| Dart Home | PowerTCP SSL for ActiveX | Custom Development | Reply | PowerTCP SSL for ActiveX Topics | Forums |
| Author | Forum: PowerTCP SSL for ActiveX Topic: SecureFTPServer - slow STOR to remote folder |
| mmchenry From: Fort Lauderdale, FL USA Posts: 25 Member Since: 10/13/03 |
posted December 9, 2003 12:09 PM I'm using SecureFTPServer with a virtual file system. When the client STORs a file, I redirect it to a folder on a file server by changing Response.Data.FileName. This works, but is very slow compared to a RETR of the same file (the retrieve is also redirected to the same file server). The RETR is 3 to 4 times as fast as the STOR. This is true with various FTP clients, so I don't think it's a client side issue. If I don't redirect the file and allow it to be stored on the FTP server's local drive, performance is the same as the RETR (fast). Client, FTP server and file server machines are on the same switched 100Mb LAN. Copying files between all three machines using Windows Explorer is very fast and similar in performance to the RETR, so I don't think there's any network problem. Any thought's on what might be causing this or what else to try? I noticed some old posts regarding this issue in a previous version of the standard FTP controls ( http://support.dart.com/postings?topicid=1953 ). Is it possible that the problem still exists in the secure FTP controls? |
Tony Priest From: Utica, NY USA Posts: 8466 Member Since: 04/11/00 |
posted December 9, 2003 12:43 PM The posting you referenced is not really related. He was complaining that the transfer was 100x slower. This was fixed by adjusting the buffer size. The problem you are experiencing is different. Probably what is happening is that the buffersize that makes for faster TCP-IP response time is probably not as efficient for the local machine to local machine traffic. What happens if instead of "redirecting" the file, you make the root be a mapped drive that is the same as where you were getting redirected? Or what about letting the file come in to the local drive first, then use IO functions to copy the file to the Network location. Do either of those methods improve performance? |
| mmchenry From: Fort Lauderdale, FL USA Posts: 25 Member Since: 10/13/03 |
posted December 9, 2003 5:31 PM It doesn't make any difference if the root and files are on the file server, or the root is local and the files are on the file server. In both cases RETR is 2-3 times as fast as STOR. Storing to the local drive then moving to the file server is better, but obviously not ideal. It's one more thing to go wrong and causes the store event to take longer to complete. Wouldn't a buffer size issue also effect the RETR speed? Are you talking about something I can tweak in the TCP stack, or is this something internal to the server control. Are you able to try this scenario yourself? It should only take a few minutes to setup. You can actually run the sample client and server on your local machine and just modify the sample server to STOR/RETR files to/from a network share. Assuming you're on 100Mb, a 10MB file will show the difference between STOR and RETR quite dramatically. |
| Tony Priest From: Utica, NY USA Posts: 8466 Member Since: 04/11/00 |
posted December 10, 2003 11:38 AM I have confirmed what you say, but I don't have a workaround. The issue has been logged as #2226 if you decide to call in and check on the progress. |
| mmchenry From: Fort Lauderdale, FL USA Posts: 25 Member Since: 10/13/03 |
posted December 10, 2003 12:44 PM Here's some more info in case it helps: Redirecting the STOR is not really the issue, the problem is just more obvious that way. Storing files on the local drive is proportionally just as slow. To create a simpler test environment I'm using your sample client with 4 different servers: the sample server, GlobalScape, Nofeel and Ability. All servers were configured as similarly as possible, the sample client used the same settings in all cases (implicit normal). Client is an XP machine, server is Win2k. Both on the same 100Mb LAN. I transferred the same 15MB file for all tests. With GlobalScape, Nofeel and Ability, STOR and RETR are both 4-5MBps (or greater) according to the sample client. With the sample server, RETR is the same as the other servers at about 4-5MBps. STOR, however, averages around 1.5MBps. It seems to start at a higher speed, then slows down as the transfer continues. It seem to pause briefly every second or so after the first few MB have transferred. Finally, I tried a test using the sample server with security set to "none". Same results, so it doesn't seem to be related to security/encryption. Hope this helps. Is there any rough time frame for the next update to the SSL Tools? I know you can't give an actual date, I'm just wondering if it weeks, months, etc. |
| Tony Priest From: Utica, NY USA Posts: 8466 Member Since: 04/11/00 |
posted December 10, 2003 12:58 PM Please contact our support department directly to get that info: support@dart.com 315-790-5456 |
| Reply | PowerTCP SSL for ActiveX Topics | Forums |
This site is powered by
PowerTCP WebServer for ActiveX
|