3

I'm trying to find a home for three ASP.NET 3.5 websites using three respective SQL Server 2008 databases. I've found one possible solution where I host the websites on a "semi-dedicated" environment with one specific provider. The problem is, the database piece of this service is lacking. So, I've found an additional solution where I host the databases with a different provider. This keeps me within budget and with two providers I respect a great deal. Unfortunately, neither provider gives me both site and database hosting as I'd prefer in one package.

My question is this:

1) Is there an definite downgrade of service if I separate the site hosting from the database hosting? As in, before I even test the possible latency, is this a bad idea? Has anyone done this successfully already?

2) Is there a good way to test the SQL Server response between those databases hosted on the same provider vs those hosted with a different provider? I'm not well-versed in testing site and database response over the Internet, so I'm vague on where to start.

Randy Burgess
  • 277
  • 1
  • 2
  • 6

2 Answers2

4

Having 2 hosting providers that each host a critical part of your site will double the chances of downtime. If either fails, your site is down. That's the biggest concern.

Latency will be your second concern. More and more data is moving to cloud models, so it's not out of the question to separate the roles into different providers, but it does introduce considerable latency.

To get a handle on performance you can eyeball it with SSMS locally and remotely. If you have tiny data sets, it may handle it well, but once you start pulling lots of data or performing many queries per page, performance can drop quickly.

For stress testing tools, the MS solutions are WCAT or Visual Studio Team System. You can also output to your page the timestamp before making a database connection, after, and as you pull back data from the database. Try that with LAN and WAN data to compare.

Ultimately, I would recommend finding a host that can handle both parts well. When they can support you end to end, they don't point fingers at each other when you try to narrow down performance issues.

Scott Forsyth
  • 16,449
  • 3
  • 37
  • 56
  • +1. Also consider the additional data transfer cost of moving all request data between 2 providers. If your providers have data traffic caps or scales, you'll be paying for a *lot* of extra data flow both way on both hosts. – Chris Thorpe Feb 10 '10 at 20:06
  • yeah, I didn't consider the extra bandwidth...but in this case, I don't think it would hurt me as much as latency. – Randy Burgess Feb 11 '10 at 00:03
1

1) Is there an definite downgrade of service if I separate the site hosting from the database hosting? As in, before I even test the possible latency, is this a bad idea? Has anyone done this successfully already?

Absolutely. I did this. For a period of about 2 weeks then I quickly realised it wasn't going to work.

2) Is there a good way to test the SQL Server response between those databases hosted on the same provider vs those hosted with a different provider? I'm not well-versed in testing site and database response over the Internet, so I'm vague on where to start.

It's a bit difficult to do this unless you can replicate the set up without getting yourself into a contract. Perhaps ask for a 30-day trial, or get a rolling monthly contract and cancel after the first month. This is what I did with my most recent provider, and I'm happily still with them.

As for how to test it, SSMS has a neat feature called "Client Execution Statistics" - it's a button on the taskbar when you've got a query tab open. Turn this on, then execute the query, and it gives you a great breakdown of how much traffic was passed, how long it spent waiting for the server to reply etc.

Do this on on remote site->db servers, and local site->db servers and compare the figures.

Andy Shellam
  • 1,878
  • 1
  • 12
  • 16