On September 26th, 2024, the Logic Apps Aviators community gathered for an inspiring and informative Logic Apps Community Day. The event was filled with insightful sessions led by a diverse group of experts from your Aviators community. Below is a recap of the presentations and key takeaways from each session, complete with links in case you missed it or want to watch again! You can also view the playlist of presentations, including the entire session, on YouTube. 

Igor Shvets: Unleash the Power of .NET C# in Azure Logic Apps 

Igor Shvets, CEO of Avant-Garde 365, kicked off the event by showcasing how C# can be leveraged in Azure Logic Apps through various practical scenarios and examples. Igor’s session highlighted the flexibility and power of integrating .NET C# with Logic Apps.

Prashant Kumar Singh: BizTalk to AIS Migration – Learnings from the field 

Next, Prashant Kumar Singh, a Solution Architect at HPE, shared his experiences and best practices for migrating from BizTalk to AIS. His insights were drawn from real-world projects, providing invaluable advice for those planning similar migrations.

Sebastian Meyer: Seamless SAP Integration with Azure Logic Apps 

Sebastian Meyer, a consultant at QUIBIQ and a Microsoft Azure MVP, demonstrated how Azure Logic Apps can streamline SAP integration with cloud services. His presentation showcased examples of enhanced efficiency and synchronization. 

Luis Rigueira: Workflow and Trigger expressions can help monitor your Logic Apps 

Luis Rigueira, an Integration Developer at DevScope, delved into the power of workflow and trigger expressions for monitoring Logic Apps. He explained how these tools can help stay notified about failures, providing details on run times and app names.

Mattias Lögdberg: Master PaaS networking for AIS 

Mattias Lögdberg from DevUP Solutions simplified the often complex world of PaaS networking options in Azure. His session clarified the different networking types and their appropriate use cases, making it easier for participants to navigate their choices.

Ankit Gupta: Mastering Azure Logic Apps: Renewing Lock Tokens in Peek-Lock Service Bus Triggers 

Ankit Gupta, a Cloud Solution Architect at TCS, illustrated how to build more resilient workflows by renewing lock tokens in peek-lock service bus triggers. His practical examples provided a clear path to seamless message processing. 

Cameron McKay: Transitioning from Pro to Low Code with Azure Logic Apps 

Cameron McKay, a Cloud Architect at MNP LLP, offered guidance for pro code developers transitioning to low code environments. His focus on best practices for maintainable and reusable workflows was particularly valuable. 

Angad Soni: Use Case: Predictive Maintenance for Manufacturing Equipment 

Angad Soni from Long View Systems demonstrated how to implement predictive maintenance systems using Logic Apps. His session on preventing downtime through scheduled maintenance tasks was both informative and practical. 

Sandro Pereira: Logic Apps: Everything you need to know about error handling 

Sandro Pereira, a well-known figure in the Azure community, shared his extensive knowledge on creating robust error handling patterns in Logic Apps. His presentation was packed with actionable advice based on his vast experience.  

Sabahat Faria: Integration between CE and FinOps using Logic Apps 

Sabahat Faria from Systems Limited focused on integrating CRM systems using Logic Apps. She provided a detailed demo on uploading data to environments like Dataverse and D365 for Finance and Operations. 

Paco de la Cruz: Extending your Azure Integration Solution with OpenAI 

Paco de la Cruz from Deloitte Engineering Australia explored how to extend Azure integration solutions using OpenAI. His session illustrated the innovative applications of AI in enhancing integration projects. 

Mick Badran: Using AI in Healthcare through Logic Apps 

Finally, Mick Badran, a veteran Azure MVP, presented a real-life use case of using AI and Logic Apps in healthcare. His session demonstrated the potential of AI-driven diagnostics and vital monitoring to support healthcare providers. 

Logic Apps Community Day 2024 was a remarkable event, filled with cutting-edge insights and practical knowledge. We extend our heartfelt thanks to all the speakers and participants who made this day a success. Stay tuned for more events and continue to post your innovations with Azure Logic Apps to #LogicAppsAviators! 

​Microsoft Tech Community – Latest Blogs –Read More 

I have noticed that we see quite a few endpoints that are triggering the Excessive NXDOMAIN DNS Queries anomaly analytics rule in Microsoft Sentinel. When I investigate these for tuning purposes, I see that the vast majority of these queries (in the in-addr.arpa domain) are for IP addresses owned by Microsoft. It appears that Microsoft have no interest in publishing reverse DNS entries, because I am unable to resolve them from any online DNS tools. The whois records do point to Microsoft, though.

What’s a good way to either stop this from happening, or eliminate the Microsoft IP address space from the query results?

​I have noticed that we see quite a few endpoints that are triggering the Excessive NXDOMAIN DNS Queries anomaly analytics rule in Microsoft Sentinel. When I investigate these for tuning purposes, I see that the vast majority of these queries (in the in-addr.arpa domain) are for IP addresses owned by Microsoft. It appears that Microsoft have no interest in publishing reverse DNS entries, because I am unable to resolve them from any online DNS tools. The whois records do point to Microsoft, though. What’s a good way to either stop this from happening, or eliminate the Microsoft IP address space from the query results?  Read More

​

With Copilot, scheduling a meeting from an email thread is quick and simple. By selecting the Schedule with Copilot option, Copilot reviews the email and generates a meeting invitation. It automatically fills in the meeting title and agenda, and attaches the email thread. The participants from the email thread are added as attendees, allowing you to easily review, modify details if needed, and send out the invitation.

#MicrosoftCopilot #Copilot #AI #Microsoft365 #MPVbuzz

​With Copilot, scheduling a meeting from an email thread is quick and simple. By selecting the Schedule with Copilot option, Copilot reviews the email and generates a meeting invitation. It automatically fills in the meeting title and agenda, and attaches the email thread. The participants from the email thread are added as attendees, allowing you to easily review, modify details if needed, and send out the invitation.
#MicrosoftCopilot #Copilot #AI #Microsoft365 #MPVbuzz  Read More

​

The DEVintersection Conference, held recently in Las Vegas, showcased the latest in Microsoft technologies.  Featuring keynotes from Microsoft leaders like Scott Hanselman, VP of Microsoft Development Community and Scott Hunter, VP Director of Product at Microsoft, on the Azure Developer Experience, along with insights from engineers, Microsoft MVPs & Regional Directors (RD), and industry experts, the event was a hub of innovation. DEVintersection founder and Microsoft MVP & RD Richard Campbell shares highlights from the event.  

Scott Hanselman, VP of Microsoft Development Community, & Safia Abdalla, Principal Software Engineer at Microsoft – photo by MVP Heather Downing

What inspired the creation of DEVintersection, and how has it evolved over the years? 

Richard: The whole idea of intersection was to bring together different groups of developers so that they could learn from each other. In the early days of DEVintersection (2012-2015), there was a focus on open web technologies. We co-located a conference called AngleBrackets with an open web content focus and then gave space to mix the two audiences. Over the years, we’ve also included data with SQLIntersection and, for a time, the M365 Conference. 

Today, DEVintersection co-locates with the NextGenAI Conference – the goal is to create an intersection between the different audiences and viewpoints to help foster a deeper understanding of both technology stacks. 

Any time you can bring more ideas together, you get powerful results! 

Do you have reflections on DEVIntersection’s growth over the years?  

Richard: DEVintersection has evolved  as the community has evolved – the demands on developers continues to grow, and we’re there to help. As new technologies emerge, we intersect them, bringing new experts and viewpoints to the DEVintersection audience. 

How does Microsoft’s participation enhance the learning and networking opportunities for attendees at DEVintersection? Were there unique insights or sessions from Microsoft this year? 

Richard: We see Microsoft as partners in telling the development story to the audience. We like to bring Microsoft engineers to the event so that they can talk about the how and why of what they’ve built. Then our industry experts add to the conversation with their experiences of deploying those technologies to customers. The intersection between industry and Microsoft is where the value lies – you can see the future from there!

MVP Veronika Kolesnikova speaking at DEVintersection – Photo by MVP Heather Downing

Can you share any success stories from past attendees who have implemented what they learned at DEVintersection? 

Richard: During the years of SQLintersection, I would host a panel discussion at the end of each of the conferences that we would record as a RunAs Radio podcast – once or twice a year, from 2014 to 2022. If you listen to those shows (and I know the SQL team does!) you’ll hear the shift of how people view  SQL Server . Ofen it’s the same attendees, year-to-year, asking progressively more complex questions. 

There too you see the gradual embrace of the cloud as your data store – initial skepticism, followed by exploration, ultimately being embraced. 

It’s the most rewarding part of creating a conference. 

How do you select the speakers and topics for each conference and ensure a diverse and inclusive lineup of speakers and sessions? 

Richard: While we do have a Call for Proposals (CFP), it’s not our primary focus for recruiting speakers, because of diversity and inclusion. The CFP process does not appeal to the largest and most diverse group of speakers. We recruit directly – being part of the community, involved with a huge array of different types of events, to directly connect with the best and brightest in the industry. And then we ask them to be part of our show. It’s the best way to get a broad spectrum of people involved in a conference. 

Curious for more? Check out videos of keynote sessions, interviews, and highlights from past DEVintersection and next GenAI Conferences.  

​Microsoft Tech Community – Latest Blogs –Read More 

Hello,

My Microsoft SQL Server error log contains very often the following line “RemoveStaleDbEntries: Cleanup of stale DB entries skipped because master db is not memory optimized.

Parallel redo is started for database ‘database_y’ with worker pool size [4].

Parallel redo is shutdown for database ‘database_y’ with worker pool size [4].

RBPEX::NotifyFileShutdown: Called for database ID: [5], file Id [0]

Starting up database ‘database_y’. ” where database_y has database ID: [5]

DBCC CHECKDB (‘database_y’)

CHECKDB found 0 allocation errors and 0 consistency errors in database ‘database_y’

Do you have any idea how to continue the investigation?

Thank you.

Best regards,

Matei

​Hello,My Microsoft SQL Server error log contains very often the following line “RemoveStaleDbEntries: Cleanup of stale DB entries skipped because master db is not memory optimized.Parallel redo is started for database ‘database_y’ with worker pool size [4].Parallel redo is shutdown for database ‘database_y’ with worker pool size [4].RBPEX::NotifyFileShutdown: Called for database ID: [5], file Id [0]Starting up database ‘database_y’. ” where database_y has database ID: [5]   DBCC CHECKDB (‘database_y’)CHECKDB found 0 allocation errors and 0 consistency errors in database ‘database_y’Do you have any idea how to continue the investigation?Thank you.Best regards,Matei  Read More

​

By Shantanu Deepak Patankar, Software Engineer Intern, and Hugo Affaticati, Technical Program Manager 2

Inefficient inference optimization can lead to skyrocketing costs for customers, making it crucial to establish clear performance benchmarking numbers. This blog sets the standard for expected performance, helping customers make informed decisions that maximize efficiency and minimize expenses with the new Azure ND H200 v5-series.

We evaluated the inference performance of the new Azure ND H200 v5-series for Small Language Models (SLMs) and Large Language Models (LLMs). The ND H200 v5-series, powered by eight NVIDIA H200 Tensor Core GPUs, offers a 76% increase in memory bandwidth over the NVIDIA H100 Tensor Core GPU of the ND H100 v5-series. We compared three models: Phi 3 (128k parameters), Mistral v0.1 (7B parameters), and Llama 3.1 (8B, 70B, and 405B parameters) to set performance standards and empower Azure customers to optimize their workloads for time or resources.

Model Architecture

Achieving optimal performance requires a clear understanding of where time is spent during the inference workload, enabling effective optimization. The first critical step is to carefully examine the parameters that directly impact performance. For the models discussed, and more broadly, these key parameters include input sequence length, output sequence length, batch size, and tensor parallelism. In this article, we measured the impact of these variables using two essential metrics: throughput and first token latency.

The inference process can be categorized into three primary components: pure computation phases (e.g., local GEMMs), pure communication phases (e.g., all-reduce), and attention phases. Analyzing the Llama3 8B model on the new ND H200 v5 virtual machine revealed that computation consistently accounts for at least 50% and up to 85% of total inference time. Communication time ranges from 10% to 25%, scaling as the number of GPUs increases from 2 to 8. In contrast, attention mechanisms consistently represent less than 10% of the total time spent, as shown in Table 1. This article aims to guide customers in striking the right balance between computation and communication when selecting their AI inference architecture, based on whether time efficiency or cost-effectiveness is their primary goal.

Tensor Parallelism
Computation (% of time spent)
Communication (% of time spent)
Attention (% of time spent)

1 GPU
83.3
0
9.2

2 GPUs
70.7
10.8
7.4

4 GPUs
56.7
24.7
6.1

8 GPUs
57.2
25.1
8.2

Table 1: Breakdown of time spent per mechanism for LLAMA 3 8B inference on the ND H200 v5 virtual machine, with an input sequence length of 1024, output sequence length of 128, and batch size of 32.

Resource optimization

Since most of the inference time is spent on computation, the GPU computational speed has a tremendous impact on the overall performance. Understanding the memory requirements ensures better GPU usage. The two main factors influencing GPU memory consumption are the model weights and the key-value cache.

Model Weights: the memory occupied by the model weights depends on the number of parameters and the quantization of the model. The memory required can be calculated using the formula: 

Memory used (in GB) = number of parameters (in billions) × precision (in bits / 8)

For example, the model weights of a LLAMA 3 model using 8B parameters and FP8 precision, would require 8 GB of memory (8B parameters x 8 / 8 = 8 GB)

Key-Value Cache: since the attention score of each token only depends on the preceding tokens, the model stores the key and value matrices in the cache to avoid recalculating attention values for every token in the sequence, accounted for by the factor 2 in the equation below.

Size of KV cache (in B) = batch size * sequence length * 2 * number of layers * (number of heads * dimension of head) * precision (in bits / 8)

For example, the key-value cache of a LLAMA 3 model using 8B parameters, FP8 precision, input size 1024, and output length 128 would require 0.5 GB of memory for a batch size of 1 (1 x (1024+128) sequence length x 2 x 32 layers x 4096 x 8 / 8 = 0.5 GB)

By using these two quantities, customers can accurately estimate the maximum batch size that the virtual machines can accommodate for their model, thereby optimizing resource utilization. The available GPU memory is calculated by subtracting the weight memory from the total GPU memory when the system is idle. The maximum batch size is then determined by dividing the available memory by the size of the KV cache required for a batch size of one. Table 2 provides several examples of these theoretical batch sizes. This approach not only simplifies the process but also helps customers avoid the trial-and-error method, which can lead to higher GPU consumption and increased costs.

Model
ND H200 v5 memory per GPU (in GB)
Number of parameters(in billions)
Weight Memory(in GB)
Available memory(in GB)
KV Cache size (in GB)
Max Batch size

LLAMA 3
140
8
16
124
0.60
206

Mistral
140
7
14
126
0.60
210

Phi-3 medium
140
14
28
115.8
0.94
123

Table 2: Theoretical maximum batch size for inference with various languageinference models (LLAMA 3 8B, Mistral, Phi-3 medium) on the ND H200 v5 virtual machine with sequence length 1152 and FP8.

Very similar results have been obtained empirically to confirm the theoretical limits. Figure 1 below highlights the maximum batch size to maximize the usage of one NVIDIA H200 Tensor Core GPU, then combined to up to the eight other GPUs of the latest ND H200 v5 virtual machine, with the corresponding throughput. By optimizing the batch size, customers can extract extra performance from each GPU, fully utilizing available resources. This ensures that every virtual machine operates at its peak capacity, maximizing performance while minimizing cost.

Figure 1: Experimental maximum batch size as a function of tensor parallelism (TP) for inference with LLAMA 3 8B on the ND H200 v5 virtual machine with sequence total length 1152.

Time optimization

For some specific workloads, time is more of the essence. While increasing the batch size can enhance throughput and maximize resource utilization, it also leads to higher latency. By measuring both latency and throughput of the inference workload, the optimal balance can be determined. For instance, when running models like Llama 3 and Mistral on a single GPU of the latest ND H200 v5 virtual machine, a batch size of 32 delivers the highest throughput-to-latency ratio, as shown in Figure 2. The optimum batch size is specific to the customer’s workload, as highlighted by the Phi-3 model, which achieves its highest ratio at a batch size of 64 with a single GPU. When scaling to two GPUs, the optimal batch size increases to 64, as illustrated in Figure 3. Although this approach may not fully utilize the available memory, it achieves the lowest possible latency for inference, making it ideal for time-sensitive applications.

Figure 2: Experimental optimal throughput to latency balance as a function of batch for inference with LLAMA 3, Phi-3 and Mistral on a single GPU of the ND H200 v5 virtual machine with sequence total length 1152, FP8, and TP 1.

Figure 3: Experimental optimal throughput to latency balance as a function of batch for inference with LLAMA 3, Phi-3 and Mistral on two GPUs of the ND H200 v5 virtual machine with sequence total length 1152, FP8, and TP 2.

​Microsoft Tech Community – Latest Blogs –Read More 

Join us in Reston, VA, on October 15th, 16th, or 17th for our Microsoft 365 Copilot GCC Readiness Days!  
 
This exclusive in-person event is your chance to learn how AI and Microsoft 365 Copilot GCC can address the unique challenges of public sector missions. Tailored for IT professionals, administrators, and decision-makers in government agencies, these readiness days offer practical, actionable insights to help you drive secure productivity, efficiency, and innovation in your organization. 

Reserve Your Spot Today! 

Spaces are limited for these in-person readiness days to ensure personalized, in-depth discussions. Register now to secure your spot and be among the first to explore how Microsoft 365 Copilot GCC can empower your teams to work more effectively. 

Register for October 17th 
Register for October 16th 
Register for October 15th 

Why You Should Attend 

Microsoft 365 Copilot GCC is set to revolutionize government workspaces, bringing AI-powered productivity into secure, compliant environments. At this event, you will uncover how Copilot can automate routine tasks, enhance collaboration, and help your team focus on higher-priority activities—all while meeting the stringent regulations of government operations. 

What You’ll Gain from the Event: 

In-Depth, Actionable Learning: Sessions will provide detailed, step-by-step guides on how to implement Copilot GCC within your agency’s existing workflows. 
Transparency and Trust: Gain insights from candid discussions about the strengths and limitations of AI in government settings, addressing your concerns around compliance, security, and data privacy. 
Real-World Solutions: Learn directly from AI thought leaders and engineers on how to navigate integration challenges and leverage responsible AI practices. 

What to Expect: 

Prepare to Become an AI-Powered Public Sector Organization: Learn how to ready your workforce for a seamless transition to Microsoft 365 Copilot GCC, with practical tips tailored to government environments. 
Responsible AI: Discover how Microsoft partners with government agencies to implement AI responsibly, ensuring compliance, security, and ethical considerations. 
Gov AI Integration Challenges and Opportunities: Engage in transparent discussions with product experts and engineers to explore best practices for deploying Copilot GCC in your agency’s environment. 

Event Agenda: 

09:00 AM – 09:15 AM: Keynote 
09:15 AM – 10:00 AM: Roadmap 
10:00 AM – 10:50 AM: How Copilot Works 
11:00 AM – 11:50 AM: How to Get Ready: Technology 
12:00 PM – 01:00 PM: Lunch 
01:00 PM – 01:50 PM: Microsoft 365 Business Chat Conversation 
02:00 PM – 02:50 PM: How to Get Ready: People 
03:00 PM – 03:50 PM: Art of the Possible Demos 

Note: The agenda is identical for each day, so select the date that best fits your schedule. 

Is Lunch Provided? 

Yes! Lunch is provided each day, thanks to our spectacular sponsoring partners: 

October 15: Planet Technologies 
October 16: Dell Technologies | Federal 
October 17: Carahsoft 

Be sure to learn more about our sponsoring partners’ Copilot offerings during the Partner Lunch & Learn session! 

Venue Details: 

Where: The Microsoft Garage, Reston – DC, 11955 Freedom Drive, Reston, VA 20190 

Engage and Share! 

Have questions or want to share your thoughts on Microsoft 365 Copilot in GCC? Join the conversation in the comments below and connect with your peers! 

​Microsoft Tech Community – Latest Blogs –Read More