philb19
asked on
10GB switch - and increased speed
Hi,
Hypothetical. I have traffic utilization of 1GB NIC at say 20% average- bursting to max of 40%
I replace with 10GB switch and NICs. - Will I see improved speed and performance at all?
Or will I just use 2% of 10GB and max 4% of 10GB on a burst?
Can someone please give technical answer. I know there are other benefits to 10GB in consolidation of number of NICs required in VM environment etc - I'm purely after an explanation or confirmation of what performance/speed gain I will get given my current utilization Thanks
Hypothetical. I have traffic utilization of 1GB NIC at say 20% average- bursting to max of 40%
I replace with 10GB switch and NICs. - Will I see improved speed and performance at all?
Or will I just use 2% of 10GB and max 4% of 10GB on a burst?
Can someone please give technical answer. I know there are other benefits to 10GB in consolidation of number of NICs required in VM environment etc - I'm purely after an explanation or confirmation of what performance/speed gain I will get given my current utilization Thanks
Your math is correct, assuming no change in traffic volume (200MB avg, 400MB max), you will simply see 2% avg and 4% max of 10GB after upgrade.
ASKER
ok thanks - so no performance gain whatsoever?
Well, It may be a mistake to assume constant traffic volume: your computers can get and send responses faster- thus they may generate MORE traffic per time unit. Application specific, I guess.
ASKER
"your computers can get and send responses faster- "
ok this is getting to my point - How do they do this? so I assume the 10GB is not just the size of the pipe. Its pushing data faster due to the 10GB chipsets? so faster processing on the NIC and switch?
ok this is getting to my point - How do they do this? so I assume the 10GB is not just the size of the pipe. Its pushing data faster due to the 10GB chipsets? so faster processing on the NIC and switch?
A "response" is a number of bytes in length, lets say 1KB.
With a 1GB network this data can be transferred in 1 milli seconds. (check my math)
With a 10GB network this 1KB can be transferred 10x faster, 0.01 milliseconds.
Note: 1 milisecond is a lot of time for a CPU to process a reply
Hmm, is that clear at all?
With a 1GB network this data can be transferred in 1 milli seconds. (check my math)
With a 10GB network this 1KB can be transferred 10x faster, 0.01 milliseconds.
Note: 1 milisecond is a lot of time for a CPU to process a reply
Hmm, is that clear at all?
ASKER
ok thanks again. I have read online that 90% of SOHO business's run fine on 1GB full duplex LAN.
My issue is whether the network is "the bottlenecK' - And from my research with a small VLAN 20-30 PC 1GB full duplex - this is rarely the case.
The arguement from developers and DBA's is to get 10GB LAN. However I see only 30-40% utilization on 1GB - Now my assumption is that (we do have Oracle) the application layer - poor SQL coding is more than likely the cause of slow oracle business application performance rather than the LAN wire speed. - I'm trying to gauge whether 10GB will make any difference at all. I don't believe the clients are waiting for network - so what will 10GB give? though I could be wrong. - thoughts?
My issue is whether the network is "the bottlenecK' - And from my research with a small VLAN 20-30 PC 1GB full duplex - this is rarely the case.
The arguement from developers and DBA's is to get 10GB LAN. However I see only 30-40% utilization on 1GB - Now my assumption is that (we do have Oracle) the application layer - poor SQL coding is more than likely the cause of slow oracle business application performance rather than the LAN wire speed. - I'm trying to gauge whether 10GB will make any difference at all. I don't believe the clients are waiting for network - so what will 10GB give? though I could be wrong. - thoughts?
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
you will see an improvement only if the network speed is your bottleneck. Perhaps your bottleneck is somewhere else.
For instance, I noticed a huge improvement by putting my sql server data files, log files, temp files all on different spindles (physical drives), the log files went to an SSD, this resulted in a 400% increase on my System Center Server
For instance, I noticed a huge improvement by putting my sql server data files, log files, temp files all on different spindles (physical drives), the log files went to an SSD, this resulted in a 400% increase on my System Center Server
I agree with David Johnson, CD, you will only see an improvement if the network is your bottle neck.
What is the time period you are seeing 30-40% network utilization? Is this an 8 hour avg.?
Remember at any specific point in time a single network connection is either 100% utilized or 0% utilized. As data is flowing over the wire it is traveling at 100% of the network connection maximum speed.
An avg. tells you what percentage of the time the network was 100% utilized. So network connection that is avg. 30-40% utilization is 100% utilized 30-40% of the time and zero% utilized 60-70% of the time.
What is the time period you are seeing 30-40% network utilization? Is this an 8 hour avg.?
Remember at any specific point in time a single network connection is either 100% utilized or 0% utilized. As data is flowing over the wire it is traveling at 100% of the network connection maximum speed.
An avg. tells you what percentage of the time the network was 100% utilized. So network connection that is avg. 30-40% utilization is 100% utilized 30-40% of the time and zero% utilized 60-70% of the time.