How UltraFast Uses Reinforcement Learning to Tackle Tough Network Conditions

How UltraFast Uses Reinforcement Learning to Tackle Tough Network Conditions

Latency and packet loss over wide area networks, the Internet and RF-based network devices (e.g., satellite, cellular, packet radio) has long been a barrier to large scale data transfers. TCP/IP’s windowing algorithm is infamous for reacting very poorly to packet loss, which reduces the amount of data TCP is willing to send per transaction, making it extremely slow yet reliable. Today, large amounts of data increasingly needs to be transferred from where it gets created to elsewhere across available networks to where it is consumed and used. Sometimes this data is purely for disaster recovery and backup, other times its for important analytics and other business processes. Edge computing promises to address some of these issues by moving workloads closer to the point of data creation, but even then, data must often be transferred to centralized locations (data centers, public clouds, SaaS services) to make use of the insights gained across many edge sites.

Buurst Fuusion’s UltraFast® Machine Learning Approach

Over the years, many different types of algorithms have been devised to try and address this network throughput optimization problem. Buurst’s Fuusion product includes a feature called UltraFast®, which overcomes the challenges posed by TCP over highly latent or lossy networks in a unique way. As we will see in this post, UltraFast utilizes a type of AI/ML technology to learn, adapt and optimize data transfers over troublesome network conditions.

The UltraFast Gambler Agent

UltraFast uses a machine learning process that uses a set of “gamblers,” or data transmission experiments, that place different “bets” on what the ideal transmission rate will be. There is no model of the network available ahead of the agent running its own experiments to learn about the network.

The main goals are to:

  1. Maximize network throughput by sending as much data as possible
  2. Avoid creating packet loss due to sending data too quickly
  3. Detect when external factors, such as other network participants, changing IP routes, and other dynamic conditions are causing congestion or interfering with packet throughput and use this information to place improved bets.

The Agent creates a set of “Gambler” processes, each running in an independent thread. Gamblers are given a “Data Transmission Bet” to place, said bet being its data transmission “rate”; i.e., a bet is the time delay between sending each packet. The data is sent to a connection at the distant end of the network, and several types of responses may occur: 1. an ACK indicating good data receipt, 2. a NAK indicating bad data receipt, or 3. no response at all, indicating a lost packet (timeout). Each Gambler process sends several data packets and records the overall success rate, i.e., how many packets were sent, how many succeeded, and how many failed. Upon completion of the Gamblers’ processing, each Gambler is assigned an overall score. The more acknowledged and successful data packets sent, the higher the score. The more NAKs or timeouts (packet losses) present, the lower the score. As we will soon see in more detail, the Agent then uses these Gambler scores to reward successful gamblers, which are allowed to “breed” and multiply during the next generation or experiment cycle. Less successful or failed Gamblers are pruned and eliminated. This process is similar to natural selection, where the strong and successful survive, and the weak and unsuccessful do not propagate.

UltraFast Reinforcement Learning Process

The UltraFast Reinforcement Learning Process The chart below depicts the UltraFast learning cycle and each step in the process.

The UltraFast learning loop runs repeatedly, processing these steps:

  1. A Monte Carlo derived genetic algorithm generates random strategies for the initial set of gamblers, then subsequently breeds new gamblers based upon last cycle’s winners’ results.
  2. A new generation of dozens of gamblers is created at the start of each cycle, each with its own rate of sending data packets.
  3. Gamblers send their data packets, measuring ACKs, NAKs, and lost packets.
  4. Each gambler’s win/loss rate is scored – more packets sent equals a higher score, lost packets or data transmission errors (NAKs) penalize the score.
  5. Each gambler’s loss-rate is compared with the current loss-zero (separately established with regularly timed packets).
  6. Winning gamblers showing the best results are rewarded by being bred, resulting in similar successful gamblers for the next cycle. The agent prunes the losers and feeds the learned results forward into the genetic algorithm. In addition to the ‘successful’ newly created gamblers, new random variants are added to further explore the newly defined boundaries, enabling the system to adapt to changing network conditions.

The above 6-step process runs continually, optimizing data throughput while minimizing packet loss and congestion, and adapting to the constantly changing and evolving complex network environment. Reinforcement learning enables UltraFast to adapt to each unique network topology and navigate its evolving traffic and routing conditions.

UltraFast Speed Test

UltraFast includes a speed test feature, which sends “iperf” data through the UltraFast optimizer as both a download stream and then an upload test. This is analogous to your typical Internet or broadband speed test, except it uses UltraFast technology to compare the throughput results vs. plain TCP/IP. In the following screenshot, we see the TCP results displayed in red (mostly hidden behind the blue UltraFast chart). The link being tested is on AWS between the Ohio region in the USA and the Capetown region in South Africa. The latency averages around 250 milliseconds round trip time, with little to no packet loss.

TCP/IP averages just 144 Mbps over this moderate-latency 1 Gbps link. We can see during the initial Download Speed test, the blue (cyan) UltraFast chart slowly increase its throughput over time, as the gamblers run and the reinforcement learning algorithm actually learns the particular characteristics of this network. Once UltraFast learns the network, it eventually is able to peg the network at near 1 Gbps at times. Then the upload test starts. Since UltraFast has already learned this network, it immediately optimizes the throughput, averaging 822 Mbps vs. TCP’s 144 Mbps.

As network conditions vary over time, UltraFast’s intelligent learning algorithm continues to observe, adapt and learn in order to continually optimize network throughput. This is very important for long-running bulk data transfer jobs in the terabytes or more size. As these long-running data transfer jobs occupy large amounts of network bandwidth over time, they are much more likely to experience competing traffic at different times of day; e.g., backup jobs running overnight, user downloads during daytime hours, and many other variables, including network routes changing the underlying network characteristics over time.

Summary

Optimizing data throughput over challenging network conditions is an age old problem – one that now has a new type of solution that uses reinforcement learning to intelligently optimize and constantly learn and adapt to changing network conditions. To learn more about the UltraFast feature of Fuusion and how it addresses challenging, high-latency and lossy network conditions, please visit the Fuusion page for more information. For more detailed insights into UltraFast, its machine learning technology and overall architecture, you can download the UltraFast technical white paper.

A Fresh Take on Edge and Cloud Services

A Fresh Take on Edge and Cloud Services

The ever-changing world of Data Storage and Management requires a never-ending quest to stay ahead of the curve. As any surfer will tell you, the ideal spot is ahead of the wave and somewhere near the crest. If you imagine the wave as your clients’ needs, it is this sweet spot, anticipating the needs of your clients and positioned to deliver – that will allow your organization to go on an epic run. That’s how we feel at Buurst, poised for an epic run.

Why do we feel this way? Because we’ve been watching the trends of the cloud and data marketplace, and we see the next wave. We knew that data storage was shifting to the cloud and see this change is accelerated by factors such as Covid-19. The work from home phenomenon, experts agree, is far from a passing trend. 90% of HR managers say they plan to allow employees to work from home more often, even after the pandemic. Work from home has increased at least 300% during the pandemic. This translates to more data generated overall, and much of it unstructured. We will not entirely know how this skews the already gigantic data predictions for 2025 and beyond for a few years yet, but we know it will.

Flexible data storage is one of the keys to the upcoming wave, one which Buurst’s SoftNAS product was explicitly designed to handle. Our company was founded on the principle of “disrupting the traditional storage market by not selling you more storage.” This approach allowed us to create an agile, cost-effective data transfer management and performance management tool that will help our customers do more than even they thought possible with their data. We do not charge by the petabyte, the terabyte, or the gigabyte. Our solution is entirely based on performance, no matter how much data flows through.

Unlike surfing, in the data world, sometimes you can ride more than one wave at a time, and we think we can do just that. Because we know that data usage and generation is not only growing but ever more often shifting to non-central locations, or the edge as industry experts have coined it. Buurst is poised to deliver a solution for managing/shifting/processing, and transferring this data. This product, Fuusion, allows our clients to seamlessly shift data from the edge to the cloud or any other central repository. Not only this, but it can perform data processing at the edge to ensure the data hits your repository in an instantly usable fashion.

In the linked article, Vic Mahadevan discusses both of these solutions and why Buurst feels poised to be the next big thing in data storage. Enjoy the read!

Hitachi Vantara and Buurst Data Transfer Management

Hitachi Vantara and Buurst Data Transfer Management

Hitachi Vantara and Buurst presented a webinar focused on the newly available joint solution of Buurst Fuusion availability with the Hitachi Virtual Storage as a Service platform.

Webinar Details

Customers are pushing data from the datacenter to the cloud, but they will task Colocation facilities as part of the Data Journey to the Cloud. It is important to have a Data Storage Platform that is built for Hybrid Cloud operations with the ability to scale from Datacenter to Edge.

Hitachi Virtual Storage as a Service provides a rich Hybrid Cloud experience for customers from Mid-Market to Enterprise. One of the biggest hurdles to achieving Hybrid Cloud Storage and Data Operations is the ability to ensure the data transfer and data operations to move and shift data between the Datacenter, to the Colocation and to the Public cloud.

Ensuring the data transfer is accomplished precisely, accurately and at the highest line speeds consistently requires strategic technology that works over all type of desperate network connectivity and latency. Failure to deliver the data is not an option, and Buurst Fuusion in strategic partnership with Hitachi Virtual Storage as a Service delivers the best in breed Data Transfer Management and Data operations in the Industry.

Combined Hitachi Virtual Storage as a Service with Buurst Fuusion ensures the data delivery in Hybrid Cloud deployments will be bullet proof.

For more information contact Buurst at sales@buurst.com.

Lessons Learned in Remote Working Before and During the Pandemic

Lessons Learned in Remote Working Before and During the Pandemic

As many parts of the world start to emerge from the pandemic, many of us working from home, it seemed a good time to share with the world Buurst’s experiences. As for Buurst, one could say that not much has changed for many of us. You see, Buurst has always employed a remote workplace strategy, finding ways to collaborate effectively with co-workers across the globe. Still, the choice to stay at home and work is not the same as being forced to by external complications. As such, we thought we’d share some insights from a company that was working from home “before it was cool.”

First, let us look at some of the reasons that Buurst has employed the remote office strategy in the first place. It all begins with our founder, Rick Braddy. According to Rick, it all began with his employment at Citrix, a company that made remote work software allowing work from home options a selling point for their company. As a result, he knew first-hand that remote work was not only feasible but could be potentially even more productive.

“From the outset, I wanted to create a ‘virtual workplace’ type of experience, both for our employees and me. Countless hours of people’s lives are wasted in traffic, not to mention the frustrations it causes everyone (and the pollution). I also observed a lot of wasted time and productivity from people’ visiting’ in the office a lot of their time. My theory was that between the increased productivity and increased work/life balance that working from home offers employees, is that it would result in lower turnover, happier people and better results.”

His experience at Buurst, operating a 100% remote workforce for over eight years, has only reinforced this idea.

Rick Braddy also mentioned that remote work allows folks to improve quality of life simply by living where they choose rather than in geographic proximity to the workplace – places where rent or housing prices might be higher, along with other expenses. I can attest to this, as I live in a small town in Alberta, Canada, 50 miles outside of Calgary, where rent and housing (and groceries) can be considerably more expensive. Not only this, but I have been able to work remotely to extend vacations, allowing me and my family to spend a bit more time in desirable locations, such as family trips to Vegas and Port Alberni on Vancouver Island. As long as I have wi-fi access, I can sneak in a couple of hours of work nearly anywhere. As a single father, this flexibility has proven indispensable.

But how do you establish a remote resource strategy for your company? Where do you find quality resources? For Buurst, the answer was in a freelancing site called ODesk, which, after a merger with rival site Elance, became known as Upwork. (Interestingly, I was a member of both sites when first starting my freelancing career. I found each to have strengths that, for the most part, carried over in the merger. Concentration on a single platform soon shored up weaknesses.) As Rick tells it, “Early on, we discovered Upwork, then known as ODesk. We began using part-time contractors early on to gain access to the skills we needed from a worldwide contractor talent pool. As Buurst grew, we continued using Upwork as a means of finding full-time employees who continue to lead the company from wherever they live.” Without this active global network, Buurst’s remote office strategy would not have been possible. Upwork ensured that Buurst (then SoftNAS) gained a trusted employment track record that attracted high-end talent while at the same time providing freelancers with the opportunities that best fit their needs.

Buurst uses several tools to keep in contact during communication and collaboration, notably several staples such as Microsoft Teams. For the development team and technical writing (as the technical writer needs to connect closely and often with the development team), Assembla has been the tool of choice for several years. Github integration and other features help keep the dev team on track, of course. Still, the ticketing system is also invaluable in tracking changes to the product as they happen or, in the best cases, well before. Asana manages tracking projects such as web design, white papers, and other activities for marketing and sales. But it was often cumbersome for those who bridged teams to keep track of notifications from a separate project management solution. Microsoft Teams proved very helpful, allowing us to link Office 365 functionality, replace Skype as a meeting tool internally, and serve as a repository hub via SharePoint for all our files and documentation, organized according to teams. The file repository and team organization proved immensely helpful in adopting Microsoft Teams for everything not directly related to development. Files are more easily found within team folders and shared readily across departments and teams, and past meetings are shareable.

More importantly, meetings are simple to organize and schedule, and team members can find one another at any moment for face-to-face calls without the need to share contacts or search a global database of users. As Vic Mahadevan, CEO of Buurst, tells us,

“I previously always enjoyed meeting the team and the camaraderie that develops face to face in the office. However, I have been very pleasantly surprised how technologies like Teams and Zoom have allowed us to stay connected to our employees, customers, and partners worldwide. We have been able to delight our customers and partners with our efficiency and productivity without skipping a beat during this pandemic.”

To create this sense of community and camaraderie, Buurst has leveraged Microsoft Teams to set up a company-wide meeting every week to highlight all things Buurst and contributions from team members.

Buurst also has leveraged meeting capabilities for fun and games to break the daily grind. A couple of times a month, an open meeting is set up to play a game with fellow Buursters. The last activity chosen was a game of Bingo, complete with prizes. Bingo resulted in some good-natured ribbing of the Bingo caller for not calling “the right numbers” and jokes about the Finance team rigging the game, with the Dev team claiming they would tweak their algorithm for better results the next time. In other words, a typical office get-together, with the same banter, the same relationships built over time and conversation, with the help of technology and shared goals.

But while it might seem that the remote office strategy is pandemic proof, there are a few challenges that have arisen. The most notable of these is that everyone else is staying home too. We, too, had to shift schedules on occasion simply because our children are home too. Noise levels during the day can determine meeting schedules and work schedules. Stephen Spector, VP Marketing, notes,

“The biggest challenge for remote work is finding a location in my house to work while still having to set up space for my two children who are both attending junior high primarily virtual this year. With my ‘office’ in the family room, having my kids at home has created plenty of daily chaos. I find that working early in the morning and taking time late at night to finish up work offers a few extra hours a day with minimum disruption.”

You might think that we parents who work from home all year long have practiced for this covid-related chaos every summer when school gets out. But unlike summer break, Covid has meant we can’t simply send children outside or to a friend’s place to gain a few more hours of blessed silence.

For those considering remote office employment, remember that it requires a different mindset. It requires discipline to ensure that the distractions inherent in the home environment do not affect your work. As Rick tells us, “It takes a shift in the mindset from ‘show up to work’ to ‘deliver excellent work results’. ” as a non-office-based company holding employees accountable for results and develop a level of mutual trust. Over the years, we have seen that some people just aren’t cut out for remote work as they lack the work ethic and self-discipline required. But for those who can separate work from where it takes place, remote working is a great way to go.” This shift may become more and more necessary until Covid-19, and its variants have run their course. Embrace the change and the challenge, and the experience will serve you well long after. Who knows, your office may be the next to be “pandemic proof.”

VDI Survey Results Highlight Value-Add of SoftNAS for User Experience

VDI Survey Results Highlight Value-Add of SoftNAS for User Experience

With more and more people working from home during the Covid pandemic, the importance of workplace resources, particularly those that allow us to communicate and collaborate, has become increasingly apparent. For example, an outage in Office 365 recently caused numerous issues with planned meetings over Microsoft Teams, email issues, and other problems. As a remote company, these types of disruptions are a significant disruption to our daily routines.  

Even if we could revert to another tool for virtual meetings or create documents on Google docs instead of Word, the problem was not that we couldnt get things done, but that it didnt feel like we were getting things done. It wasnt familiar; the day felt offThis feeling is also true for poorly configured VDI deployments. Ithe user experience is slow or uncomfortable, productivity will fall significantly.  

Considering that many organizations have invested heavily in tools for remote workers, such as Virtual Desktop Infrastructure (VDI), it is vital that rollouts go smoothly and users adopt these changes readily. These investments should not be considered a quick temporary fix. While Covid may not be around next year, the changes it has brought to worklife might well be permanent.  

%

Increase in complete workdays at home

%

Would like mix of post-pandemic and remote/in-office work

%

HR managers who plan to allow for mixed work

For example, Forbes recently posted the number of complete workdays performed at home has increased from 5% to 22% this past year. 66% of Americans agree that they would like a mix of remote and in-office work even after the pandemic. And 90% of HR managers say they plan to allow for more of a mix of remote and office work even after the pandemicIn other words, your VDI solution is not a stop-gap measureit is a permanent shift in the workplace dynamic. 

So how do we ensure that your VDI implementation is ready to tackle not only the pandemic but the post-pandemic paradigm shift?  

User Experience is Key 

Users need to feel that they are NOT working remotely, and their desktops and applications are running directly on their local PC or desktop and not on a remote server. This issue is not just about performance, though this does play a key role. Previously, IT organizations believed that persistent end-to-end performance was only possible via local server resources. In other words, significant investments in centralized servers were a must for VDI.  Or for larger organizations, regional infrastructure hubs to avoid issues such as latency. These infrastructure investments would reach end of life after a few years, and another round of investments would begin.  

VDI in the cloud is infinitely more scalable with a similar performance profile to an in-house solution running on public clouds such as AWS and Azure. However, there is a public perception, due to implementation growing pains, poor planning, or not enough investment, that cloud-based VDI is inadequate, buggy, or slow. A recent survey by ControlUp of 450 VDI administrators identified the critical challenges faced in deploying cloud VDI for various sized companies.   

Slow Login 

The top issue reported was by far excessive login times, cited by 77% of respondents. The perception this generates is that IT organizations are foisting outdated and under-performing hardware on them. But there can be numerous reasons for slow logins, including logon scripts, profiles, and group policies; however, the most significant is the accessibility of userprofiles and the files and applications behind them. To provide a seamless user experience described earlier, in which the user cannot distinguish between VDI and a work PC, a fair bit of data needs to be made readily accessible. SoftNAS provides the solution by storing critical user profile data on high-performance block storage in either AWS or Azure. In addition, this storage can be placed in the region closest to your users, avoiding latency issues. With enterprise-class data-centers storing user profiles, performance may exceed anything provided by a home office solution, especially if users are in rural or very remote locations. 

Application Performance 

The second most frequently identified issue was application performance. Applications within a VDI solution require constant and persistent access to underlying data stores. Performance depends strongly on the availability and persistence of this data. Again, the cloud storage provided to the most frequently accessed applications will largely determine the user experience and can be managed in the same manner as mentioned above – ensure the most frequently accessed data is local and high-performing. For data not commonly needed, lower-cost storage is leveraged.  

SoftNAS can also reduce storage costs by providing deduplication and compression for the data in question. Bear in mind that this can impact compute (CPU) resources required to process the data. A general rule of thumb is to plan for an additional 50% in CPU resources if compression or deduplication is used.  

Home Networking 

Another top reported issue with supporting remote users (for VDI or any remote work scenario) is connectivity itself – i.e., issues with the network at the remote user location. Even with high-quality internet service, there might be numerous users contending for the same bandwidth, particularly during this pandemic. The lack of bandwidth might include children‘s remote learning (virtual video), streaming services such as Netflix, and other such contention. If in a rural location, bandwidth or latency might be an issue in itself. In either case, the best solution is to place required storage resources as closely as possible to the user to improve response time and access. Cloud services such as AWS and Azure are very flexible in this regard. SoftNAS is equally flexible in that VMs are set up as close as necessary to the target users, and the storage used does not affect the licensing cost.  

Conclusion 

ControlUp reported additional issues, and these too can be addressed to some degree by SoftNAS. Still, typically by the exact solutions reported above – slow sessions are solved by ensuring the resources required are made more accessible and served by higher-performance storage as well, for instance. SoftNAS can help implement a solution that best fits the requirements and issues potentially identified and work with you to ensure the solution comes at the best cloud service price point possible. Remote work is here to stay, long after the pandemic. A properly configured VDI solution, underpinned by SoftNAS storage, provides the user experience needed to ready your organization for the transition.  

Fuusion Use Case: Smart Data Migration to the Cloud

Fuusion Use Case: Smart Data Migration to the Cloud

Smart Data Migration to the Cloud.

Any large and distributed business, whether it be a brick and mortar retail venture, car dealerships, energy production, mining, or manufacturing, your data is distributed widely. The larger the company, the more data is generated, typically over a wider area. Energy and mining companies have to spend a good deal of their resources on production and delivery, each with its own challenges and exploration. This means well sites or potential sites across the country, if not across the continent, or even the globe. Retail organizations develop partnerships to provide new revenue streams, whether it is a new clothing line, a new source of raw materials, or the latest gadget. This means small offices in remote locations or infrastructure to connect to newly acquired organizations.  

Whatever the case is, such companies save money by making the sites as lean as possible. Retail organizations try to leverage the existing people working on a newly acquired brand or create transitional teams to bring the new opportunity under their umbrella. Resource companies sub-contract at exploratory sites rather than sending existing employees to a distant location. This eliminates the need for salaried employees in a downturn and makes it easier to manage shorterterm operations. The downside to this approach is that there is less control with regards to the data generated – subcontracted service operators or newly acquired organizational assets may have their own way of doing things or may simply have different equipment (or a different configuration of their equipment) than the home office they report to. In short, different operators produce different data sets.  

This means that the unstructured (or differently structured at least) data from all of these edge locations needs to be processed to some degree to be used effectively. This can be done to a degree at the edge with Fuusion by deploying Fuusion Edge nodes at each site. Still, in most cases, companies want this data aggregated in one place or at least in regional hubs. Because of the large amounts of data generated daily by multiple distributed sites, small though they may be, this aggregation is increasingly happening in the cloud, often in a data lake. 

Fuusion can help with this process as well, whether edge processing is used or not. Because of the wide distribution of the data endpoints, there is almost guaranteed latency across even the bestconnected locations. Some will have limited connectivity and speeds due to rural limitations. Some will have intermittent connectivity due to satellite networks. That fairtrade coffee farm in Guatemala? Until Starbucks foots the bill, it is unlikely to have as much connectivity as the coffee shop it supplies to. There might be hundreds of smaller operations of this nature that report needed data to a central office. This means that to get the data to our AWS or Azure data lake, something needs to be done to optimize the delivery. 

Enter Ultrafast™. Fuusion’s Ultrafast allows faster data transfer over high latency or limited bandwidth connections than standard TCP. Fuusion also ensures that if transfers are interrupted, such as when a satellite connection goes down for the day, that the transfer resumes at the exact point of failure when connectivity is reestablished automatically. To maximize the available bandwidth, Fuusion Ultrafast can also throttle or open bandwidth on a schedule, ensuring that edge locations do not hog resources during peak hours, for example.  

With Fuusion deployed at edge locations, the automated transfer of raw or semiprocessed data (depending on your needs) can be made simple. Once a flow is defined, it can be saved as a template for deployment across an organization. Github integration allows for version control as well. This means that several iterations of a given flow can be stored. This can be handy if some edge locations require a similar process group but with varying degrees of complexity. 

But at the end of the day, it’s about getting the data to where it needs to be, without having to worry about whether it arrived or whether anything is missing. Fuusion automates the flow of data to the data lake (or any designated repository), data provenance, and the ability to playback a flow ensures the data arrives intact, and Ultrafast ensures it arrives in a timely fashion. Let Buurst help you get your data where it needs to be.