Using SSD on Fiber optic network
I have a 1000 base T network set up to run high speed cameras. I use the 1000BaseT network along multimode fiber runs and fiber converter cards to connect the cameras to my PC. My download times are slow. Would using a SSD dramatically speed up my acquistion time?
Depends on what you mean by "Slow." How fast is the data currently being stored, in terms of Megabtyes per second (MB/s)?
1000 base T is not as fast as the write speed of most modern disk drives. It's 1000 megaBITs per second, for a theoretical 100 megaBytes per second. But 1000 base Anything is slower than a disk drive. SATA 1 was 1.5 gigaBits per second, is is not fast enough for current drives. (Come to think of it, 1000 base T is copper. Multimode fiber is -SX or -LX, isn't it?)
1) Are you recording live from the high-speed camera and not able to keep up, or does it have adequate on-board storage so we are only talking about annoying download times?
2) What are the "slow download times" that you observe? How many actual MB/s?
3) Does the camera manual say what data rate they pump out when they are running?
4) Does the camera itself have a 1000 base connector? If not, how is it connected to the first computer or piece of networking equipment; that may be the limiting factor.
There's a good chance that I can't help you, but knowing these details will improve our chances.
While we're on the topic of "theoretical" vs "actual", I've never seen my gigE network able to transfer files between machines running SATA2 hard disks over 60 MB/s. I'm not claiming to have the fastest or best setup, but I've seen the same behavior over several different PCs on different networks over the years.
I've always assumed I was maxing out the write speeds of my hard disks on my slow file server.
I've seen charts record performance HDD write speed up to 160 MB/s. I think gigE should top out at 119 MB/s.
I'll be getting my SSDs this week, and I'd be able to confirm my idea that my transfers should be faster. That's assuming the HDDs I'm using have a higher read rate then 60 MB/s. The SSD will have no problem writing faster then my network can transfer.
Sorry for the late update, I am seeing 5 megaBYTES per second downloads.
To quickly answer some questions posed above ...
The cameras record to a buffer that is recorded once the camera is triggered. This happens real time. I suspect thats one of the main reasons why high speed cameras are so expensive, because of the high performance buffers contained inside them. I am not sure exactly how this works, but files can be as large as 10 gigabytes in one recording and the cameras are designed to record this all live. I then access this buffer by plugging in cat6 into the back of the camera.
Heres a quick overview of the setup
High speed camera --> Cat 6 ---> Omnitron Gigabit fiber converter --> multimode fiber xxxxx number of feet --> fiber converter rack (gigabit) ---> cat 6 --->acquisition computer (in this case a Dell t7400)
I will have to get back to you on the specs on the dell t7400 but I have 4 1tb harddrives in RAID 0 7200rpm each.
I have contacted the camera manufacturer regarding output speed of the camera, but this should not be the problem as far as I know.
At one point there were certain network protections installed on my t7400 which I think may have slowed it down somehow. I am in the process of having our IT guy benchmark both my t7400 and an "untouched" t7400 we have in another lab. Will report back.
Bottom line, I'm still getting 5 megabytes/sec.
Well, at least the data is being buffered at the camera, so you are not losing it.
This is one of those cases where I suggest benchmarking sub-parts of the lashup.
Do a write-speed benchmark on the drives, just to make sure that you don't have an insanely slow RAID array. If your test comes out at 5 MB/sec, then your network is irrelevant.
Then put the camera and the PC in the same room (I sincerely hope that this is possible) and connect them directly with a crossover cable if that will work, or at least connect them both to the same gigabit copper hub.
If these sub-tests run at gigabit speeds, then your problem seems to be somewhere in the converters / fiber transmission. If these sub-tests run at 5 MB/sec, you are that much closer to solving the problem.
Actually, if the second test that I described yields 5 MB/sec and the first test showed much better speeds, you want to try two separate tests: That PC to another PC (like the untouched t7400) and the camera to another PC. All of this is just basic isolate-the-components testing to find where your weak link is.
See you in another week?
Ok, thanks for the quick reply.
We are thinking on the same wavelength.
I unfortunately am prohibited from doing any administrative tasks on any machine which makes dealing with our camera networks and other aspects of my job very difficult...redtape...privilege limitation I guess...anyway...I got our IT guy in to run a benchmark on the acquisition T7400 as well as the untouched T7400. The acq. computer scored a 990 while the untouched computer scored a 1080. I am sorry, I do not know what kind of benchmark he used, I will ask him tomorrow. I don't have enough free time at work to be devoted solely to this issue and unfortunately have to deal with a ton of other things so bare with me. Anyway, I am thinking the two identical machines may have scored differently due to the nature of their usage. The untouched machine remains relatively unused/extremely light usage while the acquisition computer is constantly dealing with downloading and moving huge image files to a server. I am thinking it may run slower due to the amount of data on its disks and how it is constantly being rearranged. With this in mind, I am writing off the notion that the security measures in place of the acquisition computer are whats causing it to slow down.
I did get the time to run a true download speed test after they were benchmarked. I set up an identical acquisition network using all of the equipment I would use while field testing albeit on a small scale. Camera --> short run of cat6 --fiber card--multimode fiber --- fiber card --- cat 6--- acq computer. I went into my camera software and saved the acquisition configuration and network configuration so I could apply it when I tried this set up on the untouched machine. Long story short the untouched machine ran the download 10 seconds faster. I calculated the transfer speed to be roughly 6.88 megabytes per second. I can live with the 10 second difference so I am not so much worrying about my acquisition computer being corrupted by its security measures or disk space issues.
I then did exactly what you thought and connected the camera and machine directly with cat6. The results were the same. With that being said, I am still waiting to speak with the camera manufacturer. At least I know the fiber cards are not what is chocking it.
One interesting thing is that I can change the cameras network settings from "fast/less reliable" to "slow/more reliable". There is a slider bar with some nominal values applied. I guess that I am modifying packet sizes here? Problem is "fast/less reliable" is exactly that, faster but less reliable. I cannot risk grenading the data during the transfer as many of the images gathered from the high speed cameras are not repeatable and the tests can cost upwards of a million dollars. Losing the data is not an option and is the most stressful part of operating the camera network during test events.
I will have to run another test with it switched to "faster/less reliable" just to see how much it will max out my speed. Recalling previous experience, it should half the download speed, essentially giving me 12-13megabytes/sec. Should I be expecting more from a gigabit connection? A lot of searching has only provided my with the theoretical 120megabyte/sec max.
PS just reading over your reply above and although I dont remember the exact scores of the harddrives during the harddrive portion of the benchmark. I know they scored near the top end of the range in the report, so I assume its not the drives. I will see if I can report back those scores tomorrow.
Perhaps the IT guy could run some of the tests described?
If a direct connection camera-to-computer with a good crossover cable runs at 6.88 MB/sec, at best 70 mbits/sec, then you either have
1) 100 megabit Ethernet on the PC
2) 100 megabit Ethernet on the camera
3) A really lousy cable
3) A broken gigabit port somewhere.
If either device is only 100 megabit Ethernet, then running gigabit Ethernet between them won't make any difference. You could run Infiniband between them. My best guess is that at least one of the two, camera or computer, has 100Base-TX instead of 1000Base-TX. That extra zero is pretty important.