Does SDS scale?

Question

Can SDS v1 scale beyond a <10GB single instance?

http://blogs.conchango.com/simonmunro/archive/2009/03/31/sql-data-services-does-not-scale.aspx

Answer 1

Greetings,

I will look into it deeper (as its an interesting view).

but for my understanding:
- partition can be 10 GB in size and the database can be N partitions.

so if we got 20 GB's of financial records we need to shard it over 2 partitions.
and the 2 partitions will go on different instances, depending on their activity they will be moved to appropriate server.

now we also got 2 GB of users, 5gb of costumer records.
these will be 2 partitions, however they will be hosted on the same PC.

So while 10gb is a "hard" limited for a single instance, its only for 1 partition. as SDS will spread its data over multiple instances with 1 entry point to keep track of the partitions and instances.

This Entry point has now 4 partition registered, with 3 seperated endpoints linked to the partition.

This is, to my knowledge, how the speed and replication is keep in tracks with  even the largest database.

Kind Regards,
Sebastian

Answer 2

Thanks Sebastian.

I don't doubt that SDS can scale out if multiple partitions are used.  The challenge is ACID across partitions for your average database developer.  I do hope that someone from MS helps clarfiy this because without a clear understanding of scalability, how can we architect systems to use SDS?

Simon

Answer 3

Check out the latest video from Tech-Ed.  http://www.msteched.com/online/view.aspx?tid=1878a3f2-53a9-4c74-a4d7-d2cc7e4de70c

I think it will answer your question.  I had the same issues but after rethinking how the app works I can get things in multiple servers, data providers, etc. without any re-coding of the app.  Existing apps are harder to scale then new ones of course.

---

Aaron Stanley King Microsoft Consultant Perpetual Technologies, Inc. http://www.perptech.com

Answer 4

We are offering a 1GB SKU and 10GB SKU. You can partition your data based on the needs and our guarantees for scalability and consistency still apply.  For Eg: for 20GB data, you can partition the data in to two 10GB databases and they may be placed on two different machines for scalability. I would encourage you all to read our blog http://blogs.msdn.com/ssds/default.aspx for more details. We are also targeting to offer a public CTP during late summer, please stay tuned for updates.

Thanks,
Abi (MSFT)