hospital NHS Foundation Trusts and primary care services in southeast London, and we apologize for the inconvenience this is causing to patients and their families.” Despite the disruption, emergency care remains available, and patients are advised to continue accessing services as usual, including dialing 999 in emergencies.

According the The Record, the NHS suffered 138 ransomware attacks in just 2022 and 2023.

Experts with Cyware and Horizon3.ai offer perspective:

Emily Phelps, Director, Cyware had this to say:

   “These ransomware attacks targeting healthcare and critical infrastructure reinforce the urgent need for a collective defense security approach. This incident, which has  disrupted critical healthcare services, highlights the vulnerabilities inherent in siloed systems that slow down response times. 

   “A unified defense strategy and modernized security operations not only mitigate risks but also ensure a quicker and more efficient response, safeguarding essential services and protecting patient care.

Stephen Gates, Principal Security SME, Horizon3.ai follows with this:

   “What all organizations must come to terms with is that their exploitable attack surface is no longer just their own – it now encompasses their suppliers’ attack surfaces as well. A successful attack against a smaller supplier, who is part of a larger supply chain, can in fact cause disruption to their upstream buyer community. Therefore, it is critical for organizations to not only continuously assess their own security postures, but the security postures of their suppliers as well.

   “The challenge here is how do you continuously assess yourself – and your suppliers? Traditionally, organizations relied on two assessment approaches: Checkbox assessment exercises or manual penetration tests. The first does not always guarantee better security and the latter is nearly impossible to perform in a continuous fashion.

   “However, there are autonomous assessment solutions and approaches readily available that can continuously (and preemptively) discover exploitable vulnerabilities and other weaknesses so organizations can resolve them before attackers discover them. It’s time for both public and private sector organizations to adopt these solutions and approaches before this sort of event becomes the norm

Third party attacks are the new problem of the moment. Which means that you need to ensure that you take into account the security practices of your suppliers and partners if you want to stay secure. Otherwise you get situations like this.

Does The Updated Adobe Photoshop Terms Of Use Gives Them The Right To Access All Your Work?

Posted in Commentary with tags on June 5, 2024 by itnerd

From the “this seems a bit suspect” comes this discovery by Sam Santala about the terms of use for Adobe Photoshop:

Of course this set off a bit of a tizzy on the Internet. And it sounds pretty bad on the surface. But there’s another view to this.

The way this is written, it sounds like it only refers to content that you upload to Adobe. Meaning that stuff that is local is stuff that they don’t have access to. That’s fine. Sort of. There’s also references to looking for things like CSAM and adult material. Now I am a computer nerd and not a lawyer, but that sounds like a cover your a$$ sort of thing to me. There’s also verbiage about using machine learning to improve their products. That’s something that I can see people having issues with as there are people out there who don’t want AI models trained using their stuff.

Adobe could really do itself a favour by clarifying this. Because until they do, people are going to have the impression that Adobe products are basically spyware when in reality they’re likely not. Thus I’ve pinged Adobe over this and if I actually get a response, I’ll post it here.

Equinix Makes Data Center Heat Available to Warm Nearby Neighbourhoods and Buildings

Posted in Commentary with tags on June 5, 2024 by itnerd

Equinix – the world’s digital infrastructure company – is calling for municipal planning agencies, energy utilities and heat network operators in Canada and around the world to join the Equinix Heat Export program to unlock the value of the residual heat generated in its International Business Exchange™ (IBX®) data centers.

As the world continues to digitize and there is an increasing demand for massive amounts of processing power to train AI data, current spec CPUs and GPUs are running constantly and are producing residual heat. Equinix is constantly innovating in order to operate in more sustainable manners and the newest innovation is by recovering residual heat from IBX data centres and exporting it to nearby buildings and communities.

TR5, one of Equinix’s IBX data centres in Toronto, is sharing heat throughout multiple residential buildings as well as a hotel, university, and local shopping centre in downtown Markham. TR5 is also servicing multiple buildings with domestic hot water all year long, through a partnership with Markham District Energy.

Equinix also plans to export heat in its newest IBX in Paris, PA10, and transfer it to the Plaine Saulnier urban development zone and the Olympic Aquatic Centre, which will host several events during the 2024 Summer Olympics.

Equinix is dedicated to expanding its Heat Export program in suitable locations, where data centers can serve as valuable heat sources for local communities.

Noah Nkonge, Design Delivery Principal, Equinix, details what data centre heat export is and how it works in the newest Equinix blog, which can be read here.

RansomHub May Have Possibly Evolved From The Knight Ransomware Project

Posted in Commentary with tags on June 5, 2024 by itnerd

With the new RansomHub ransom gang becoming very active in the last couple of months, security researchers have come to believe that the ransomware-as-a-service evolved from the Knight ransomware project. Analysts have found multiple similarities between the two – the list is included in a Bleeping Computer article.

Darren Williams, CEO and Founder, BlackFog:

“The spawning of new ransomware gangs from older ones is not unusual, and we have seen this regularly over the last several years as they continue to share and leverage multiple variants at the same time. The simple fact is the risks associated with cybercrime are so low that its significantly less dangerous than any other form of crime, especially when you consider that most originate in China and Russia where law enforcement is not cooperative. The focus of organizations needs to be on both defensive and preventative strategies to protect all digital assets from the inside. Focusing on data exfiltration to prevent a data breach and loss of data is a good first step that many organizations fail to recognize in their security posture.”

The risk that ransomware gangs face need to be increased. Otherwise there is zero deterrent in terms of them launching attacks on companies. Hopefully countries around the world get the message so that these gangs disappear out of fear of the consequences of their actions.

Apple Should Be Less Concerned With AI And More Concerned About The Quality Of Their Software

Posted in Commentary with tags on June 5, 2024 by itnerd

Now I will admit that this isn’t a new complaint from yours truly. In fact, if you want to see the last times I complained about Apple’s software quality, I’ll leave some links here, here and here. The fact is that over the last few years, Apple’s software quality has taken a nosedive. While that’s bad, it’s likely to get worse.

Enter AI.

Apple is perceived by many to be behind the curve when it comes to AI. So as a result, Apple appears to be going all in on AI if you believe the rumour mill. Resulting in the fact that we’ll likely see AI mentioned a whole lot next week when Apple does the World Wide Developer’s conference keynote. However I am here to say that this is the wrong move. If I were Apple, I’d be fixing the issues that exist in their operating systems and software to make them stable and reliable rather than trying to shove new AI features, or new features of any sort which will likely make things even less stable than they already are. Take this for example. Joanna Stern of the Wall Street Journal has this to say about Apple’s Screen Time feature:

In no universe should it take a reporter from a major newspaper to make a company fix a problem that has existed for years. But at least she got a response. Apple from personal experience has often taken bug reports from users and developers and you’ll never hear from them again. I think they do that because Apple’s focus is on the “new hotness” rather than ensuring that their products have stability and features work as advertised. And this is likely to continue with whatever software Apple announces next week. In other words, you should not expect watchOS, macOS, iOS or anything else that Apple announces next week to be more stable when they finally ship to consumers in the fall. Instead, you should be expecting the opposite. Though I suspect that you’ll hear Apple say that you got all these cool AI features for you to use. To me, that’s not a good trade and I would happily trade AI or any other new cool feature for stability.

I am pretty sure that there are many, many others who would echo that sentiment.

I’m Doing A Lunch And Learn On How Not To Get Scammed On June 14th In Downtown Toronto

Posted in Commentary with tags on June 5, 2024 by itnerd

This is some shameless self promotion. But it’s shameless self promotion on an important topic. I am doing another lunch and learn on how not to get scammed. This is an important topic these days as scams seem to be everywhere. Thus I’ve put together a roughly one hour lunch and learn on this topic so that it arms you with information to protect yourself. 

The location that I will be doing this lunch and learn from is a downtown Toronto restaurant called Harbord House. I highly recommend this restaurant as it has good food and a good beer selection, and I thank them for being willing to host this lunch and learn. The lunch and learn will be on June 14th at 1PM in the upstairs portion of the restaurant. If you can attend, please let me know.

And if you think that someone could benefit from this lunch and learn, feel free to forward this to them. 

Cisco and Splunk Launch Integrated Full-Stack Observability Experience for the Enterprise

Posted in Commentary with tags on June 5, 2024 by itnerd

Today at Cisco Live, the company unveiled the first of its innovative observability integrations with Splunk, a Cisco Company, as the organizations combine their industry-leading technologies to accelerate full-stack observability (FSO) for the entire enterprise. 

Building on the recent landmark acquisition, Cisco and Splunk are now launching an integrated full-stack observability experience for the enterprise, enabling unparalleled visibility and real-time insights to standardize observability in one solution. The new integrations and innovations are designed to help customers unlock unified visibility across any environment and any stack while harnessing powerful real-time analytics for faster, more accurate detection, investigation and response. 

Key Announcements include: 

  • Splunk Log Observer Connect for Cisco AppDynamics: Combines the power of Splunk Platform with Cisco AppDynamics Application Performance Monitoring (APM) to drive faster, in-context troubleshooting across on-premises and hybrid environments. 
  • Cisco AppDynamics integration with Splunk Enterprise / Splunk Cloud and Splunk ITSI: Reduce alert noise, improve troubleshooting and gain exec-level visibility via integration across Cisco AppDynamics, Splunk Enterprise / Splunk Cloud and Splunk ITSI. 
  • Cisco AppDynamics on Microsoft Azure: Expansion of cloud-hosted observability offerings now brings Cisco AppDynamics APM services to SaaS-hosted Microsoft Azure, fostering the support of multi-cloud strategies across new regions. 
  • Cisco AI Assistant for Cisco AppDynamics: Integrated into the AppDynamics Help Center, the new AI assistant empowers users with meaningful guidance and insights to make informed, intelligent decisions faster and more accurately than ever before. 
  • Advanced AI in Splunk IT Service Intelligence (ITSI): Leverages advanced AI and machine learning capabilities to help teams quickly and easily configure and implement dynamic, adaptive thresholds, manage and optimize configurations, and proactively surface insights into the health of ITSI knowledge objects, such as KPIs, services and entities. 

You can read through the full overview of Cisco Live announcements here.