Very slow source

[cooch1811]

Sorry if this is a repeated topic, but I can't seem to find my answer!

I have a MacPro Dual 2.8GHz QuadCore Xenon running 10.5.2.

 

Typical performance I get is listed below.

 

Other info:

1) This computer just replaced a Dual 2GHZ G5 running 10.4.x, which ran the same scripts to the same locations on the same servers with the same ftp accounts. The G5 typically ran 350 to 500 MB/minute.

2) Other G4's on the same network back up the same way (ftp), and run 150 to 250 MB/minute.

3) all backup sets are files

 

Please help!

My 50GB backups are killing me!

 

cooch

 

 

 

∆ Retrospect version 6.1.138

automatically launched at 2/28/2008 8:00 AM

+ Retrospect Driver Update, version 6.1.13.101

 

+ Normal backup using Peony_DUDS_daily at 2/28/2008 8:00 AM

To backup set Peony_DUDS_Set…

 

- 2/28/2008 8:00:24 AM: Copying DUDS…

2/28/2008 8:16:48 AM: Execution completed successfully.

Completed: 43 files, 734.1 MB, with 22% compression

Performance: 44.9 MB/minute

Duration: 00:16:24 (00:00:04 idle/loading/preparing)

 

+ Normal backup using Peony_Main at 2/28/2008 8:17 AM

To backup set Peony_Main_Set…

 

- 2/28/2008 8:17:06 AM: Copying Main on Files…

2/28/2008 8:18:21 AM: Execution completed successfully.

Completed: 10 files, 61.7 MB, with 57% compression

Performance: 60.6 MB/minute

Duration: 00:01:15 (00:00:14 idle/loading/preparing)

 

- 2/28/2008 8:18:23 AM: Copying Retrospect 6.1 on peony…

2/28/2008 8:18:47 AM: Execution completed successfully.

Completed: 2 files, 1880 KB, with 0% compression

Performance: 11.0 MB/minute

Duration: 00:00:24 (00:00:14 idle/loading/preparing)

 

- 2/28/2008 8:18:49 AM: Copying Retrospect on peony…

2/28/2008 8:19:09 AM: Execution completed successfully.

Completed: 3 files, 151 KB, with 0% compression

Performance: 1.2 MB/minute

Duration: 00:00:20 (00:00:13 idle/loading/preparing)

 

2/28/2008 8:19:11 AM: Execution completed successfully.

Total performance: 46.5 MB/minute with 58% compression

Total duration: 00:02:03 (00:00:41 idle/loading/preparing)

 

+ Normal backup using Peony_myScraps_daily at 2/28/2008 8:20 AM

To backup set Peony_myScraps_Set…

 

- 2/28/2008 8:20:17 AM: Copying MyScraps on Files…

2/28/2008 8:20:17 AM: No files need to be copied.

2/28/2008 8:20:32 AM: Execution completed successfully.

Duration: 00:00:15 (00:00:12 idle/loading/preparing)

 

Quit at 2/28/2008 8:20 AM

[Mayoff]

The G5 probably backs up faster then the G4 because it is a faster computer.

 

The log section you talk about doesn't show any 50GB backups, so I am not sure what you are asking.

[CallMeDave]

Certainly 44.9 MB/minute performance is painful, but your post is lacking in some critical descriptions:

 

> the same scripts to the same locations on the same servers with the same ftp accounts

What does this describe? Are your File Backup Sets stored on a remote volume?

 

 

> Other G4's on the same network back up the same way (ftp), and run 150 to 250 MB/minute.

> all backup sets are files

File Backup Sets are not Internet Backup Sets. What, exactly, is your setup and configuration?

[cooch1811]

The G5 probably backs up faster then the G4 because it is a faster computer.

 

The log section you talk about doesn't show any 50GB backups, so I am not sure what you are asking.

 

Sorry I wasn't clear on these points!

1) The computer with the problem is a new mac Pro as described in the post.

2) The references to the G5 and G4 were merely to say that we have had other computers on the same network that had much faster performance

3) The log was simply to show one example of a recent daily backup; not of a 50GB backup. The large backups occur less frequently for re-cycles.

 

The question I am asking is basically "why would the Mac Pro perform so slowly compared to the others, and what should I look for?"

 

Thanks!

 

 

 

[cooch1811]

Certainly 44.9 MB/minute performance is painful, but your post is lacking in some critical descriptions:

 

> the same scripts to the same locations on the same servers with the same ftp accounts

 

What does this describe? Are your File Backup Sets stored on a remote volume?

 

 

The backup sets are "Internet Backup Sets". We are using ftp to send the files to the hard drive of a G4 on our internal network.

 

 

 

> Other G4's on the same network back up the same way (ftp), and run 150 to 250 MB/minute.

> all backup sets are files

 

File Backup Sets are not Internet Backup Sets. What, exactly, is your setup and configuration?

 

Sorry; see above. All computers are using Internet Backup Sets over the same network (at different times) to the same G4 computer hard drive.

 

Thanks!

[Mayoff]

FTP/Internet Backup sets are not usually a very fast way to do a backup. Think about this, you have 50GB of data, being copied to an FTP server in 16 MB chunks, the result will be 1000's of "chunks" on the FTP server. This will result in slower speeds.

 

What happens if you write the data to the network server using a file backup set instead of FTP? What is the speed like?

[cooch1811]

FTP/Internet Backup sets are not usually a very fast way to do a backup. Think about this, you have 50GB of data, being copied to an FTP server in 16 MB chunks, the result will be 1000's of "chunks" on the FTP server. This will result in slower speeds.

 

I understand this, but it doesn't answer my basic question. We have done similar backups using slower computers, and their backup speeds were faster.

 

The computer has blazing speed, the network runs 100MB/sec, the harddrives write at least 100MB/sec, yet Retrospect runs at ~50 MB/minute, which is at least 100 times slower than the backup path capacity.

 

What happens if you write the data to the network server using a file backup set instead of FTP? What is the speed like?

 

Same slow speeds

[Mayoff]

Then this tells us the problem is not writing to the FTP site (good info to know).

 

If you go to Configure>Clients and Get Info on the source client, what is the connection speed displayed? How does it compare with other clients that backup fast? Feel free to attach a screenshot of the get info window.

[CallMeDave]

If you go to Configure>Clients and Get Info on the source client, what is the connection speed displayed?

 

I don't see the original poster identifying using client software at all.

 

- Confirm that Retrospect is running on the machine in question, and that the Source(s) are local attached volumes on that machine?

 

 

Dave

[cooch1811]

I don't see the original poster identifying using client software at all.

 

- Confirm that Retrospect is running on the machine in question, and that the Source(s) are local attached volumes on that machine?

We are pretty simple people here, so I'm getting a little confused about the client questions, because there are no clients listed...We don't have a client/server setup, just a simple computer-to-computer communication.

 

Based on your questions, I'm thinking that maybe we are not set up the way we should be. So, I'm going to back up a second (no pun intended) and tell you how we are set up:

 

We have a simple home network with 7 macs connected through a Linksys router and several switches. We use Retrospect Backup 6.126 to back up 2 of the computers to one G4 400 computer's internal hard drive via ftp. Currently, one of the computers we back up is a new Mac Pro, and the other is a G5. The Mac Pro backups run at roughly 60 MB/minute speeds, and the G5 runs at roughly 250 MB/minute speeds.

 

We create backup scripts using the Easyscript feature on each computer, and we follow the wizard choosing the internet option for backup media. We enter the ftp info, using the local computer name (like computer1.local) instead of an ip address because the ip's change every time we shut down computers.

 

The Mac Pro is a very recent purchase. It took the place of the G5, which has now been physically moved. However, before being moved, the G5 was connected where the Mac Pro now is, it was doing the job of the Mac Pro, and was running backups at 250+ MB/minute. So...as far as we can tell, the only difference in setup is we traded the G5 for the Pro.

 

As for the G4 mentioned in a previous post: it has been replaced by the G5. The G4 ran backups at roughly 150 MB/minute, and now the G5 in its place runs around 250MB/minute, which makes sense.

 

Does this make it clearer what we are doing?

 

 

 

 

[Mayoff]

We use Retrospect Backup 6.126 to back up 2 of the computers to one G4 400 computer's internal hard drive via ftp. Currently, one of the computers we back up is a new Mac Pro, and the other is a G5. The Mac Pro backups run at roughly 60 MB/minute speeds, and the G5 runs at roughly 250 MB/minute speeds.

 

What method is Retrospect using to talk to the other computers? Are you using filesharing or did you install the Retrospect Client software? How does Retrospect "see" this other computer?

 

Is this a wired connection or a wireless connection?

[CallMeDave]

We use Retrospect Backup 6.126 to back up 2 of the computers...

 

This is the place where some confusion begins. Do you have a copy of Retrospect (each with its own license) installed on each of these two computers?

 

The answer to the question quoted above would have made this clear; is Retrospect running on the machine that's giving you unexpectedly slow performance? And is the Source for those backups the internal hard drive of that same machine?

[cooch1811]

What method is Retrospect using to talk to the other computers? Are you using filesharing or did you install the Retrospect Client software? How does Retrospect "see" this other computer?

 

We have no "client software". We installed Retrospect onto the Mac Pro, then we "talk" to the G4 via ftp. Retrospect "sees" the other computer via an ftp connection. Since we are doing neither of your questioned methods, I assume we aren't doing something right...

 

 

 

Is this a wired connection or a wireless connection?

 

We have tried it both ways. Currently, it is wireless using AirPort to the 100MB/sec router which is 2 feet away.

 

 

[cooch1811]

This is the place where some confusion begins. Do you have a copy of Retrospect (each with its own license) installed on each of these two computers?

 

Yes, we have a copy of Retrospect installed on each machine. I know we purchased a license, but I'd have to check if we purchased 2.

 

The answer to the question quoted above would have made this clear; is Retrospect running on the machine that's giving you unexpectedly slow performance? And is the Source for those backups the internal hard drive of that same machine?

 

Yes and yes.

 

 

[CallMeDave]

Upthread in Post # 103662 you answered Mayoff's question about the speed seen in another Type of Backup Set. Can you provide a complete description of what you did for that test, and what you saw when you did it?

 

It's critical for readers to have a clear understanding of exactly what you're doing and what you're seeing.

 

 

Dave

[cooch1811]

Upthread in Post # 103662 you answered Mayoff's question about the speed seen in another Type of Backup Set. Can you provide a complete description of what you did for that test, and what you saw when you did it?

 

It's critical for readers to have a clear understanding of exactly what you're doing and what you're seeing.

 

With all due thanks and respect, I have laid out everything I can. I have no more info to offer

 

The "tests" are nothing more than EasyScript backups over our network using the internet option.

 

"What we are seeing" is simply 50MB/minute source speeds on a dual quad core Mac Pro machine.

 

There is nothing special going on: just one Mac internal hard drive talking to another using Retrospect's standard ftp setup.

 

No external drives; no client/server; no filesharing; just an unacceptable situation.

 

 

[Mayoff]

For me, I am still confused about the configuration, so it is hard for me to offer suggestions. Maybe you can take some screenshots of the backup script and Configure>volumes and attach them to the forum post? That will help us understand how you have this configured.

 

Do you get different performance when using a file backup set instead of an FTP backup set? What about writing to a drive directly plugged into the backup computer?

[rhwalker]

I'll offer my two cents and then get out of the way because I really don't have any suggestions either.

 

It appears that the configuration is backwards from the standard Retrospect installation. The standard installation would have Retrospect running on one machine and Retrospect client running on the machines to be backed up, in a "pull" configuration for the backup (Retrospect pulls the files from the clients).

 

Here, there appear to be multiple copies of Retrospect floating about, one on each machine to be backed up, and the machines "push" their backups onto a local FTP server over a wireless connection through an Airport router. Steve, correct me if my understanding of your configuration is wrong.

 

Retrospect's client/server model is optimized for UDP throughput. Here, we have three added variables: (1) FTP, not UDP; (2) wireless through an Airport router, which may have its own throughput issues and windowing issues for the protocol; and (3) the FTP destination server, which may have its own issues through the Airport and with Retrospect.

 

To me, the best first test would be the one suggested by (CallMe) Dave, namely, to evaluate backup performance to a local disk file backup set.

 

A good second test might be to evaluate throughput with Retrospect's client/server model.

 

Russ

[cooch1811]

It appears that the configuration is backwards from the standard Retrospect installation. The standard installation would have Retrospect running on one machine and Retrospect client running on the machines to be backed up, in a "pull" configuration for the backup (Retrospect pulls the files from the clients).

 

Thanks! This is helpful

 

Here, there appear to be multiple copies of Retrospect floating about, one on each machine to be backed up, and the machines "push" their backups onto a local FTP server over a wireless connection through an Airport router. Steve, correct me if my understanding of your configuration is wrong.

 

You are basically correct! However, one machine is wireless and one is hardwired ethernet. The wireless one has been tested hardwired as well, with similar slow results.

 

 

Retrospect's client/server model is optimized for UDP throughput. Here, we have three added variables: (1) FTP, not UDP; (2) wireless through an Airport router, which may have its own throughput issues and windowing issues for the protocol; and (3) the FTP destination server, which may have its own issues through the Airport and with Retrospect.

 

Some of these may be at issue, or maybe a combination. However, I suspect something else is thrown in here, because we have had faster backups using identical setups. The only other variable that I haven't mentioned is that the slow computer uses Leopard, versus Tiger for the others...

 

 

 

To me, the best first test would be the one suggested by (CallMe) Dave, namely, to evaluate backup performance to a local disk file backup set.

 

A good second test might be to evaluate throughput with Retrospect's client/server model.

 

I will take a stab at both of these, and let you know

 

 

Thanks all.

[CallMeDave]

I have laid out everything I can...

 

Perhaps. But in the course of doing so, you have provided either conflicting, or simply inaccurate information.

 

In post # 103593 you were asked by Mayoff "What happens if you write the data to the network server using a file backup set instead of FTP? What is the speed like?"

 

To which you answered "Same slow speeds" and even added a sad face.

 

But now, you state unequivocally "The "tests" are nothing more than EasyScript backups over our network using the internet option.... just one Mac internal hard drive talking to another using Retrospect's standard ftp setup."

 

So I hope you can forgive me for insisting on accuracy and consistency. The head of EMCInsignia tech support has asked you a specific question, twice. The answer you gave the first time was contradicted by a later post, and the second time was simply ignored.

 

Yes, you should be able to have Retrospect installed on as many machines as you have licenses for, and yes, you should be able to "push" the data via ftp to a central ftp server (even though Internet Backup Sets are somewhat outdated, and work better if the program is hacked to make the minimum segment size larger then the hard-coded 16 MB that it is now), and yes, the client/server works better but is not the only option.

 

We hear that what had been a working configuration no longer works as expected, although multiple things are different with the new setup (different CPU architecture, different OS version).

 

Confirming that a File Backup Set stored on the same computer that's currently acting as an ftp server, that is accessed via AFP from the machine running Retrospect would be a good test. As Mayoff suggested, knowing if this happens _only_ on ftp communication or if this happens when Retrospect communicates using other protocols would be very helpful to know. It would also help to confirm if ftp transfers using other software, such as CyberDuck or Transmit, are fast or slow.

 

Dave

[rhwalker]

Steve, this piece of information that it is only slow on Leopard adds a new dimension; may be some issue in the Leopard networking code. I know that some people have been having throughput issues with Leopard and wireless.

 

I still suggest that you perform the tests that Dave and Robin have suggested (local backup) to identify whether the problem is with networking or Retrospect. If it's with the networking, well, there is little that Retrospect can do unless you are going to switch to the client/server model that is optimum for Retrospect's network operations.

 

If you are still wanting to stay with your current unique FTP configuration, you might want to increase the FTP segment size, which might improve throughput and also will increase the backup set size. It's a hack. Details in this thread here:

 

ResEdit hack to Retrospect

 

Russ

[CallMeDave]

...this piece of information that it is only slow on Leopard adds a new dimension...

 

Russ, I'm surprised that you took this at face value.

 

As I read the thread, there are _two_ unique attributes to the machine that is slow. The first is that it's a MacPro, running Retrospect under Rosetta emulation. And the second is that it's running Mac OS X 10.5.

 

Either of these, or both, or neither, might be the cause of the slow transfers.

[rhwalker]

As I read the thread, there are _two_ unique attributes to the machine that is slow. The first is that it's a MacPro, running Retrospect under Rosetta emulation. And the second is that it's running Mac OS X 10.5.

 

Either of these, or both, or neither, might be the cause of the slow transfers.

And nothing I said disagreed with this. My carefully-worded point was that another possibility had been added to the mix. Your and Robin's test case is a needed first step, and no results have been provided.

 

Russ

[CallMeDave]

I suppose that while we await additional information, we can amuse ourselves by bickering about how we're each trying to be helpful in our own unique way.

 

I don't believe that your point was carefully-worded. When you wrote: "it is only slow on Leopard" you implied that it is, in fact, only slow on Leopard. But we don't know that, we only know that it's slow on this machine, and that this machine happens to be running Leopard.

 

If I were college educated I'd probably know the fancy Latin name for this logical fallacy. Maybe I can find it on the internets before I run out of Ch33toz.

 

 

Patiently,

 

Dave

[cooch1811]

So I hope you can forgive me for insisting on accuracy and consistency. The head of EMCInsignia tech support has asked you a specific question, twice. The answer you gave the first time was contradicted by a later post, and the second time was simply ignored.

 

Please excuse me for not having a phd in network terminology, and for not genuflecting to the higher orders on this forum. I mentioned several times that we are simple people here, and since I am confusing you with my ineptitude, I will not return.

 

Please delete my profile.