Fault Tolerance?

[marc at corky.net]

Last time I checked was around a year ago.  I don't remember the exact 
figures but it was *very* slow.   Since things may have changed since 
then, you should run tests with a fresh version.  The latency may very 
well be sufficient for your app.

MySQL 5.1, which includes NDB, is indeed still very much a beta.


Peter Chiu wrote:
> Ok Marc. Sounds like you have experience there.. could you share how slow it is? Thanks.
>
> Best,
> Peter
>  
>
> -----Original Message-----
> From: marc at corky.net [mailto:marc at corky.net] 
> Sent: Friday, September 28, 2007 5:01 PM
> To: Peter Chiu
> Cc: Mdecandia; memcached at lists.danga.com
> Subject: Re: Fault Tolerance?
>
> You would be surprised how *SLOW* it is to fetch even the most simplest 
> objects from NDB.   This has to do with the way mysql-cluster is 
> designed and cannot be overcome so easily.
>
> Forget about using NDB as a cache.
>
>
> Peter Chiu wrote:
>   
>> Totally agreed. What about building a HA memory cache based on NDB, running on machines with plenty of RAM (64GB)? That way NDB would cache most of the stuff in memory for fast access, while at the same time provide HA. Ideas?
>>
>> Best,
>> Peter
>>  
>>
>> -----Original Message-----
>> From: memcached-bounces at lists.danga.com [mailto:memcached-bounces at lists.danga.com] On Behalf Of Mdecandia
>> Sent: Friday, September 28, 2007 3:50 PM
>> To: memcached at lists.danga.com
>> Subject: Re: Fault Tolerance?
>>
>>
>> On Sep 27, 2007, at 22:16, Dustin Sallings write: 
>>   
>>     
>>> On Sep 27, 2007, at 21:56, Paul Scott wrote:
>>>
>>>     
>>>       
>>>>> Agreed. Wouldn’t it be great though to have a mem-based HA datastore?
>>>>>         
>>>>>           
>>>> I would certainly vote +1 on that idea!
>>>>       
>>>>         
>>> 	You do realize you wouldn't have anything remotely like the  
>>> performance of memcached, don't you?  You'd need something along the  
>>> lines of two-phase-commit if you want any kind of correctness.  If  
>>> you don't want correctness, then why are you worried about HA?
>>>
>>>     
>>>       
>> We use memcached to store items fetched from a slow service, not from
>> database. 
>> Performance difference between this service and memcached are huge.
>> May be really convenient to have an HA feature on memcached servers.
>>
>>   
>>     
>>> 	If you lose a node, how do you plan on rematerializing?  A complete  
>>> synchronization would block both nodes in a two-node cluster.
>>>     
>>>       
>> We've introduced libketama consistant hashing to reduce effects on server
>> faults but 
>> it will be useful to have a redundant caching system between servers to be
>> really fault toulerant.
>>
>>   
>>     
>>> 	How would you handle conflicts during rematerialization after a  
>>> netsplit?
>>>
>>> 	Is it acceptable to block all clients during a netsplit (pending  
>>> some sort of magical synchronization that knows what to do when  
>>> conflicts occur)?
>>>
>>> 	After you get all of the pieces in place, are you sure you'd have  
>>> something that would be any faster than any solution that isn't  
>>> completely in-memory?
>>>
>>> -- 
>>> Dustin Sallings
>>>     
>>>       
>>  Michele 
>>
>>  
>>  --
>>  Email.it, the professional e-mail, gratis per te: http://www.email.it/f
>>  
>>  Sponsor:
>>  Problemi di Liquidità? Con Logos Finanziaria 30.000 € in 24 ore a
>> dipendenti e lavoratori autonomi con rimborsi fino a 120 mesi clicca qui
>>
>>  Clicca qui: http://adv.email.it/cgi-bin/foclick.cgi?mid=2907&d=20070928
>>
>>
>>   
>>     
>
>