1

I have similiary situation like in question Increase transfer rate with Link Agreggation

But with some different purposes - need increas backup speed. ServerA - backup media server (Win2008r2); ServerF - file share (Win2012); and switch (24x 1Gbit support LACP, Link Agg etc)

Both of servers has two 1Gbit network adapters.

Problem - backup task runs too many hours (more 3TBs in each task). After perfome configuration LACP on both servers and switch ports - upset: copy file between hosts - speed not growth upper 113megabytes\sec. I try measure if iperf - same result. One positive thing - while different clients copy files each use approximate 1Gbit (140-150megabytes together).

After seaching on web - found same discussion were explaned aggregation speed behavior: no speed increase for one stream (IP > IP).

Maybe some other protocols can help me speedup backup jobs tasks ?

UPDATE 1: In this brochure (page 9) https://blog.pingus.de/wp-content/uploads/2016/11/Backup-Exec-Best-Practices-Guide-v.4.0.pdf comparsion table consist mark about teaming with 200MB\s speed. Video https://www.youtube.com/watch?v=O7V0qz6Q1mw

3
  • 1
    Unfortunately this is the situation. If you are simply doing a transfer from one host to another host in a single stream, only one of the links will be utilized so you will not go above 1Gb. You will need to increase the 1Gb to something like 10Gb to see greater performance. Alternatively, break up the source or destination to multiple servers so that more than one stream is utilized over your LAG.
    – SleepyMan
    Feb 15 '17 at 16:18
  • Regarding backup exec, you can see in the video (in Windows task manager, at 2"50) that the traffic doesn't actually go over 2.4Gb/s over a 10Gb/s interface. The 14K MB shown in job rate involve compression and / or block level differential backup.
    – JFL
    Feb 16 '17 at 10:30
  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you could provide and accept your own answer.
    – Ron Maupin
    Aug 16 '17 at 22:11
1

Link aggregation increases the bandwidth for multiple flows. A backup job (single flow) will remain the same at what a single link can do because each flow only uses a single link. Multiple flows use multiple links, and the bandwidth is higher on average across the multiple flows.

Really, the only thing you can do is to use a faster network link. You probably want a dedicated 10 Gb, or faster, link between the production and backup server.

3
  • What about this (page 9) blog.pingus.de/wp-content/uploads/2016/11/…
    – JJbaz
    Feb 16 '17 at 9:51
  • You can keep wishing and hoping, but you have three different answers here, all telling you the same thing. If you want to increase the bandwidth, you really need to get a faster link. You could have a dedicated 10 Gbps link between the production and backup servers, and that would also take a burden off the production network. Separating storage on its own network is often practiced because bulk-type transfers need to be a low priority on a production network, but not on a dedicated storage network.
    – Ron Maupin
    Feb 16 '17 at 14:21
  • right now i do some test. Want compare backup speed (in work hours period) by dedicated 1Gbit link (one link for fileshare users, second only for backup stream) or via teamed 2x 1Gbit. Simulate ~1Gbit traffic users-stream and start backup job.
    – JJbaz
    Feb 17 '17 at 8:37
1

Link aggregation works under specific scenerio:

If the switch has 1 host connected to it with ingress traffic at 100 mbps. Now at the egress port has 5 links connected through etherchannel capable of providing 1gbps. You will still find your speed stuck at 100 because the ma incoming traffic is 100 mbps.

Secondly, load balancing takes place based on several parameters, if the parameter is source ip and you have just one host, in your case while copying files, you will not see significant increase because the source of all packets is the same.

Hope this helps.

1

As explained by @john and @Ron Maupin, link aggregation is so that a single flow will always use the same physical link.

There's no standard network protocol that will help you in your scenario.

Regarding the backup exec documentation and video

  • the PDF mention teaming, but this is useful when running concurrent backups toward the same device
  • the video clearly show that the traffic doesn't go higher than 2.4Gbs.

First you have to make sure that the network is the culprit. Quite often the limiting factor is the disk subsystem rather than the network.
Depending of the software and feature used (compression...) the CPU may also limit the transfer speed. This is valid for both sides of the transfer.

Outside of network itself, you may change your backup strategy. Weekly full backup with daily incremental backup, compression, block-level differential backups are some way you can investigate, but all of this is for serverfault rather than network engineering.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.