|
|
 |
 |
 |
 |
 |
 |
|
|
|
From: June Mullins (junemullins, earthlink dot net) Date: 2001.10.19 - 09.26 MDT
Perry's added some good points - I obviously didn't understand the whole
story about the portal. Let's discuss this further at the meeting on
Tuesday.
June
Perry and Lorae Merritt wrote:
> I read through Junes revisions. I inserted my comments in the doc.
>
> Hopefully I'm misreading it or June was sick while she wrote it.
>
> P
>
> ----- Original Message -----
> From: "June Mullins" <junemullins, earthlink dot net>
> To: <antera, cryptofreak dot org>
> Sent: Wednesday, October 17, 2001 3:01 PM
> Subject: Arch doc (finally)!
>
>
>> My apologies for taking so long - I was babying myself with my cold.
>> It's almost all better now.
>>
>> I just made a few changes to the arch doc, which I hope make it more
>> clear. I added back some
>> of Perry's words from his original arch doc, because I felt the new
>> version was kind of too vague
>> about what we are actually trying to build. I also added examples in
>> some places for the same
>> reason. Finally, I changed Data Manager to Request Manager, because
>> Persistent Data Manager
>> is too confusing close to Data Manager. If anybody can think of a
>> better name, that's fine too. I'd
>> appreciate comments and amendments to the doc. Even though you guys did
>> a great job of
>> explaining the architectural changes you made while I was gone, I'm sure
>> I could have missed some
>> important points.
>>
>> Here are some questions we came up with last meeting:
>> - Is there a problem with timeouts when we need to restore a file from
>> tape? Do we block on open,
>> and wait until the entire file is restored? Do we block on read, and
>> wait only until we can satisfy the
>> read (i.e., continue to restore the file in the background)? Jay,
>> would it be hard for you to run some
>> experiments with your current driver code - put in a sleep before
>> responding to reads or opens, and
>> see what happens with CIFS, NFS, or local requests?
>> - Can we (should we) take advantage of HSM-aware applications?
>> - Do we need dynamically expandable storage? With RAID protection?
>> - Would it be reasonable to require the customer to have a tape library
>> with more than one tape drive?
>> Two tape drives would be useful for autorestore during disaster
>
> recovery.
>
>> - We need to know more about what the different management policies
>> would be.
>> - What do we want to build for a prototype?
>>
>> I also want everyone to know that I am single-handedly responsible for
>> driving up the price of
>> Tricord stock. I sold all my stock Monday at $1.20. It is now $1.74.
>> It galls me that I did
>> something useful for Tricord.
>>
>> Don't forget - we have a meeting next Tuesday at the library at 1:30.
>>
>> Kevin and Brian - are you employed yet?
>>
>> June
>>
>>
>> Software Architecture-rev2.doc
>>
>> Content-Type:
>>
>> application/msword
>> Content-Encoding:
>>
>> base64
>>
>>
This archive was generated by hypermail 2b30 : 2001.10.20 - 03.00 MDT |
