Infiniband/MT25208/10u7 throughput
Heya all,
Working on setting up an HPC cluster of sorts. Running into some funky issues with throughput of Infiniband on Solaris. I'm unable to get any sort of reasonable speed out of the cards; throughput seems limited to around 60 MB/s. Tested using FTP and also the netio package I stumbled across here
Currently using a pair of t5220's for testing, each with a Mellanox MHGA28-XTC card. The cards have the latest firmware installed, and according to Mellanox, use the MT25208 chipset. They show up under prntconf as 'pciex15b3,6282'. The t5220s are running Solaris 10u7. The Infiniband fabric is running over a Voltaire 9024 switch.
Noticed when I installed the cards were not being recognized. The Infiniband Update 3 solved this; but I noticed it's apparently meant for 10u6? Not sure if that's causing problems or not. Has anyone else used IU3 on 10u7?
Are there TCP settings I need to tweak/etc.? I'm thinking that might be it, but I'm having a hard time finding documentation about Infiniband on Solaris. :)
I've also tested the cards in a pair of x86 boxes running SuSE Enterprise Linux 11; there, transfers were up in the range of 300MB/s and were being bottlenecked by disk speeds. So I'm fairly certain the cards, cables and switch are fine.
Dear Sir
Problem solved?
I want you to teach if you know something, because I am newbei Solaris user.
Perhaps, I have the same problem as yours. (But hardware environment is totally different.)
I examined Voltaire400Ex, 410Ex, and 600Ex.
Throughput is all low.
I tried in this environment.
# Dell R610
# HCA : Voltaire 410Ex-D (Mellanox MT25204)
# Switch : Voltaire 9024D
Now, I have 410Ex only .
Therefore, I am using 410.
(1) Solaris 10 5/09 s10x_u7wos_08 X86 ( ib_updates_3d_s10u7)
#1. netperf(v2.4.5)
+5485.80(10^6bits/sec)+
#2. netio(v1.26)
+713309 KByte/s+
(2) CentOS 5.3 (2.6.18-128.el5) ( OFED-1.4.1)
#1. netperf(v2.4.5)
+10819.86(10^6bits/sec)+
#2. netio(v1.26)
+1220885 KByte/s+
On solaris, I was able to improve the performance by tuning . (a little)
However, it is still slow.
Can I improve it more?
I think that Solaris device driver has the problem.
Similar Messages
-
Infiniband bad outgoing Throughput on 10Gbit HCA
Hi @ all,
I have a problem. I would like to use Solaris with ZFS to provide storage for a glusterfs server with nfs.
My test environment:
Node 1: CentOS 6.2 with OFED 1.5.4.1
Node 2: OI 151a4 with native IB and before a Solaris 11 at both Solaris and OpenIndiana have the same result.
If I run a test with iperf:
From CentOS to OI throughput around 4.90Gbit/s
[root@dev-cos62 ~]# iperf -c 1.1.1.2
Client connecting to 1.1.1.2, TCP port 5001
TCP window size: 193 KByte (default)
[ 3] local 1.1.1.1 port 36173 connected with 1.1.1.2 port 5001
[ ID] Interval Transfer Bandwidth
[ 3] 0.0-10.0 sec 5.66 GBytes 4.86 Gbits/sec
From OI to CentOS throughput only 900Mbit/s
Croot@dev-oi:~# iperf -c 1.1.1.1
Client connecting to 1.1.1.1, TCP port 5001
TCP window size: 256 KByte (default)
[ 3] local 1.1.1.2 port 35841 connected with 1.1.1.1 port 5001
[ ID] Interval Transfer Bandwidth
[ 3] 0.0-10.0 sec 1.13 GBytes 968 Mbits/sec
My IB Hardware is a new Mellanox InfiniScale switch and some older 10Gbit Mellanox HCAs. (MTLP23108)
A second dd test with a ramdisk shared over nfs tel me the maximum of solaris/oi infiniband outgoing throughput "write" is max 1Gbit.
Have anybody any idea?
thx and many greets from germany
Andreas
Edited by: 942419 on 23.06.2012 17:47Do you see this problem with S11 FCS too?
In any case, you could try increasing the ndd /dev/tcp setting
'tcp_naglim_def' from 4K to 64K and see if it helps.
#ndd /dev/tcp tcp_naglim_def 65535 -
Using InfiniBand as interconnect
Is anyone done or know someone has RDS ov IB or IP ov IB implementation as interconnect in RAC environment? If so please give some feedback and thanks.
Hi,
Infiniband becomes more and more popular. I recently worked with some customers that use Infiniband as their network infrastructure and so also for the RAC interconnect. Voltaire seems to be an upcomming vendor in this market (this is not a stock recommendation !), but Cisco is very active, too, of course.
Tests have shown that IB in general provides better throughput. Using the RDS protocol can speed up the communication in addition:
http://www.oracle.com/corporate/press/2006_oct/openworldsf06-18.html?rssid=rss_ocom_pr
and:
http://www.oracle.com/technology/deploy/availability/pdf/oow06/S281216_Tsien.pdf
Please, let me know, if you are looking for something in particular.
Thanks. -
Coherence and myrinet/infiniband?
We have some networks that run TCP/IP over myrinet and infiniband. Has anybody tried using this kind of nets with Coherence and in that case what was the experience performance as well as reliability wise? What configuration changes did you make in Coherence to get
maximum perfromance?
Best Regards
MagnusHi Magnus,
In some small scale performance testing on IB, I was able to achieve throughputs of around 600MB/s between two IB connected Coherence JVMs. In order to get rates this high though I needed to use jumbo frames of max size (16KB I think) and correspondingly big data objects.
Mark | Oracle Coherence -
AP1231G-A-K9 access points - very slow throughput - Is TKIP the issue?
I recently setup our small office network using the following setup:
Cablemodem <--> router <--> 1231AP(role root bridge with wireless clients) <-> 1231AP(role non-root bridge with wireless clients)
Code on both APs: 12.3(8)JEE
Office network generally has less than 3 wireless clients connected at any one time to either AP.
AP's are a mere 50' apart; clients are all less than 30' from either AP; they all show excellent signal and connected at 54mbps signaling rates.
All is/has been working very well & very stable with the exception of speed. We have business class service from RR, approx 25mbps dl, 2mbps ul. Any hardwired client to the router switch ports are able to download at speeds averaging 23mbps. Any wireless client connected to either AP is never able to exceed download speeds of 5mbps. With no other wireless clients connected except my one test client, I was not able to exceed 5mbps throughput from either AP that I connected to.
I can confirm that the ethernet connection between the router and root bridge is up at 100mbps-FD and not showing any errors:
ap#sh interfaces FastEthernet0
FastEthernet0 is up, line protocol is up
Hardware is PowerPC405GP Ethernet, address is 0013.60cf.bb29 (bia 0013.60cf.bb29)
MTU 1500 bytes, BW 100000 Kbit, DLY 100 usec,
reliability 255/255, txload 1/255, rxload 1/255
Encapsulation ARPA, loopback not set
Full-duplex, 100Mb/s, MII
ARP type: ARPA, ARP Timeout 04:00:00
Last input 00:00:00, output 00:00:00, output hang never
Last clearing of "show interface" counters never
Input queue: 0/160/0/0 (size/max/drops/flushes); Total output drops: 0
Queueing strategy: fifo
Output queue: 0/40 (size/max)
5 minute input rate 5000 bits/sec, 0 packets/sec
5 minute output rate 1000 bits/sec, 1 packets/sec
8054605 packets input, 3141009145 bytes
Received 46005 broadcasts, 0 runts, 0 giants, 0 throttles
0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored
0 watchdog
0 input packets with dribble condition detected
4076106 packets output, 411952731 bytes, 0 underruns
0 output errors, 0 collisions, 4 interface resets
0 babbles, 0 late collision, 0 deferred
0 lost carrier, 0 no carrier
0 output buffer failures, 0 output buffers swapped out
Wandering thru the cli on either AP shows that all wireless clients are indeed connected at 54mbps to their respective AP and the two AP's are connected happily at 54mbps signaling:
Address : 0013.1a37.b3e0 Name : ap
IP Address : 192.168.0.120 Interface : Dot11Radio 0
Device : 11g-bridge Software Version : 12.3
CCX Version : NONE
State : Assoc Parent : Our Parent
SSID : Tsunami
VLAN : 0
Hops to Infra : 0 Association Id : 44
Tunnel Address : 0.0.0.0
Key Mgmt type : WPA PSK Encryption : TKIP
Current Rate : 54.0 Capability : WMM ShortHdr ShortSlot
Supported Rates : 1.0 2.0 5.5 6.0 9.0 11.0 12.0 18.0 24.0 36.0 48.0 54.0
Voice Rates : disabled
Signal Strength : -51 dBm Connected for : 75169 seconds
Signal to Noise : 26 dB Activity Timeout : 14 seconds
Power-save : Off Last Activity : 1 seconds ago
Apsd DE AC(s) : NONE
Packets Input : 1050695 Packets Output : 296536
Bytes Input : 474651248 Bytes Output : 96734573
Duplicates Rcvd : 0 Data Retries : 63646
Decrypt Failed : 0 RTS Retries : 0
MIC Failed : 0 MIC Missing : 0
Packets Redirected: 0 Redirect Filtered: 0
Here is a config snippet from the AP non-root bridge with wireless clients:
dot11 ssid Tsunami
authentication open
authentication key-management wpa
guest-mode
infrastructure-ssid optional
wpa-psk ascii 7 (snipped)
bridge irb
interface Dot11Radio0
no ip address
no ip route-cache
encryption mode ciphers tkip
ssid Tsunami
speed basic-1.0 2.0 5.5 6.0 9.0 11.0 12.0 18.0 24.0 36.0 48.0 54.0
station-role non-root bridge wireless-clients
bridge-group 1
bridge-group 1 subscriber-loop-control
bridge-group 1 spanning-disabled
interface FastEthernet0
no ip address
no ip route-cache
duplex auto
speed auto
bridge-group 1
bridge-group 1 spanning-disabled
interface BVI1
ip address dhcp client-id FastEthernet0
no ip route-cache
bridge 1 route ip
(The AP root-bridge with wireless clients config is identical to this config with the exception of the station-role and a static IP on the BVI1 interface.)
Are these very slow thoughput speeds normal of this hardware combination?
I did much searching/googling and found claims that by eliminating TKIP it almost doubles the actual wireless speeds our clients can obtain. Is there any truth to this?
Any suggestions or recommendations without changing hardware would be very welcome.
Thanks in Advance!
D.
=============Ok, thanks for the explanation - I understand. But even at a 22mbps signaling rate shouldn't I be seeing throughputs greater than 5-5.5mbps especially since this location is literally 100% free of any outside interference and the interfaces definitely show the clients and non-root bridge (when connected) all being at the highest rate of 54mbps? I tried even in the same room, approximately 40' away, total line of sight, no obstructions, between my laptop and the root AP.
I disconnected the non-root bridge and connected directly to the root bridge during my testing. I was still only able to achieve approx 5.5mbps download. Adding back in the non-root bridge and re-connecting to it I notice slightly lower throughput, approx 5mbps. During testing, my laptop was the only device connected to the network, all other clients were shut off.
Here are the int stats (I've never reset the counters):
Root Bridge:
RATE 1.0 Mbps
Rx Packets: 2178725 / 49 Tx Packets: 39 / 0
Rx Bytes: 335124036 /7595 Tx Bytes: 4965 / 0
RTS Retries: 61 / 0 Data Retries: 5 / 0
Non-Root-Bridge:
RATE 1.0 Mbps
Rx Packets: 2323120 / 50 Tx Packets: 141 / 0
Rx Bytes: 336455923 /7595 Tx Bytes: 17869 / 0
RTS Retries: 2 / 0 Data Retries: 56 / 0
All the other rates, 2-12mbps show single or double digit packet/byte counts until I get to the 36mbps section of each interface:
ap#sh int Dot11Radio0 stati
DOT11 Statistics (Cumulative Total/Last 5 Seconds):
(snipped for brevity)
Root Bridge:
RATE 36.0 Mbps
Rx Packets: 915395 / 1 Tx Packets: 2345589 / 9
Rx Bytes: 93420936 / 70 Tx Bytes: 3370791285 / 874
RTS Retries: 0 / 0 Data Retries: 573981 / 4
RATE 48.0 Mbps
Rx Packets: 2163192 / 2 Tx Packets: 216861 / 0
Rx Bytes: 222455730 / 404 Tx Bytes: 182817967 / 0
RTS Retries: 0 / 0 Data Retries: 106808 / 0
RATE 54.0 Mbps
Rx Packets: 987986 / 0 Tx Packets: 168923 / 0
Rx Bytes: 190467269 / 0 Tx Bytes: 61665042 / 0
RTS Retries: 0 / 0 Data Retries: 34424 / 0
Non-Root Bridge:
RATE 36.0 Mbps
Rx Packets: 2368679 / 0 Tx Packets: 965419 / 0
Rx Bytes: 3396819830 / 0 Tx Bytes: 90880825 / 0
RTS Retries: 0 / 0 Data Retries: 242686 / 0
RATE 48.0 Mbps
Rx Packets: 341870 / 0 Tx Packets: 2156282 / 1
Rx Bytes: 216497093 / 0 Tx Bytes: 215775536 / 210
RTS Retries: 0 / 0 Data Retries: 478619 / 0
RATE 54.0 Mbps
Rx Packets: 1469926 / 15 Tx Packets: 2529678 / 15
Rx Bytes: 411722698 /1122 Tx Bytes: 1366306113 /5159
RTS Retries: 0 / 0 Data Retries: 198532 / 0
I will try disabling the rates below 12mbps and re-test.
I would like to try disabling all encryption and try as well.
Do you know if the AP's will associate if there is zero encryption? -
A question very similar to mine exists here.
I have a SOFS cluster (3 hosts). I connected each without nic teaming at first and later tested with nic teaming. I'm using a single 10GbE Netgear M7100-24X switch. The CSV is configured as a 2-way mirror through storage space using a SAS JBOD with
24 disks. Each host is configured the same way with 32 GB of RAM. 6 GB is set for CSV cache.
I ran ntttcp test (v5.28) with 8 threads. Sending to the SOFS host, I get over 1100 MB/s throughput. Receiving from the SOFS, I get just under 680 MB/s throughput. So the switch looks to be working fine.
When using LAN Speed Test (Lite), connections directly to the file share folders (\\host#\c$\ClusterStorage\Volume1\Shares\folder) for a 200MB file for each server averages to about 700 Mbps write and 2000 Mbps read. Connection to the cluster role (\\sofs\folder),
results in 90 Mbps write and 2000 Mbps read. However, after waiting for a minute for it to start running, the speed test starts and pauses repetitively. I know this doesn't mean much because it isn't testing transfers from SMB to SMB.
Since I can't set up another SMB to test SMB to SMB transfer, I'm jumping straight to Hyper-V. In VMM, I added SOFS file share folder to an existing vm cluster. After that, I migrated a vm to one of the hosts in the cluster with high availability checked
and saw that it indeed used the \\sofs\folder.
Using LAN Speed Test (Lite) on that vm and back to that particular vm host, I'm getting under 90 Mbps write and 340 Mbps read. If you recall the earlier results directly to \\sofs\folder, the write speed is similar to just regular file transfer speed, but
the read speed is 6 times lower. Sending with ntttcp, I'm getting an average of 11 MB/s throughput, which does explain the 90 Mbps write. And receiving from the host, I'm getting an average of 42 MB/s throughput, which also explains the 340 Mbps read speed. But
another vm hosted by the same server without SOFS is giving me 350 MB/s sending and 360 MB/s receiving to and from that host respectively. Although way faster, this does seem a little bit slower. I then ran Passmark Network Test to be thorough. Max speed
of the vm using SOFS is 100 Mbps sending and 330 Mbps receiving. The vm without SOFS is 7500 Mbps sending and 6000 Mbps receiving. I don't know why ntttcp differs from Passmark this much. (Maybe ntttcp not as optimized for 10GbE?)
But disregarding discrepancies on the results for the vm without SOFS, it is still clear that the vm with SOFS as storage is way slower. To rule out nic teaming as the solution to my problems, I've set up nic teaming (switch independent and dynamic) to all
the SOFS hosts. I didn't get much difference in the results. As I do not have another switch, I don't think nic teaming helped with the load balancing. And I haven't set up link aggregation (MLAG) on the switch either.
Is this the speeds that I should be getting, or are there other optimizations or configurations you can suggest? I'll be honest, a single vm on SOFS doesn't lag very much if at all despite its awful throughput I'm currently getting. What I'm scared of is
if I put 50 vms and have SQL Server run off the SOFS.I'm using normal 10 GbE NIC with just RSS; I should have mentioned that I'm only using Intel X520-T2 nics.
As for getting 7500 Mbps throughput on the VM host without using SOFS, let me clarify. On host A, I have two virtual machines. One VM is set up to use SOFS and HA, the other one is just on host A itself. The VM on SOFS is giving me 90
Mbps write and 340 Mbps read. And the VM stored on the host itself is giving me 7500 Mbps write
and 6000 Mbps reading. BTW, typically I get better read speeds than writes, it might be the day and hour. Earlier today when I reran the benchmarks, it was 7300/7500 Mbps write/read. The VM with the SOFS is still the same however.
I was told before that we don't need RDMA at a Microsoft conference, but now I think it's only true for a lab environments... The company I work for does not have
the budget to buy RNIC and SFP+ switches for now.
I'm going to try to implement these
solutions first. I currently only using a single VNIC. Give me a day or two.
What I'm currently wondering about is that the DNS servers on the network are not on the 10 GbE network. I'm wondering if the data is staying on the 10 GbE network and not going out to the 1 GbE network first. This is a total guess. -
Hard drive performance and data throughput
I am using my macbook pro for work primarily and part of that entails creating/restoring images of other macs. I've had the best luck with SuperDuper however the process is still VERY slow. For instance at this moment with no other applications open other than S.D. and firefox the copy speed is under 5MB/s from my MBP to an iMAC via fire wire.
I am looking for suggestions to increase the performance/IO in the hopes to speed up the process. When purchasing this system the 7200rpm drive was not an option (15") which is unfortunate. I realize that both hard drives in the operation will cause the variable in speed but I want the sending drive as fast as possible.
My thoughts right now are to purchase a 7200rpm external drive to store backup images and also send from. This would cut out any possible IO on the drive that my mac is performing to run the operating system. Another thought was to upgrade my mac to a 7200rpm drive and use the current 5400rpm drive as the storage for images...in the hopes that it would still provide an increase in restoration speed since it wouldn't be running OSX on it.
Any thoughts or ideas? Experiences? My MBP has the 5400rpm I believe and 2GB of ram.
ThanksI'll try and explain a bit better. I'm not restoring
the same image to different types of macs. I create
images of OTHER macs using my macbook pro to perform
the process as well as store the backup image.
Thanks for the clarification. I do that too, but when I do I use my Mac Pro to clone a Mac via Target Disk Mode to an external FireWire 800 drive.
it helps but its a usb2 enclosure with a somewhat
older hard drive that is only 30Gb. I am looking at
purchasing a firewire 800 external drive but I will
see how this other unit works for now since we
already have it.
Part of your throughput problem may be the overhead issues with USB 2. FireWire uses its own chipset so is more independent of the CPU, and FireWire can sustain high-speed transfers at a higher level. USB is CPU-bound and is more vulnerable to CPU demands from other apps or background processes or other USB devices. So even though USB 2 has a higher theoretical peak (480Mbps), FireWire (400Mbps) actually does better in the real world.
About USB 2 vs. FireWire 400 performance
I'm not sure if FireWire 800 would help because your slowest drive in the chain may not be fast enough to take advantage. -
Poor 2.4 GHz Wi-Fi Throughput (Hub 5)
Hello,
Even though Windows 7 reports my Wi-Fi signal strength as 'Excellent', as does inSSIDer (-25dBM), the throughput itself is very poor.
Here is a LAN test to begin with just to prove the broadband is okay (37.50 Down | 9.49 Up):
http://www.speedtest.net/result/3435996049.png
Here is a Wi-Fi test, using the same server 1 minute later (6.85 Down | 7.74 Up):
http://www.speedtest.net/result/3435997528.png
Nexus 7 Tablet Wi-Fi (18.30 Down | 9.35 Up):
http://www.speedtest.net/my-result/a/794982309
Xperia S Smartphone Wi-Fi (12.33 Down | 9.25 Up):
http://www.speedtest.net/my-result/a/794985718
No matter what I try, I cannot get the rock solid throughput I had with my HH3 :-(
Things I have tried:
Changing Wireless interface type from 150Mb/s to 300Mb/s and then back again.
Manually selecting a non-overlapping channel, i.e., 1, 6 or 11.
Find less crowded channels using inSSIDer and changing it to one of those.
Disabling 5GHz wireless.
Resetting wireless settings to default.
Complete resetting to factory defaults.
Turning off all nearby electrical devices.
Moving the router.
Standing on 1 leg and pulling a silly face.
Any help on this matter is much apreciated.
Paul.Update:
After 2 hours on the phone (yes, 2 hours!), I have a new Hub5 arriving tomorrow and will report back if that one is the same as the first.
I'll try to make this long story short. Here goes...
Phoned BT faults approx 3:30pm and spoke to a woman, possibly off-shore going by the accent. Explained that the wired connection was rock solid at over 37Mb/s but 2.4GHz wireless was very poor at around 6Mb/s.
When she got to the, "Have you got another microfilter you can try" bit, I suspected she wasn't altogether aware of the problem I had.
Asked to speak to a supervisor who tried telling me nothing was wrong with the hub. I asked for a replacement as I suspected it could be faulty. After he tried fobbing me off with pompous talk and refusing to send a new one because it 'was not faulty', I asked to cancel the contract as I was not happy with aspects of the package I had just signed up to and I believe I was still within the 'cooling-off' period.
He put me on hold for 25 minutes while he spoke with sales to get a new one sent out.
Call was cut off!!
He rang me back and put me through to sales who were unaware of his apparent 25-minute conversation with them. Sales were unable to send out a new Hub5 as their systems were unable to do so and they were forced to put me back through to faults.
Spoke to another patronising lady and I gave her an ultimatum. 1. Send a new Hub5 or 2. Cancel my Infinity. She confirmed a replacement will be with me tomorrow!
2 hours on the phone to get that sorted and I'm fairly IT savvy. I can only cringe at how some less tech-minded people get on. -
Extremely low message throughput with MQ 3.6
We performed some load tests to determine maximum throughput of the Sun JMS framework when used in conditions similar to our application.
The achieved throughput is extremely low: around 10 messages / second while we expected around 100. Could you please verify what may be the reason?
We encountered one particular problem: the time of closing JMS producers increases much during the test. Message throughput decreases proportionally. Time periods of other phases of the JMS API usage remain constant. What can be the reason?
detailed information:
Our configuration used:
server machine:
cpu: Intel Celeron 2.8 Ghz
ram: 1GB
os: CentOS release 4.4 (Final), 2.6.9-42.0.3.EL
application server: Sun Java System Application Server Enterprise Edition 8.1 2005Q2 UR2
java version: 1.5.0_04
we use default imqbroker configuration
client machine:
Intel Celeron 2Ghz
ram: 512MB
os: Fedora Core release 3, 2.6.12-1.1381_FC3
java version: 1.4.02
Test description (find attached test unit sources presenting our way of using JMS API):
Load test is performed by running a number of concurrent test units against the JMS broker. Number of units is constant in time. Additionally, every unit test at the end of its life cycle launches another unit to keep long test time perspective.
Each test unit is self-contained. It contains of a producer and a consumer (MessageListener). It sends messages to itself . Also each message unit sleeps for some time to simulate message processing.
Messages are non-persistent, no durable subscriptions are used. All units are using one shared Queue for whole messaging. Message selectors are utilized to guarantee that messages are delivered to intended receiver.
By this test we wanted to determine a maximum value of message throughput for which all messages are delivered successfully and in some reasonable time (say less than 1 minute).
Some variations of test units are possible, e.g. JMS connections and/or sessions can be open/closed for every sent message or shared among multiple producers. But (contrary to our expectations) we encountered no visible differences in message throughput and message delivery time.
Code of test units:
Sender.java
package pl.ericpol.jmstest;
import javax.jms.Connection;
import javax.jms.JMSException;
import javax.jms.MessageProducer;
import javax.jms.Session;
public class Sender{
private UnitTest unitTest = null;
private Connection connection = null;
private Session session = null;
private MessageProducer producer = null;
private String selector = null;
public Sender(Connection con, Session session, MessageProducer producer, String selector, UnitTest unitTest){
public void send() {
try {
boolean closeSession = false;
boolean closeProducer = false;
if(this.session == null){
this.session = this.connection.createSession(false, Session.AUTO_ACKNOWLEDGE);
closeSession = true;
if(this.producer == null){
this.producer = this.unitTest.createProducer(this.session);
closeProducer = true;
this.producer.send(this.unitTest.createMessage(this.session, this.selector));
if(closeProducer){
this.producer.close();
this.producer = null;
if(closeSession){
this.session.close();
this.session = null;
} catch (JMSException e) {
e.printStackTrace();
Receiver.java
package pl.ericpol.jmstest;
import javax.jms.Connection;
import javax.jms.ConnectionFactory;
import javax.jms.Destination;
import javax.jms.JMSException;
import javax.jms.Message;
import javax.jms.MessageConsumer;
import javax.jms.MessageListener;
import javax.jms.MessageProducer;
import javax.jms.Session;
import javax.jms.TextMessage;
public class Receiver extends Thread implements MessageListener{
private Connection connection = null;
private Session session = null;
private Destination destination = null;
private MessageConsumer consumer = null;
private Connection sendConnection = null;
private Session sendSession = null;
private MessageProducer producer = null;
private String qname = null;
private String selector = null;
private int messagesToReceive = -1;
private int delay = -1;
private UnitTest unitTest = null;
private boolean active = true;
private long[][] localStats = null;
private Boolean monitor = new Boolean(true);
public Receiver(ConnectionFactory cf, Connection sendConnection, Session sendSession, MessageProducer producer,
String code, UnitTest unit){
try {
this.connection = cf.createConnection();
this.session = this.connection.createSession(false, Session.AUTO_ACKNOWLEDGE);
this.destination = this.session.createQueue(this.qname);
this.consumer = this.session.createConsumer(this.destination, UnitTest.KEY + " = '" + code + "'");
this.consumer.setMessageListener(this);
this.connection.start();
this.producer = producer;
this.sendSession = sendSession;
this.sendConnection = sendConnection;
} catch (JMSException e) {
e.printStackTrace();
} catch (NumberFormatException e1) {
e1.printStackTrace();
public void run() {
if(this.consumer == null){
} else {
this.sleep(0);
this.close();
TestManager.getInstance().unitTestFinished();
public synchronized void onMessage(Message arg0) {
Message message = arg0;
if(this.active){
if(message == null){
} else {
if(message instanceof TextMessage){
this.registerDeliveryTime(this.localStats.length - this.messagesToReceive, message);
this.sleep(this.delay);
synchronized(this.monitor){
if(--this.messagesToReceive == 0){
this.unitTest.messagesReceived();
} else {
if(this.active){
this.send();
} else {
} else {
if(! this.active){
this.notify();
public synchronized void deactivate(){
public int getMessagesToReceive(){
return this.messagesToReceive;
private void send(){
try {
boolean closeSession = false;
boolean closeProducer = false;
if(this.sendSession == null){
this.sendSession = this.sendConnection.createSession(false, Session.AUTO_ACKNOWLEDGE);
closeSession = true;
if(this.producer == null){
this.producer = this.unitTest.createProducer(this.sendSession);
closeProducer = true;
this.producer.send(this.unitTest.createMessage(this.sendSession, this.selector));
if(closeProducer){
this.producer.close();
this.producer = null;
if(closeSession){
this.sendSession.close();
this.sendSession = null;
} catch (JMSException e) {
e.printStackTrace();
private synchronized void sleep(int delay){
private void registerDeliveryTime(int index, Message message){
private synchronized void close(){
try {
if(this.producer != null){
this.producer.close();
if(this.sendSession != null){
this.sendSession.close();
this.sendSession = null;
if(this.sendConnection != null){
this.sendConnection.close();
this.sendConnection = null;
if(this.consumer != null){
this.consumer.close();
this.consumer = null;
if(this.session != null){
this.session.close();
this.session = null;
if(this.connection != null){
this.connection.close();
this.connection = null;
} catch (JMSException e) {
e.printStackTrace();
TestUnit.java
package pl.ericpol.jmstest;
import java.util.HashMap;
import java.util.Iterator;
import javax.jms.Connection;
import javax.jms.DeliveryMode;
import javax.jms.JMSException;
import javax.jms.MessageProducer;
import javax.jms.Session;
import javax.jms.TextMessage;
import com.sun.messaging.ConnectionFactory;
import com.sun.messaging.QueueConnectionFactory;
public class UnitTest extends Thread{
public static final String KEY = "Type";
private ConnectionFactory cf = null;
private Connection sendConnection = null;
private Session sendSession = null;
private MessageProducer producer = null;
private int delay = -1;
private int messagesPerCall = -1;
private int loop = -1;
private int timeOut = -1;
private int maxLoops = -1;
private boolean waitingFlag = false;
private boolean messagesReceived = false;
public UnitTest(int loop){
try {
this.cf = new QueueConnectionFactory();
this.applyProps(this.cf, Properties.getInstance().getSunProps());
this.sendConnection = this.cf.createConnection();
if(sharedProducer){
this.sendSession = this.sendConnection.createSession(false, Session.AUTO_ACKNOWLEDGE);
this.producer = this.createProducer(this.sendSession);
} else if(sharedSession){
this.sendSession = this.sendConnection.createSession(false, Session.AUTO_ACKNOWLEDGE);
this.sendConnection.start();
} catch (NumberFormatException e){
MyLogger.logger.error("number format exception!!");
} catch (JMSException e) {
e.printStackTrace();
public void run(){
String selector = String.valueOf(Randomizer.getRandInt(Integer.MAX_VALUE));
Receiver receiver = new Receiver(this.cf, this.sendConnection, this.sendSession, this.producer, selector, this);
receiver.start();
long startTime = System.currentTimeMillis();
this.send(selector);
synchronized (this) {
if(! this.messagesReceived){
this.sleep(this.timeOut);
long finishTime = System.currentTimeMillis();
receiver.deactivate();
if(++this.loop < this.maxLoops && TestManager.getInstance().getStatus()){
UnitTest newTest = new UnitTest(this.loop);
newTest.start();
} else {
TestManager.getInstance().workerFinished();
public void messagesReceived(){
private synchronized void sleep(int delay){
private void applyProps(ConnectionFactory cf, HashMap props){
private void send(String selector){
Sender sender = new Sender(this.sendConnection, this.sendSession, this.producer, selector, this);
sender.send();
public MessageProducer createProducer(Session session){
MessageProducer producer = null;
String qname = (String) Properties.getInstance().getOtherProps().get(Properties.JMS_QUEUE_NAME);
try {
producer = session.createProducer(session.createQueue(qname));
producer.setDeliveryMode(DeliveryMode.NON_PERSISTENT);
} catch (JMSException e) {
e.printStackTrace();
return producer;
public TextMessage createMessage(Session session, String selector){
try {
TextMessage message = session.createTextMessage();
message.setStringProperty(UnitTest.KEY, selector);
String messageLenAsString = (String) Properties.getInstance().getOtherProps().get(Properties.JMS_MESSAGE_LENGTH);
int messageLen = Integer.parseInt(messageLenAsString);
StringBuffer buf = new StringBuffer();
buf.append(selector).append("-");
for(int i = 0; i < messageLen; i++){
buf.append("x");
message.setText(buf.toString());
return message;
} catch(NumberFormatException e){
MyLogger.logger.error("bad message length!!");
} catch (JMSException e) {
e.printStackTrace();
return null;
}If you go here:
http://www.sun.com/software/solaris/get.jsp
and
1.Check Sun Java Enterprise (or the Sun Java Application Platform Suite),
2.Click Get Download and Media
3. Then select Systems for windows (the bar at the top)
I beleive you can get Message Queue 3.6 SP3 from it, by only installing the Message Queue component.
Note: it is a large download (500 Mb for Application Platform), especially for Sun Java Enterprise...
TE -
Disk throughput drops when inserting data packages in write-optimized DSO
Hi all,
we are currently testing our new freshly installed SAN.
To see the performance gain in BI, I'm currently doing some test loads.
And during the monitoring of those loads, I noticed something I'd like someone to explain :-):
I execute a DTP from PSA to a write-optimized DSO.
The n° of parallel processes = 9
Update method = serial extraction, immediate parallel processing
N° of records transferred: +23.000.000
Ok, in the first phase (read the PSA) only one process is used (serial extraction). When I look in OS07, I notice we have very good throughput: +66.000 TransfKB/s. Very nice!
But as soon as BI starts inserting the data packages, and parallel processing kicks in, the throughput drops to 4K or something, and sometimes we get 20K at max. That's not too good.
We have a massive SAN , but the BI system does not seem to use it?
I was wondering why this is the case. I already toyed around with the package size, but it's always the same.
Also I noticed that the allocated processes don't seem to be active. I allocated 9 BTC processes to this load.
They are all used, but we only see 3 inserts at the same time, max. Also in the DTP-monitor, only 3 packages are processed at the same time. As it's a write-optimized DSO, RSODSO_SETTINGS does not apply I presume.
Any ideas?
tnx!Hi,
can you pls try to give some filetr in DTP and try to pull the data.
I am not sure why first data package is taking long time and otehr data package is taking less time..
Do you have any start routine..If datapak = 1.. the do this logic..
Pls check..
regards
Gopal -
Network user access to local disk to solve I/O throughput issue?
Not sure if this is the best forum for this, but...
THE ISSUE:
I work at a small university and am responsible for the Mac network and video dept (only 1 xserve and about 20-odd clients). Anyway, everyone uses network accounts. I have 3 iMacs for Final Cut and so far everyone just captures to the server over gigabit ethernet. We're only using DV25 (~4MB/sec transfer) and it hasn't been an issue until this week. Now as people begin to stress more about final projects, however, and are spending more time, the suites are being used more during the same times. Even though the throughput is usually less than 10MB/sec, which (I think?) is well under what gigabit ethernet supports, we're consistently getting stalled video and dropped frames. We've often had multiple people working on projects (playback) simultaneously in the past, but never capturing simultaneously before. Also, another factor that may come into play, is my 3TB RAID 5 that they're using has become rather full at this point (down to 200MB) so perhaps read/write times are being affected?
Now, we're looking at adding 6 more Final Cut suites, which will potentially increase my throughput load by 200%, so I really need to find a solid solution (and fibre cards are not an option since they're iMacs--and we don't have that budget anyway).
THE QUESTION:
Since the issue seems to only occur while capturing and not during playback, it seems the easiest solution would be to give the editor group access to the local hard drive (and then copy from there to their network storage so they can work on any given station later) which I'm not sure how the best way to do this would be. OR, should gigabit ethernet be capable of sustaining the throughput I need and there's some other problem I should be looking at to fix that?
Related, if they do capture to the local disk (or even an external hard drive, which we may end up going to), is there some script or something that can be run every night to automatically copy files from X location on a local drive to Y location (say, a common media storage location) on the server RAID?Hi
+". . . my 3TB RAID 5 that they're using has become rather full at this point (down to 200MB). . ."+
I think the problem you're currently seeing starts (and probably ends) here. Regardless of what your data comprises of the general rule of thumb is to make sure you keep at least 10% of free disk space for any hard drive let alone a RAID. Most people would generally recommend 15% for a RAID 5. A rough calculation based on your 3TB RAID would mean you should keep at least 300GB free. You're well below that unless 200MB is a typo? I'm surprised the whole thing has not ground to a halt?
Video Capture, Editing and Playback in a network home environment is always going to place a considerable strain on any network unless you've prepared the ground first. For example 60 seconds of RAW Video is roughly 200MB in size. Multiply this for a classroom of 16 and that's a lot of data. Even if you're not using hi-def you're still talking a lot of data.
There are better ways of doing this.
HTH?
Tony -
How to setup a infiniband network
How to setup an infiniband network??
Our Customer has 2 x t5-4 wirh ZS3 using infiniband. Do you have any white paper where explain how we can do
Thanks in advancePerhaps you can start with the product documentation for those T5-4 systems:
http://docs.oracle.com/cd/E29659_01/index.html
When you are on that page, find the Search field at the top and type "infiniband" into it.
You might find additional information specific to the hardware by doing that. -
High synch rate with VERY slow throughput
Help!!!
I'm hoping that the one of the mods on here may be able to intervene and help get my issue sorted.
I had my home broadband upgraded to adsl2 on the 8th of Feb. after 5 years of hassle free service on up to 8 Meg service.
From the day of upgrade (of which has given no significant increase in speed by the way), I've had an issue with fluctuating throughput whilst my synch rate remains rock solid at 6.5-7mb/s.
On any given day I can get my full expected throughput of between 6 and 6.5mb/s but can have long periods of dial up speeds at 40-50kbp/s which can be morning, day or night, so not a peak time bandwidth cap/throttle issue. I have given it the 10 day stabilisation before taking up with tech helpdesk, but after nearly 2 weeks of calls and going round in circles I'm running out of patience.
Can I just add I'm an Openreach engineer who deals with broadband/SFI faults daily so have made sure the issue does not lie within the network or premises. I've also taken a hub and connected it direct on to the adsl out port of the exchange mdf and witnessed the same speed problems, so am in no doubt that the problem is with BT. When I try to explain to the helpdesk advisors it doesn't seem to register. Whether it's down to a fault on my port or a general congestion issue on the backhaul I don't know but just need someone to get to the bottom of it for me
I am awaiting a call back with an update in the next 2 days but judging by the way thing have gone down so far I'm not confident I'll get a result, so was hoping someone on here could help.in order to get things moving you need to post your router adsl stats and also btspeedtester results. would help if you can post a good set where download is ok and then a poor set for comparison
If you like a post, or want to say thanks for a helpful answer, please click on the Ratings star on the left-hand side of the post.
If someone answers your question correctly please let other members know by clicking on ’Mark as Accepted Solution’. -
Slow throughput on 7.1 Mb/s line
I ran some speed tests today to check my 7.1 Mb/s line as I do occasionally and I found out that even though I was syncing with the CO at 7661 my download speed tests are only coming back in the 1.1 to 2.0 range. Some were actually well below 1.0 today. Previously I was routinely getting speed tests in the 6.5 to 6.6 range on this line, which has been provisioned at this speed for 3 and a half years. My current transceiver stats are as follows:
Transceiver Statistics
Transceiver Revision: 7.2.3.0
Vendor ID Code: 4
Line Mode: ADSL2+ Mode
Data Path: Interleaved
Transceiver Information Downstream Path Upstream Path
DSL Speed (Kbits/Sec) 7661 863
Margin (dB) 6.0 14.0
Line Attenuation (dB) 45.0 25.0
Transmit Power (dBm) 19.4 12.1
I know that my downstream margin is low at 6.0 but this is a VERY STABLE line that will go weeks without losing sync even with that low margin. The ADSL2+ is really giving me some help at my attenuation.
So far today for troubleshooting I tried replacing my 7-year-old Westell 327W B90 modem with a brand-new Westell 327W D90 modem that I had stored away in a box. There was no difference in the speed tests when I used either modem. They were both running the 4.04.03 firmware which is the latest for the Westell 327W modems.
I also tried uninstalling and reinstalling my NIC card as I'm connected via ethernet cable to my modem. That made no difference in my speed tests either.
I did run a Network Diagnostic Tool at M-Lab with the following results:
Upload speed
0.73 Mb/s
Download speed
1.1 Mb/s
Network latency: 5.8e+2 msec round trip time
Jitter: 8.1e+2 msec
Your system: Windows XP version 5.1
Java version: 1.7.0_04 (x86)
TCP receive window: 255552 current, 255552 maximum
9.62464E-4 packets lost during test
Round trip time: 70 msec (minimum), 875 msec (maximum), 580.12 msec (average)
Jitter: 805 msec
0 seconds spend waiting following a timeout
TCP time-out counter: 935
130 selective acknowledgement packets received
No duplex mismatch condition was detected.
The test did not detect a cable fault.
Network congestion may be limiting the connection.
No network address translation appliance was detected.
0.9756% of the time was not spent in a receiver limited or sender limited state.
0.0% of the time the connection is limited by the client machine's receive buffer.
Optimal receive buffer: 261685248 bytes
0 duplicate ACKs set
So it looks like I have a network congestion issue going on. I'm wondering if it has to do with the router that I'm on or if my local CO (Fall River, MA) is having a more widespread problem. I'm thinking of keeping an eye on it for the next few days and see if it gets back to normal on its own. I'm very hesitant to contact Verizon as the last time I did a few years ago they cut my speed down to 5 from 7.1 and took away my ADSL2+. I had to talk to executive customer relations to get them back and it was not an easy task!
Does anyone have any suggestions or opinions about my situation?PROBLEM RESOLVED
I contacted the Verizon Direct Forum at DSL Reports and received a telephone call from someone at the executive customer relations department at Verizon. He was able to have someone from my local Central Office who does network management/DSL provisioning contact me directly by phone. I was changed over to a "less congested" router right while I was on the phone with this person from my local CO. My ADSL2+ and 7.1 speed package were left intact on my line during the change to a new router. All seems to be well as I am now getting actual speed throughput of 6.5 to 6.6 even during peak evening hours! -
Slow throughput Ironport S370 Proxy CPU 100%
We have a cluster of 3 x Ironport S370's all running 7.7.0-753
The throughput is really poor we have a 500Mbps Internet connection which at it's peak is only getting to 120Mbps as the Ironports don't seem to be able to handle the traffic.
The Proxy CPU% is always close to 100% but the overall CPU is usually at no more than 30% at times it can take up to 60 seconds to load the initial page particularly if the site is an HTTPS site.
We have
22 Identities
62 Access policies
6 decryption Policies
Our maintainer says that having this number of Identities / policies should not be an issue but I have my doubts.
Can anyone advise as it's really become a major issue, Output from the rate and status commands are below.
%proxy reqs client server %bw disk disk
CPU /sec hits blocks misses kb/sec kb/sec saved wrs rds
99.00 285 373 1293 1193 26484 21838 17.5 550 100
99.00 286 209 1313 1335 28682 24532 14.5 635 80
99.00 285 182 1323 1359 37083 33529 9.6 1351 0
100.00 231 132 1051 1113 34816 34151 1.9 355 0
98.00 253 161 1171 1195 39668 37236 6.1 1363 0
99.00 294 256 1225 1469 51371 43304 15.7 1117 40
96.00 346 525 1166 1763 31882 23300 26.9 1328 0
98.00 302 228 1258 1534 30385 25565 15.9 1302 0
99.00 295 149 1200 1597 26253 22888 12.8 816 0
98.00 275 199 1020 1536 35237 31443 10.8 838 0
99.00 288 184 1131 1574 35019 26688 23.8 1433 0
99.00 262 116 1073 1437 24744 23228 6.1 1306 0
105.00 307 292 1165 1610 24249 20236 16.6 1061 0
Status as of: Thu Oct 16 08:28:10 2014 GMT
Up since: Wed Oct 15 15:21:19 2014 GMT (17h 6m 51s)
System Resource Utilization:
CPU 28.2%
RAM 82.6%
Reporting/Logging Disk 16.0%
Transactions per Second:
Average in last minute 266
Maximum in last hour 296
Average in last hour 118
Maximum since proxy restart 296
Average since proxy restart 9
Bandwidth (Mbps):
Average in last minute 25.461
Maximum in last hour 49.605
Average in last hour 16.400
Maximum since proxy restart 49.605
Average since proxy restart 1.365
Response Time (ms):
Average in last minute 179
Maximum in last hour 526
Average in last hour 192
Maximum since proxy restart 17710
Average since proxy restart 3165
Cache Hit Rate:
Average in last minute 16
Maximum in last hour 25
Average in last hour 7
Maximum since proxy restart 25
Average since proxy restart 0
Connections:
Idle client connections 1276
Idle server connections 1170
Total client connections 1638
Total server connections 1890In the release notes it states...
IMPORTANT: During testing of AsyncOS 7.7.0, Cisco observed performance changes ranging from +
33% to - 16%, depending on the model and configuration. Performance degradation risk is limited to
S160 & S360 models and models S370 and S660 that are running the web proxy without security
services. If you experience performance degradation with AsyncOS 7.7.0, Cisco recommends that you
revert to AsyncOS 7.5.x.
http://www.cisco.com/c/dam/en/us/td/docs/security/wsa/wsa7-7/Release_Notes/WSA_7-7-0_Builds_after_725_Release_Notes.pdf
Maybe you are looking for
-
How do I set-up my Dataset to pull data for Service Requests for a report
I am using SQL Server 2012 Report Builder 3.0. I have my Data Source added - DWDataMart I want to build my Dataset to query and find a very specific item. Basically I am creating a daily report that will show what each "team" or Support Group (AD Obj
-
Can't make jpeg in Photoshop CC
I have Photoshop CC (with Windows 8). I can't make jpeg and pdf in Photoshop CC. Is this normal or there is something wrong with my computer? (Very annoying, please help!) Thanks
-
I purchased iphone 5S in Saudi Arabia; there are iphones promoted with Face Time app, sold here widely. When I purchased it, there was no explicit ad for Face Time with it. I have Gold, 64 GB. When I extensively searched for this app, i just found i
-
When I put up a new photo album in Dreamweaver and put it live on my site, only the thumbnails show up. What am I doing wrong
-
hi could any body tel me pricing procedure in sap............