P2P Caching

Old, inactive threads
Munka
Posts: 289
Joined: Sat Oct 22, 2005 8:22 pm
Location: Rural NSW

Post by Munka » Wed Nov 14, 2007 3:36 pm

Spanner_Man wrote: Maybe its just me but i know some other users on here are a little pedantic (like myself) when it comes to config's on their own systems, so this question isn't just for myself but for others.
No it is not just you, I am interested to know how a user paying the extra premium for cached data/extra quota, can have some clue that said data is actually coming from the cache.

On the other hand I realize it is early days, but the devil is in the detail, particularly when it comes to sales.
Munka

James
Exetel Staff
Posts: 1991
Joined: Mon May 09, 2005 10:27 pm

Post by James » Wed Nov 14, 2007 3:44 pm

Munka wrote:
Spanner_Man wrote: Maybe its just me but i know some other users on here are a little pedantic (like myself) when it comes to config's on their own systems, so this question isn't just for myself but for others.
No it is not just you, I am interested to know how a user paying the extra premium for cached data/extra quota, can have some clue that said data is actually coming from the cache.

On the other hand I realize it is early days, but the devil is in the detail, particularly when it comes to sales.
The same way that a user paying for his dsl account can have some clue that they actually downloaded said data from his browsing/downloads/etc activities?

tocpcs
Posts: 523
Joined: Sun Aug 26, 2007 10:01 am
Location: Online

Post by tocpcs » Wed Nov 14, 2007 4:40 pm

James wrote:
Spanner_Man wrote:
James wrote:All p2p requests will go through this system automatically and find the bits of data necessary if it's available.

There is no need to set your software to do anything different.
James,
You misunderstand me, or perhaps i wasn't making myself clear.

To ensure (ie force) the particular client/program to use the cache system is what i am talking about, for users that want to make sure for an actual fact (settings within the program/client) that all p2p data will be passed/requested through the cache.

Maybe its just me but i know some other users on here are a little pedantic (like myself) when it comes to config's on their own systems, so this question isn't just for myself but for others.
There is no need to force it, it will do it automatically.
BUT checking to see if its got to go to the cache is a process that no doubt consumes process power.

Hitting it directly just makes sense! It reduces load on the systems at the front line for filtering it, if the traffic doesn't need to be filtered and instead just gets directed at the appropriate point.

James
Exetel Staff
Posts: 1991
Joined: Mon May 09, 2005 10:27 pm

Post by James » Wed Nov 14, 2007 4:48 pm

tocpcs wrote:
James wrote:
Spanner_Man wrote:James,
You misunderstand me, or perhaps i wasn't making myself clear.

To ensure (ie force) the particular client/program to use the cache system is what i am talking about, for users that want to make sure for an actual fact (settings within the program/client) that all p2p data will be passed/requested through the cache.

Maybe its just me but i know some other users on here are a little pedantic (like myself) when it comes to config's on their own systems, so this question isn't just for myself but for others.
There is no need to force it, it will do it automatically.
BUT checking to see if its got to go to the cache is a process that no doubt consumes process power.

Hitting it directly just makes sense! It reduces load on the systems at the front line for filtering it, if the traffic doesn't need to be filtered and instead just gets directed at the appropriate point.
It has an entire blade of servers, it has more than enough processing power to cater for any/all requests; besides there is no way to see what's actually on it, it is all in fragmented bits and bytes.

tocpcs
Posts: 523
Joined: Sun Aug 26, 2007 10:01 am
Location: Online

Post by tocpcs » Wed Nov 14, 2007 5:01 pm

James wrote:
tocpcs wrote:
James wrote: There is no need to force it, it will do it automatically.
BUT checking to see if its got to go to the cache is a process that no doubt consumes process power.

Hitting it directly just makes sense! It reduces load on the systems at the front line for filtering it, if the traffic doesn't need to be filtered and instead just gets directed at the appropriate point.
It has an entire blade of servers, it has more than enough processing power to cater for any/all requests; besides there is no way to see what's actually on it, it is all in fragmented bits and bytes.
Yes, but I'm not suggesting "viewing" it. The suggestion is hitting the cache directly, or even, limiting to cache only.

James
Exetel Staff
Posts: 1991
Joined: Mon May 09, 2005 10:27 pm

Post by James » Wed Nov 14, 2007 5:06 pm

tocpcs wrote:
James wrote:
tocpcs wrote: BUT checking to see if its got to go to the cache is a process that no doubt consumes process power.

Hitting it directly just makes sense! It reduces load on the systems at the front line for filtering it, if the traffic doesn't need to be filtered and instead just gets directed at the appropriate point.
It has an entire blade of servers, it has more than enough processing power to cater for any/all requests; besides there is no way to see what's actually on it, it is all in fragmented bits and bytes.
Yes, but I'm not suggesting "viewing" it. The suggestion is hitting the cache directly, or even, limiting to cache only.
What's the point of that?

Graham C
Posts: 90
Joined: Mon Apr 12, 2004 8:16 pm

Post by Graham C » Wed Nov 14, 2007 5:13 pm

James wrote:
tocpcs wrote:
James wrote: It has an entire blade of servers, it has more than enough processing power to cater for any/all requests; besides there is no way to see what's actually on it, it is all in fragmented bits and bytes.
Yes, but I'm not suggesting "viewing" it. The suggestion is hitting the cache directly, or even, limiting to cache only.
What's the point of that?
To utilise the extra 'cache bandwidth' provided in the hypothetical P2P/cache plans spoken of initially?

Klaas
Posts: 609
Joined: Sat Apr 03, 2004 1:48 pm
Location: Sydney

Post by Klaas » Wed Nov 14, 2007 7:19 pm

I think we should all wait for the trial/tests to commence, to see if the claimed performance improvement eventuates, before worrying about how to tweak it.

I've got to admit, though, I think the idea of limiting data requests to the cache only, is not utilising caching technology properly.

If you hit the cache directly, and the data you want is only partly cached, you're not going to get the whole file, I would have thought.

tocpcs
Posts: 523
Joined: Sun Aug 26, 2007 10:01 am
Location: Online

Post by tocpcs » Wed Nov 14, 2007 8:35 pm

Klaas wrote:I think we should all wait for the trial/tests to commence, to see if the claimed performance improvement eventuates, before worrying about how to tweak it.

I've got to admit, though, I think the idea of limiting data requests to the cache only, is not utilising caching technology properly.

If you hit the cache directly, and the data you want is only partly cached, you're not going to get the whole file, I would have thought.
That'd be fine. Wait until the whole file is cached.

Hitting the cache directly would mean that you avoid any Peer traffic from outside the cache. ie. hitting the cache means all your traffic stays inside the network, and if the cache doesn't have it, you can either wait for it to get it (ie, someone else making a request that results in the part being cached, or go the traditional method).

The theory here is to stop all access except for cache access to save bandwidth.

tocpcs
Posts: 523
Joined: Sun Aug 26, 2007 10:01 am
Location: Online

Post by tocpcs » Wed Nov 14, 2007 8:36 pm

Klaas wrote:I think we should all wait for the trial/tests to commence, to see if the claimed performance improvement eventuates, before worrying about how to tweak it.

I've got to admit, though, I think the idea of limiting data requests to the cache only, is not utilising caching technology properly.

If you hit the cache directly, and the data you want is only partly cached, you're not going to get the whole file, I would have thought.
Just on the performance, is there a date in mind as to when performance data will be available?

CrackerJak
Posts: 37
Joined: Wed Nov 07, 2007 4:03 am
Location: NSW

Post by CrackerJak » Thu Nov 15, 2007 4:47 am

I'm very interested in this p2p cache system, in particular what type of material will be cached. Will the cache be limited to "legitimate" traffic such as linux iso's and freeware? If not how will you decide what gets cached?

4 terabytes isn't that much storage for the cache when you consider most of your members download 88gigs a month, unless your very selective on what you cache.

ForumAdmin
Posts: 3663
Joined: Sun Jan 04, 2004 2:31 pm
Location: Sydney

Post by ForumAdmin » Thu Nov 15, 2007 8:35 am

The physical installation of the caching system is now complete.

The routing with our providrs will need to be modified by them which will delay the start of the trial until the end of next week.

ForumAdmin
Posts: 3663
Joined: Sun Jan 04, 2004 2:31 pm
Location: Sydney

Post by ForumAdmin » Thu Nov 15, 2007 12:19 pm

CrackerJak wrote:I'm very interested in this p2p cache system, in particular what type of material will be cached.

1) Will the cache be limited to "legitimate" traffic such as linux iso's and freeware?

2) If not how will you decide what gets cached?

3) 4 terabytes isn't that much storage for the cache when you consider

4) most of your members download 88gigs a month, unless your very selective on what you cache.
1) No.

2) The servers decide based on P2P signatures

3) 4 terabytes is a great deal of cache.

4) Firstly the 'average' Exetel ADSL user downloads 8 gb per month. Secondly many people download the same files at the same time - the sole principle of using cache.

Klaas
Posts: 609
Joined: Sat Apr 03, 2004 1:48 pm
Location: Sydney

Post by Klaas » Thu Nov 15, 2007 3:46 pm

tocpcs wrote:
Klaas wrote:I think we should all wait for the trial/tests to commence, to see if the claimed performance improvement eventuates, before worrying about how to tweak it.

I've got to admit, though, I think the idea of limiting data requests to the cache only, is not utilising caching technology properly.

If you hit the cache directly, and the data you want is only partly cached, you're not going to get the whole file, I would have thought.
That'd be fine. Wait until the whole file is cached.

Hitting the cache directly would mean that you avoid any Peer traffic from outside the cache. ie. hitting the cache means all your traffic stays inside the network, and if the cache doesn't have it, you can either wait for it to get it (ie, someone else making a request that results in the part being cached, or go the traditional method).

The theory here is to stop all access except for cache access to save bandwidth.
And if everyone does that, no data would be loaded into the cache to start with :)

I understand the point you are trying to make, but I think it conflicts with the whole idea of caching technology (either P2P data or any other caching process).

sh0nky
Posts: 143
Joined: Sat Sep 15, 2007 9:09 pm
Location: Bankstown/Cronulla

Post by sh0nky » Thu Nov 15, 2007 8:03 pm

Agreed totally... The whole point of this cache, if i've read these forums correctly, is to store data for easy access and quicker data/file transfers. For data to be stored by the cache it has to be sourced from elsewhere the 1st time. After this process has happened once, the need to source the data/file from elsewhere is eliminated.....making the whole process quicker!
I can understand the desire to want to ''Hit'' the cache direct, but if everyone did this, especially when the cache is accumulating data/files during its initial implementation, the benefits of the cache will take much longer to utilize,due to there being nothing in the cache to start with... making it slower!
Hard work never killed anyone ..but im not taking any chances :)

Locked