Post by account_disabled on Sept 14, 2023 10:56:47 GMT
As IT companies increasingly apply artificial intelligence, machine learning, and so-called AI Ops technologies to network management, network data has come to play an important role in the success of companies. AI/ML technologies require more data to learn personal networks, derive insights, and make recommendations. Unfortunately, many companies run into problems when inputting network data into AI tools.
In other words, network departments need Phone Number List to modernize how they access network data before adopting AI technologies.
Enterprise Management Associates recently surveyed 250 IT professionals about their experience with AI/ML-driven network management solutions and found the answer: “AI-Driven Networks: Leveling up Network.” Management” was published as a report. The report identified data issues as the second difficult technical challenge encountered when applying AI/ML to network management. Network complexity was identified as the biggest technical challenge.
Additionally, 90% of responding companies experienced at least one serious issue related to network data when piloting AI/ML solutions.
A VP of IT at a $9 billion financial services company recently said, “AI Ops needs data to accelerate workflows. Without that data, AI Ops can’t be used. The first thing you need to do with any AI project is data. “To prepare, you need to see the data, understand it, and check the differences,” he advised.
Let’s take a look at the key data-related issues cited by IT experts in the survey.
data quality
The biggest problem cited by 46% of respondents is data quality. IT companies quickly discover garbage data that leads to garbage insights. They also struggle with errors, formatting issues, and non-standard data. This problem is likely to arise when IT companies input data from multiple siled tools into third-party AIOps solutions. The average IT company uses at least 4 to 15 tools for network management and monitoring. Each tool maintains its own database with different levels of quality. Problems arise when AIOps solutions attempt to correlate insights from these data sets.
security risk
39% of respondents struggle with security risks when sharing network data with AI/ML systems. Many companies create and sell AI-centric network solutions based on the cloud. If IT departments want to analyze network data, it must be sent to the cloud. Some industries, such as financial services, avoid sending network data to the cloud and instead store it on on-premises servers. Unfortunately, there are not many network companies that sell AI data lakes in an on-premise version. This is because cloud scalability is needed for analysis.
Some companies combine anonymized data from all customers for global analysis of the network. You can see all comprehensive trends regardless of variables such as regional boundaries or industries. However, some customers are uncomfortable with these advantages of AI/ML solutions. This is because we do not want anonymized data to be combined in this way.
network overhead
The third data-related challenge pointed out is network overhead. 36% of responding companies were concerned about the cost of moving massive data sets from off-premises to a cloud-based data lake. This data transfer often consumes excessive bandwidth. Some vendors alleviate this problem by processing data at the edge of the network in a local area and sending metadata to the AI cloud for analysis. Companies evaluating AI-centric network solutions should ask how AI/ML vendors are addressing this issue.
Data Segmentation
The last problem cited by 32% of respondents was the lack of data segmentation capabilities. This is because data is not collected at sufficient intervals to provide insights to AI solutions. This problem manifests itself in a variety of ways. Some SD-WAN vendors limit the rate at which network telemetry is collected. This is because telemetry traffic affects network performance.
Some monitoring tools limit the interval at which they collect network data using SNMP. This is because the higher the response rate, the more unstable the monitoring platform becomes. Some network switches and routers also affect business performance, limiting the frequency of flow record creation. Recently, some network vendors have begun to alleviate this problem by using switch silicon optimized for generating more granular data, but such hardware is prohibitively expensive.
Evaluate the network data you've already collected
Even if you have no plans to use AI/ML solutions for network management right now, it is a good idea to evaluate the current state of your company's network data. The network operations department also responded that data quality is usually the biggest issue, regardless of whether AI is considered.
For example, companies need to determine whether there are any blind spots in their networks that will become apparent with the use of AI. You should evaluate the quality of data collected and held by the tools you currently use. Is this collected data prone to errors? Data must also comply with standards. If the tool tags data with metadata, can it be analyzed by a third party? Data standardization ensures that data can be read by other systems. Data collection intervals should also be considered. A lot can happen between 5 and 10 minutes between NMP polling intervals.
In other words, network departments need Phone Number List to modernize how they access network data before adopting AI technologies.
Enterprise Management Associates recently surveyed 250 IT professionals about their experience with AI/ML-driven network management solutions and found the answer: “AI-Driven Networks: Leveling up Network.” Management” was published as a report. The report identified data issues as the second difficult technical challenge encountered when applying AI/ML to network management. Network complexity was identified as the biggest technical challenge.
Additionally, 90% of responding companies experienced at least one serious issue related to network data when piloting AI/ML solutions.
A VP of IT at a $9 billion financial services company recently said, “AI Ops needs data to accelerate workflows. Without that data, AI Ops can’t be used. The first thing you need to do with any AI project is data. “To prepare, you need to see the data, understand it, and check the differences,” he advised.
Let’s take a look at the key data-related issues cited by IT experts in the survey.
data quality
The biggest problem cited by 46% of respondents is data quality. IT companies quickly discover garbage data that leads to garbage insights. They also struggle with errors, formatting issues, and non-standard data. This problem is likely to arise when IT companies input data from multiple siled tools into third-party AIOps solutions. The average IT company uses at least 4 to 15 tools for network management and monitoring. Each tool maintains its own database with different levels of quality. Problems arise when AIOps solutions attempt to correlate insights from these data sets.
security risk
39% of respondents struggle with security risks when sharing network data with AI/ML systems. Many companies create and sell AI-centric network solutions based on the cloud. If IT departments want to analyze network data, it must be sent to the cloud. Some industries, such as financial services, avoid sending network data to the cloud and instead store it on on-premises servers. Unfortunately, there are not many network companies that sell AI data lakes in an on-premise version. This is because cloud scalability is needed for analysis.
Some companies combine anonymized data from all customers for global analysis of the network. You can see all comprehensive trends regardless of variables such as regional boundaries or industries. However, some customers are uncomfortable with these advantages of AI/ML solutions. This is because we do not want anonymized data to be combined in this way.
network overhead
The third data-related challenge pointed out is network overhead. 36% of responding companies were concerned about the cost of moving massive data sets from off-premises to a cloud-based data lake. This data transfer often consumes excessive bandwidth. Some vendors alleviate this problem by processing data at the edge of the network in a local area and sending metadata to the AI cloud for analysis. Companies evaluating AI-centric network solutions should ask how AI/ML vendors are addressing this issue.
Data Segmentation
The last problem cited by 32% of respondents was the lack of data segmentation capabilities. This is because data is not collected at sufficient intervals to provide insights to AI solutions. This problem manifests itself in a variety of ways. Some SD-WAN vendors limit the rate at which network telemetry is collected. This is because telemetry traffic affects network performance.
Some monitoring tools limit the interval at which they collect network data using SNMP. This is because the higher the response rate, the more unstable the monitoring platform becomes. Some network switches and routers also affect business performance, limiting the frequency of flow record creation. Recently, some network vendors have begun to alleviate this problem by using switch silicon optimized for generating more granular data, but such hardware is prohibitively expensive.
Evaluate the network data you've already collected
Even if you have no plans to use AI/ML solutions for network management right now, it is a good idea to evaluate the current state of your company's network data. The network operations department also responded that data quality is usually the biggest issue, regardless of whether AI is considered.
For example, companies need to determine whether there are any blind spots in their networks that will become apparent with the use of AI. You should evaluate the quality of data collected and held by the tools you currently use. Is this collected data prone to errors? Data must also comply with standards. If the tool tags data with metadata, can it be analyzed by a third party? Data standardization ensures that data can be read by other systems. Data collection intervals should also be considered. A lot can happen between 5 and 10 minutes between NMP polling intervals.