PROCESS WARP is a distributed processing system that allows applications to run in a virtual space made up of multiple connected nodes. This allows unlimited scaling of computing resources and the ability to migrate processes between nodes seamlessly. It aims to break down the distinction between standalone and cloud applications by allowing desktop programs to utilize remote resources and move freely between offline and online usage. The system uses a special virtual machine implementation with its own CPU and memory model to achieve this distributed processing without a centralized server. It has potential benefits for increasing performance, reducing costs, and providing flexibility in application deployment.

1. PROCESS WARP
distributed processing system
Yuji Ito
facebook : https://www.facebook.com/ito.yuuji
github : https://github.com/llamerada-jp
I talk about PROCESS WARP that is a distributed processing system, I'm developing.

2. PROCESS WARP : A new infrastructure
for distributed processing system. All
machines are equal. It doesn’t require a
special machine as a server.
Typical system : Formed by client/
server. It requires a server, if the server is
down, service will stop.
※ Icons by Crystal Project (LGPL).
Aim

3. Feature
• Applications run in virtual
space, that is made up by a
group of nodes.
• Makes it possible to use any
amount of computer resources
and control nodes by a single
process.
• Makes it possible to warp a
process to another node with
keep alive.

4. Advantage
Increase total processing power
according to number of joining
nodes.
• Decrease cost to maintain
service without relation to
number of nodes.
• Service provider can choose a
good topology for system
having a lot of nodes.
join

5. Advantage
It breaks the border between a
standalone application and a
cloud application.
• User can use stand alone
application similar to cloud
application on PROCESS
WARP.
• User can use application
offline, after a migration
process.
(Under active development)
Because a process image can migrate, the
user can use the application offline.
Share the process space among the
connected computers and freely pass the
running application between machines.

6. Architecture
Host : native environment like UNIX, or web-browser
CPU : Interpret LLVM-IR
instructions without using
native stack
Memory : Store memory
image partially per node
other host
VM
Process Virtual Machine : Dump and synchronise (or pass)
memory image; Control PCB (Process Control Block); Call API
Use native APIs
Synchronise memory image, Control PCB
I'm developing a special virtual machine to realise previous slide's system.
LLVM is a compiler infrastructure that is used in many environments.

7. Actual status
• You can run virtual machine on FreeBSD, Linux, OSX, and Web-
browser.
• And run sample C/C++ program on virtual machine.
• Multiple threads can be distributed across several machines.
• Virtual machines have minimal APIs like “printf”.
• I published source code with MIT license and anyone can try it on
web-browser.
• http://www.processwarp.org/
• https://github.com/processwarp

8. Demo
#include <stdio.h>
#include <stdlib.h>
#include <pthread.h>
char* ret = "a";
static void *thread_func(void *vptr_args) {
int i;
for (i = 0; i < 20000; i++) {
printf(" b [%d]n", i);
}
return (void*)ret;
}
int main(void) {
int i;
pthread_t thread;
void* retval;
if (pthread_create
(&thread, NULL, thread_func, NULL)
!= 0) {
return EXIT_FAILURE;
}
for (i = 0; i < 10000; i++) {
printf("a [%d]n", i);
}
if (pthread_join(thread, &retval) != 0) {
return EXIT_FAILURE;
}
printf("ret %016llx %16llxn",
(unsigned long long)ret,
(unsigned long long)retval);
return EXIT_SUCCESS;
}
This program is a simple multi thread
application. This program makes two
threads and increments a counter on each
thread.

9. Next milestones
• Remove server (currently needs control server) and implement
peer to peer transfer.
• Make test, document, and more APIs to run existing applications.
• Develop I/O functions including storage and socket layers.

10. Thank you!