Just running some small benchmarks between EDB, DBISAM and Interbase. EDB performs about as fast as Interbase, but it uses five times the network bandwidth. IB runs around 1 Mbps and EDB runs at 5 Mbps, with wall times of 52 and 54 seconds respectively for inserting 100K records. Interesting.

Arthur

>Just running some small benchmarks between EDB, DBISAM and Interbase. EDB performs about as fast as Interbase, but it uses five times the network bandwidth. IB runs around 1 Mbps and EDB runs at 5 Mbps, with wall times of 52 and 54 seconds respectively for inserting 100K records. Interesting.
>

Not being a C/S user, and assuming you're inserting the same data, I wonder about compression. Have you used a packet sniffer to see what's actually being transfered?

Two questions:

1. What about DBISAM?
2. What is a wall time?

Roy

Wall time is the elapsed time based on the clock on the wall.

DBISAM performs rather poorly. It runs about 4 Mbps but the elapsed time is just over 4 minutes. Not even remotely close to the other two.

No, I didn't run a packet sniffer. I don't care what's going on, it was just a simple test to see how EDB compared against something that costs almost ten times as much. I was pleasantly surprised at how well it did.

----------------------------------------------
Roy Lambert wrote:

Not being a C/S user, and assuming you're inserting the same data, I wonder about compression. Have you used a packet sniffer to see what's actually being transfered?

Two questions:

1. What about DBISAM?
2. What is a wall time?

Roy

On 10/3/2013 10:13 PM, Arthur Williams wrote:
> Just running some small benchmarks between EDB, DBISAM and Interbase. EDB performs about as fast as Interbase, but it uses five times the network bandwidth. IB runs around 1 Mbps and EDB runs at 5 Mbps, with wall times of 52 and 54 seconds respectively for inserting 100K records. Interesting.

I'd definitely suggest you try this with EDB Session compression setting
enabled ( default is 0 which is none).

Setting is called RemoteCompression and value runs from 0..9. Manual
recommends 6 for good speed/size compromise.

Raul

Arthur

>Wall time is the elapsed time based on the clock on the wall.

Oh.

On a faster machine EDB turns in a markedly better performance. Running the same code, bandwidth usage jumps to 21 Mbps and cpu usage averages 30% yielding an elapsed time of 16 seconds. IB jumps up to 9 Mbps and has an elapsed time of 18 seconds, with a comparable 30% use of the cpu. So both systems increased their speed by a factor of 4 but IB needed 8 times the bandwidth to do so, whereas EDB only doubled its bandwidth usage.

I tried the compression setting, but found it to make zero difference beyond a mild (2-3%) increase in CPU usage. I suspect the record (3 ints) is simply too short for compression to have any effect.


>>
Raul wrote:

On 10/3/2013 10:13 PM, Arthur Williams wrote:
I'd definitely suggest you try this with EDB Session compression setting
enabled ( default is 0 which is none).

Setting is called RemoteCompression and value runs from 0..9. Manual
recommends 6 for good speed/size compromise.

Raul

My mistake, EDB used 4 times its original bandwidth. For some reason I was thinking it used 10 Mbps in those trials.

>>
Arthur Williams wrote:

On a faster machine EDB turns in a markedly better performance. Running the same code, bandwidth usage jumps to 21 Mbps and cpu usage averages 30% yielding an elapsed time of 16 seconds. IB jumps up to 9 Mbps and has an elapsed time of 18 seconds, with a comparable 30% use of the cpu. So both systems increased their speed by a factor of 4 but IB needed 8 times the bandwidth to do so, whereas EDB only doubled its bandwidth usage.

DBISAM shows the same 4x improvement in speed although its bandwidth usage jumps to 40 Mbps.

>>
Arthur Williams wrote:

On a faster machine EDB turns in a markedly better performance. Running the same code, bandwidth usage jumps to 21 Mbps and cpu usage averages 30% yielding an elapsed time of 16 seconds. IB jumps up to 9 Mbps and has an elapsed time of 18 seconds, with a comparable 30% use of the cpu. So both systems increased their speed by a factor of 4 but IB needed 8 times the bandwidth to do so, whereas EDB only doubled its bandwidth usage.

Arthur,

<< Just running some small benchmarks between EDB, DBISAM and Interbase. EDB
performs about as fast as Interbase, but it uses five times the network
bandwidth. IB runs around 1 Mbps and EDB runs at 5 Mbps, with wall times of
52 and 54 seconds respectively for inserting 100K records. Interesting. >>

When you say "bandwidth", do you mean just the raw amount of data
sent/received ?

Can you send me the EDB benchmark code ?  I'm curious as to what the actual
operations are (SQL, navigational, etc.), and I might be able to shed some
light on the bandwidth usage.  It may be due to some memory alignment going
on with respect to encryption of requests/responses.

Thanks,

Tim Young
Elevate Software
www.elevatesoft.com

I'm referring to the total network traffic going out over the nic. I don't care what's going on, I was just doing some small testing of some of the db software I have. I bought something like six db systems last year and now I'm sorting through them to see which support subscriptions I plan to renew and which ones I'm dropping, that's all.

I did already renew DBISAM.


>>>>>>>>>>>>>>>>>>>>
"Tim Young [Elevate Software]" wrote:

Arthur,

<< Just running some small benchmarks between EDB, DBISAM and Interbase. EDB
performs about as fast as Interbase, but it uses five times the network
bandwidth. IB runs around 1 Mbps and EDB runs at 5 Mbps, with wall times of
52 and 54 seconds respectively for inserting 100K records. Interesting. >>

When you say "bandwidth", do you mean just the raw amount of data
sent/received ?

Can you send me the EDB benchmark code ?  I'm curious as to what the actual
operations are (SQL, navigational, etc.), and I might be able to shed some
light on the bandwidth usage.  It may be due to some memory alignment going
on with respect to encryption of requests/responses.

Thanks,

Tim Young
Elevate Software
www.elevatesoft.com