[2014] Optimizing I/O and Memory for Tablet Devices

[2014] Optimizing I/O and Memory for Tablet Devices

Table of Contents

1. Introduction
2. Optimizing Virtualization and VM Memory Management
3. Running Windows in a Virtual Machine on Android
4. Hardware Specifications
5. References: Limbo and Intel KVM
6. Configuring QEMU and Kernel with PAE Support
7. Virtualized IO Devices
8. Improving VM Loading Time
9. Ballooning Policy for Memory Management
10. Interface between Android and QEMU Virtual Devices
11. Assigning More Than 1GB of RAM to a Virtual Machine

Introduction

In this article, we will discuss our efforts in the Android and KVM (Kayvyun QMU) virtualization technology. Our focus will be on optimizing virtualization and VM memory management while running Windows 8.1 in a virtual machine on Android KitKat OS. We will explore the hardware specifications used for our tests and discuss the references we utilized, including Limbo and Intel KVM. Additionally, we will cover topics such as configuring QEMU and the kernel with PAE support, virtualized IO devices, and improving VM loading time. We will also Delve into the ballooning policy for memory management, the interface between Android and QEMU virtual devices, and the assignment of more than 1GB of RAM to a virtual machine.

Optimizing Virtualization and VM Memory Management

To ensure smooth and efficient virtualization on Android, we have focused on optimizing our virtualization and VM memory management processes. By running Windows 8.1 on Android KitKat OS, We Are able to provide users with a seamless experience while utilizing virtualization technology. Our efforts have been aimed at enhancing the performance and memory utilization of virtual machines, thereby improving the overall user experience.

Running Windows in a Virtual Machine on Android

Running Windows in a virtual machine on Android opens up a world of possibilities for users. With the Atom tablet and our optimized virtualization techniques, users can seamlessly switch between the Android and Windows operating systems. By utilizing screenshots and a tablet interface, Windows is treated as an application within the Android environment. This integration allows for multitasking, Bluetooth and Wi-Fi connectivity, battery status monitoring, and audio and 2D/3D graphics rendering within the virtual machine. However, there are challenges to overcome, such as the loading time for Windows and memory management within the limited virtual address space.

Hardware Specifications

To achieve optimal performance, we utilized a Paytrail Intel Atom processor with four gigabytes of RAM and 64 gigabytes of eMMC storage. For our tests, we allocated three gigabytes of RAM for the virtual machine, with one gigabyte of RAM and 32 gigabytes of virtual disk assigned to both Android and Windows. It's important to note that Windows was running a 32-bit OS on the Android platform.

References: Limbo and Intel KVM

In our efforts to optimize virtualization, we worked with two key references: Limbo and Intel KVM. Limbo is a wrapper program that allows us to run QEMU on Android. Intel enabled Limbo with KVM support, adding necessary system calls and functions to improve performance. Samsung even contributed to Limbo by updating its QEMU version to 1.7, the most recent and stable version at that time. By leveraging these references, we were able to enhance the virtualization capabilities on Android.

Configuring QEMU and Kernel with PAE Support

To enable Windows to utilize more than one gigabyte of RAM within the limited virtual address space in Limbo on 32-bit hosts, we had to configure QEMU and the kernel with Physical Address Extension (PAE) support. This allowed Windows to manage its memory efficiently while running in the virtual environment. By enabling the VM to use more than one gigabyte of RAM, we were able to provide a better user experience without sacrificing system performance.

Virtualized IO Devices

To ensure a seamless user experience, we virtualized crucial input/output (IO) devices that are frequently used on tablet devices. This included multitouch, Bluetooth, Wi-Fi access, battery status monitoring, and audio and 2D/3D graphics rendering. By virtualizing these devices, users can enjoy full functionality within the virtual machine environment.

Improving VM Loading Time

One of the challenges we faced was the loading time for virtual machines. Users expect quick and responsive performance when launching applications, including virtual machines. To address this challenge, we implemented various techniques to reduce the time it takes to load a virtual machine snapshot. By utilizing snapshot images and optimizing the initialization process of the QEMU virtual machine, we were able to achieve significant improvements in loading time.

Ballooning Policy for Memory Management

Memory management is a critical aspect of virtualization technology. To optimize memory utilization, we implemented a ballooning policy that adjusts the level of memory in response to host memory pressure. By monitoring host memory and controlling balloon inflation and deflation in the virtual machine, we ensure that memory resources are effectively distributed and prioritize important applications. This dynamic memory management approach helps enhance the overall performance and responsiveness of the virtual machine environment.

Interface between Android and QEMU Virtual Devices

To facilitate seamless communication between Android and QEMU virtual devices, we implemented an interface architecture. This involved modifying BlueDroid and Limbo, as well as integrating QEMU virtual device models and the Android event handling mechanism. The interface allowed for the virtualization of Bluetooth, multi-touch, audio, and battery status monitoring, ensuring smooth interaction between the Android and virtual machine environments.

Assigning More Than 1GB of RAM to a Virtual Machine

Traditionally, one of the challenges in virtualization was allocating more than one gigabyte of RAM to a virtual machine. However, through our enhancements in Limbo and QEMU, we were able to assign more than 1GB of RAM to a virtual machine running on Android. By leveraging the unused memory address space in Limbo, we expanded the capabilities of virtual machines, allowing for improved performance and expanded memory capacity.

Conclusion

In this article, we have discussed the efforts made in Android and KVM virtualization to enhance the virtualization technology and improve the performance and functionality of virtual machines running on Android devices. We have explored various techniques, such as optimizing memory management, reducing VM loading time, and interfacing between Android and QEMU virtual devices. These enhancements provide users with a seamless experience when running Windows in a virtual machine on Android. The future of virtualization on Android looks promising, with continued advancements and optimizations in memory management, device virtualization, and overall performance.

Highlights

• Optimal virtualization and VM memory management techniques for Android and Windows

• Hardware specifications and references utilized for virtualization

• Enhancements in QEMU and kernel configuration for efficient memory utilization

• Virtualization of crucial IO devices for seamless user experience

• Techniques to improve VM loading time and reduce startup delays

• Ballooning policy for dynamic memory management in virtual machines

• Interface architecture for smooth interaction between Android and QEMU virtual devices

• Assignment of more than 1GB of RAM to virtual machines on Android

• Continued advancements in Android virtualization for improved performance and functionality.

• Pros:

  • Seamless integration of Android and Windows operating systems
  • Enhanced performance and memory utilization in virtual machines
  • Efficient memory management through ballooning policy
  • Improved loading time and responsiveness of virtual machines
  • Smooth interaction between Android and QEMU virtual devices

• Cons:

  • Limited virtual address space for Windows running on Android
  • Challenges in allocating more than 1GB of RAM to virtual machines

FAQs

Q: Can I start Windows guests from a snapshot image? What happens if the guest crashes or needs to be rebooted?
A: Yes, You can start Windows guests from a snapshot image. If the guest crashes or needs to be rebooted, you can revert to the snapshot image to ensure the consistency of the VM state.

Q: How do you handle user-installed applications in the virtual machine?
A: When the user installs applications in the virtual machine, we typically take an additional snapshot. This allows us to preserve the state of the VM, including the installed applications, for future use.

Q: How does the ballooning policy manage memory between the host and guest systems?
A: The ballooning policy dynamically adjusts the memory allocation between the host and guest systems Based on memory pressure. This ensures optimal memory utilization and prioritizes important applications in both environments.

Q: Can you explain the interface architecture between Android and QEMU virtual devices?
A: The interface architecture involves modifying BlueDroid and Limbo, integrating QEMU virtual device models and the Android event handling mechanism. This allows for seamless communication and interaction between Android and QEMU virtual devices.

Q: What are the primary benefits of assigning more than 1GB of RAM to a virtual machine on Android?
A: By assigning more than 1GB of RAM to a virtual machine, we enhance the performance and memory capacity of the virtual environment. This allows for smoother multitasking, improved application performance, and enhanced overall user experience.