[Storagesync] Re： recent issues discussed (plain text)

["[e-mail]" <[e-mail]>]

Hi，Fei,

>Is this a traditional and normal way to download a file from a storage server?
That is used by some existing downloading acceleration technologies such as CDN, CDNI.  I mean, moving data center closer to end users.

>It seems that you want to describe it from clients' perspective.
>User request URL, then DNS and CDN will help the client machine to finish the downloading
Yes, that's exactly what I mean. 

>You mean the content or data created by companies will be large and easy to handle with?
Actually，I mean, in downloading, the content can be stored to the "hot server" in advance. But, in uploading, when and where the end users request to upload is random.

>Uploading: UATN should assign the best edge server and select a optimal uploading route to serve a uploading.
>Can we just enable "temporary storage" at the nearest node?

In some case, IMO,  that may be OK, like "hotspot" .

Best wishes

Xiaowei
  





 
发件人： Fei Song
发送时间： 2016-01-08 22:32
收件人： [e-mail]; storagesync
主题： 回复: Re: [Storagesync] recent issues discussed (plain text)
Dear Xiaowei
 
Thanks for your email. Please see my reply inline
 
--------------
Fei Song
>Hi,Fei
>IMO, the main work includes:
>1. Identification of pre-uploaded content and Redirection of the uploading request: 
>     As we known, in downloading, URL/URI  -->  "DNAME” of the DNS  -->  uCDN
 
Is this a traditional and normal way to download a file from a storage server?
It seems that you want to describe it from clients' perspective.
User request URL, then DNS and CDN will help the client machine to finish the downloading
 
>Uploading: How to identify the pre-uploaded content and redirect the request to UATN?
 
There are already some discussions about matadata and Etag. Might be helpful in this issue.
 
>2. Move the random & mass content generated by end users efficiently
>Downloading: content generated by companies, large and manageability
 
You mean the content or data created by companies will be large and easy to handle with?
 
>Uploading:  When and where end users upload content is random, mass and relatively small
 
Indeed. it is hard to estimate the pattern of usage.
 
>3. Best edge server and optimal uploading route 
>Downloading: moving content to the hot edge server in advance.
 
It will be possible if users' moving trajectory and content's name is available before requesting.
 
>Uploading: UATN should assign the best edge server and select a optimal uploading route to serve a uploading. 
 
Can we just enable "temporary storage" at the nearest node?
 
>
>Regards
>
>Xiaowei Qin
> 
>--------------
>Fei Song
>>Hi,Fei>If I did not get wrong, this issue is complicated and should be seperated into multiple steps. >like protocol improvement, deployment optimization, etc.I deeply agree with you. That is a great problem, but, IMO, is worth doing.  If needed, I will outline what should we do about that.Best wishes!Xiao wei Qin
> 
>That would be great! Please do that. then we can discuss in depth via mailing list.
>Thanks for that!
> 
>>
>> 
>>
>>
>>
>>