Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

My problem is that when I transfer files to or from 2 of my ESXi 4.1 servers using Veeam's FastSCP utility, the transfer seems to be limited to about 5 MBs when I can transfer the same file from the same source to another ESXi server on the same switch at 60 MBs. Just to be clear, this is just transferring to the data store and not to a VM.

I have 3 independent ESXi servers, each running ESXi 4.1 installed from the same media (using the HP verision of the ESXi 4.1 distributable). 2 of the machines exhibit this slow transfer rate while the other transfers at a normal speed (to and from my desktop is 60 MBs average).

All three servers are HP DL380 G5 servers with the following differences

  • Working
    • PCI Express Intel NC364T Gigabit (4 port)
    • 2 4-core Intel E5450
  • Under-Performing
    • Integrated Broadcom NC373i Gigabit (2 ports)
    • PCI Express Intel NC360T Gigabit (2 ports)
    • 2 4-core Intel E5440

Steps taken

  • Installed latest firmware from HP
  • Carefully compared ESXi Configuration for differences
  • switched my transfer between each of the 4 network connections, all 4 behave the same.
  • confirmed that the physical switch ports are healthy and that the cables are good

Anyone have any suggestions about what the problem could be or something to investigate?

share|improve this question
    
What type of storage you have? –  Vick Vega Apr 12 '11 at 23:43
    
Are the NICs teamed in any way? –  Holocryptic Apr 13 '11 at 0:05
    
Vick: Both the fast and the slow servers have the HP Smart Array P400 controller with local storage of 6 10k drives in RAID 5 –  Dennis Allen Apr 13 '11 at 13:32
    
Holocryptic: The switch has not been configured for link aggregation yet, so for the file transfer they are effectively independent 1 Gigabit full-duplex connections. –  Dennis Allen Apr 13 '11 at 13:34
    
Vick: The difference that I did not account for in the storage controller is that the fast server has a 512MB BBWC kit installed and is configured for 50/50 read/write acceleration. –  Dennis Allen Apr 14 '11 at 3:10
add comment

3 Answers

It turns out that the Storage Adapter was exhibiting consistent write latency of 25 ms or more, even when the system was idle (no file transfers going on). This is a known issue with the HP Smart Array P400 when you do not have the BBWC (batery backedup write cache) module installed and write acceleration enabled.

Using the array configuration utility you can turn on write buffering, which did help some at the trade off of a slight risk of data loss. I have a 2 hr UPS in addition to a backup generator so I am not too worried about power loss.

Thanks for all your help

share|improve this answer
add comment

HP has a new advisory related to this.

http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?objectID=c01832427&jumpid=em_alerts_us-us_Jun11_xbu_all_all_1302806_87735_optionsandaccessoriesbladesystemproliantserverssoftware_critical_006_1

"HP recommends that customers purchase an optional Battery Backed Write Cache (BBWC) Module to achieve optimal disk subsystem performance on ProLiant G6 and G7-series servers running VMware ESXi 4.0 when the configuration includes the following Smart Array controllers:

Smart Array P410i Controllers Running VMware ESX 4.0
Smart Array P410 Controllers Running VMware ESX 4.0
Smart Array P411 Controllers Running VMware ESX 4.0
Smart Array P212 Controllers Running VMware ESX 4.0
Smart Array P712m Controllers Running VMware ESX 4.0
Note: This issue may affect other HP controllers not listed in this advisory that do not include Battery Backed Write Cache.

DETAILS Certain Virtual Machine (VM) operations such as cloning and powering on VMs may significantly impact disk subsystem performance on the VMs configured with the above mentioned Smart Array controllers that do not include Battery Backed Write Cache.

Note: When the Battery Backed Write Cache is not included in the configuration, even moderate disk I/O can negatively impact server performance."

share|improve this answer
add comment

vmfs fragmented? Was the formatting done differently, block/cluster allighment? Do you have other machines running on the host? Have you updated the firmware/BIOS on the controller and or array bios.

share|improve this answer
    
I have no VMs running on the slow host, and there are no files in the data store. The fast host has 5 VMs running on it. As I put in the steps taken I ran the latest hardware update from HP which did upgrade the firmware and BIOS and the controller and the NIC. –  Dennis Allen Apr 13 '11 at 15:23
    
I built the array using the setup from the controller post, so there are no block configuration options. Perhaps the HP SmartStart utility would give more information? –  Dennis Allen Apr 13 '11 at 17:25
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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