So anyone who has had to live through reading my noob questions and posts (
) may know I already mentioned Ninja's weakness in not working to use 100% of the processor for any large system. I have a 970X (6 core, around 110 GFlops, about four times as powerful as a Quad Core Q6600) that will not be fully used by Ninja. I also tried it on a cluster - same problem.
I'm working with Ninja almost daily now, even though I'm not placing live trades through it. Despite the bugs, I'm coming to like the UI, support, efficiency (when on a smaller normal workstation), automated trades, etc. better than others. I've decided to propose using Ninja with our firm, however the biggest hurdle will be the same thing as my personal dislike - the inability to use all of the CPU on larger systems. Given other threads here, as well as comments I've heard from two friends, this is not a unique problem.
I'm getting ready to build a new system... it will not be a complex cluster, just dual Xeon 5600s working out to around 250-500GFlops CPU only. There isn't much reason to build it if Ninja continues to be unpredictable in using all of the CPU power. I'm not a software engineer, but I am a programmer, and understand this isn't a unique problem among software - I would be willing to be my money it's an easy fix.
I'm not suggesting you should add to your business plan to go up against Bloomberg, however Ninja IS an extraordinary platform at a good price ($1.5k/license) and is something I'd like to spend even more time (and money) with, especially with my new system. Other people are in the same position.
Bottom line: commit a day (or even less) to figuring out why Ninja won't use all of the CPU on larger systems. It might be as easy as fixing an algorithm so it doesn't confuse a large CPU to mean stay on one core (it looks like Ninja is designed to switch between one and multiple cores).
I'll post a reply when more comes on my use with Ninja in datamining, and my friends and coworkers' opinions on using it. Best of luck, c
) may know I already mentioned Ninja's weakness in not working to use 100% of the processor for any large system. I have a 970X (6 core, around 110 GFlops, about four times as powerful as a Quad Core Q6600) that will not be fully used by Ninja. I also tried it on a cluster - same problem. I'm working with Ninja almost daily now, even though I'm not placing live trades through it. Despite the bugs, I'm coming to like the UI, support, efficiency (when on a smaller normal workstation), automated trades, etc. better than others. I've decided to propose using Ninja with our firm, however the biggest hurdle will be the same thing as my personal dislike - the inability to use all of the CPU on larger systems. Given other threads here, as well as comments I've heard from two friends, this is not a unique problem.
I'm getting ready to build a new system... it will not be a complex cluster, just dual Xeon 5600s working out to around 250-500GFlops CPU only. There isn't much reason to build it if Ninja continues to be unpredictable in using all of the CPU power. I'm not a software engineer, but I am a programmer, and understand this isn't a unique problem among software - I would be willing to be my money it's an easy fix.
I'm not suggesting you should add to your business plan to go up against Bloomberg, however Ninja IS an extraordinary platform at a good price ($1.5k/license) and is something I'd like to spend even more time (and money) with, especially with my new system. Other people are in the same position.
Bottom line: commit a day (or even less) to figuring out why Ninja won't use all of the CPU on larger systems. It might be as easy as fixing an algorithm so it doesn't confuse a large CPU to mean stay on one core (it looks like Ninja is designed to switch between one and multiple cores).
I'll post a reply when more comes on my use with Ninja in datamining, and my friends and coworkers' opinions on using it. Best of luck, c
Comment