For our remote users on Forticlient VPN:
We get about 1 MB/second when Getting files from PDM Professional. (Ping to server is 88ms)
Internet speed at either end is definitely not the limiting factor. Doesn’t matter if both ends are on gigabit internet (nominally 125 MB/s), that’s still about the fastest it gets.
This kind of performance also causes 30 minute check-ins, which doesn’t encourage good practices.
I simply stopwatch time a big file to test. What kind of speed do other users see over VPN?
Would it be better if we got cloud hosting in a data center for our PDM server?
We switched to Zscaler VPN…I’ll need to test at home to compare. How much data is getting uploaded for that 30 minute check-in?
We also just got forcibly moved to the cloud for all our DC servers so its all slower than it used to be in the Office. 1-2ms ping to servers just went to around 35-40ms. Working on CAD files off network servers for the few departments that aren’t using PDM went from 30 seconds to several minutes for every open and save on some larger files. Some stuff in the cloud makes sense but large CAD file servers doesn’t, I wasn’t given a choice though. Also my server budget doubled in the cloud…so slower access and double the cost. I used to use older PDM servers to test/dev, now I have to pay for cloud servers for test and dev.
Please do let me know when you test it. I don’t know what is possible with a VPN-PDM setup.
We tried WireGuard VPN, which uses a supposedly superior connection tech compared to Foritclient, but didn’t see any change in PDM (tho it does seem smoother for remote desktop work).
How much in my casual 30 min check-in example? Not sure, for an existing file in PDM I’m not sure how much is actually new. We do have some big files, assemblies around 300MB. Tested simple upload/addition of new file just now, and it was 0.6MB/s.
Cloud hosting was one possible approach I was considering. Sounds like a bummer for you. Seems like Saves wouldn’t take much longer (since still saving to local drive, no?), just check-in.
I wonder if any of this would apply to you: https://www.javelin-tech.com/blog/2020/03/using-solidworks-pdm-remotely-during-covid-19/
“”“Option 2: Disable automatic updates in the PDM Add in within SOLDIWORKS. Go to Tools > SOLIDWORKS PDM > Options and on the “Server” tab, uncheck all options under “Automatically update the tree in the extra tree pane” and ensure that the option “Reload tree when selecting a component that has been unloaded” is set to “Never”.”“”
This is only for the non PDM users. There are a few departments that use SolidWorks but not PDM. More on the order side as we didn’t have the capacity to store order data in PDM. They would eat through all of my PDM disk space in a year or less.
We do have users toggle those depending on their location and preference. They made some performance improvements to those settings I think but it will still slow things down some. A few users turn off the PDM-add-in completely. They just do their checkout/in outside of SolidWorks in Windows.
Sorry, I registered the “Not PDM” part of the comment too late.
Measuring speed just by using the progress window. I find a 40-100MB file, un-cache it, then Get latest and use a stopwatch. Not fancy, but actual Get speed is what I’m interested in. Or for upload speed I add a new file. The faster your speed, the bigger the file you need to get a decent measurement of the seconds.
Ditto on the latencies. It seems there more time in client chatter to the servers than actual bulk file moving, so latencies seem to be king.
When I use VPN at home (~40MB down 10MB up service, through the VPN and other appliances seems to vary every time I check.) Get file times are not so bad, check ins can be slower. Many factors can affect, there’s a lot of network chatter during the check in process.
Thanks bnemec, can you post a simple speed measurement here, next time you get a chance to test PDM on VPN?
**For all involved - note the difference between little b and big B. I measured 1 MegaByte a second (=8Mbps). Internet speeds are often hyped as Mbps, Megabits per second. Use either unit, but be clear (though MByte/sec is far more useful real world).
My mistake, that’s Mbps from my ISP at home, bits not bytes.
I’ll take a box home this weekend and VPN in then do some ping tests to the sql and archive servers. Both my house and work are rural ISPs so the back haul to the www is busy. We see significant drop in connection speeds to the cloud hosted ERP depending on time of day. Less in the summer, but during school year we can tell when the kids all fire up their devices around 3:30-4:00 our latencies from main office to OCI goes to crap.
You have a slower upload speed thus the slower check-in. Although I would expect “Get” to be worse since there are more files involved…unless you checked out all references and checked them all in. I have 1gb up/down fiber at home so I guess it would be the same?
Brian-M
I just cleared my local cache and did a “Get” on an assembly drawing and all references at work. 236 total files at 1.3gb for the entire data set. I’ll try this weekend at home on VPN and see how it goes.
Frederick_Law
Fair enough. I’m also interested in actual check-in times. But it’s more messy, more variables.
I picked a modest assembly. 9500 KB sldasm with 3 parts (11068, 6966, 5023 KB). I used Copy Tree so they were fresh. Took 95 seconds. 9.5+11+7+5 = 32.5MB. → 0.34 MB/sec.
Then I checked-out, added a boss to all 3 parts, saved assembly → Check-in Took 98 seconds, 0.33 MB/sec.
Didn’t expect it to be that close the 2nd time.
IMHO the ideal scenario would be (if possible) to connect to a local machine in the company via a remote desktop application using the VPN with a normal PC or notebook.
To avoid speed problem during data transfer and the lag between PDM SQL and remote PC causing slow downs in the whole PDM interface. there is even the risk a disconnection from sql server could cause the reset of some registry settings related to PDM in the workstation (like columns, there is an SPR and happened to one of our engineers)
HP offers a program (free of charge until this years end) that allow to connect remotely, supporting multi display and remote 3D mouse: ZCentral Remote Boost. It is better than MS remote desktop as it uses the remote workstation GPU.
Ok, my numbers in the office on 1gb ethernet vs home on 1gb fiber but connected via wifi. I may need to try work again using wifi to get a fair comparison. Servers are in Azure.
Data set:
Assembly drawing and references, 236 files, 1.3gb total.
Office - 1gb ethernet:
Get (Show dialogue)- 3 seconds
Get (Operation) - 410 seconds
Home - 1gb fiber - VPN:
Get (Show dialogue)- 3 seconds
Get (Operation) - 858 seconds
My home wifi is on the other side of my house so there could be some speed loss. Checking the connection, I have 3 to 4 bars and speed fluctuates between 108 to 180 Mbps. Ideally I need to try plugging directly into the router.
I myself used the solution the solution of Teradici also internally in the office, my workstation is in the server room, which provides a pleasant climate in the office at the current high temperatures.
Download a STEP file or any assembly not in Vault yet.
So it’s a new assembly and check it in.
You can delete it from Vault after the test and repeat the test later.
Make 2 sets of assemblies. Both with similar total files size. 100MB if possible.
One set with more files then the other set.
Good point, and we do sometimes do this. This post is partly to determine if your system is healthy, and the speeds we see are just normal slow. In the future we may put more resources into the remote desktop approach.
Thanks Jason! So 1300MB/858 = 1.5 MB/s on VPN. Similar to our 1 MB/s. I doubt you will see faster plugged in with an ethernet cable at home, since your wifi speeds are not the weak link. Ethernet is more reliable, but in general I found the same times on wifi (in the case of receiving from PDM).
3.2 MB/s at the office seems poor. I think we usually experience roughly 40-80 MB/s Get Latest transfers when connected at the office.
You need a common data set for benchmark.
If data set is different, result cannot be related.
Large file test bandwidth.
Lots of small files test latency.
I believe problem is latency.
Would be nice if you can see what the server is doing during test.
