Tilted Forum Project Discussion Community  

Go Back   Tilted Forum Project Discussion Community > Interests > Tilted Technology


 
 
LinkBack Thread Tools
Old 10-13-2004, 06:06 AM   #1 (permalink)
Insane
 
Location: Michigan
MS SQL Server Performance/Redundancy?

Ok, a little background. We have about 1000 clients that hit our servers hard. We have 7 or so webservers and 4 database servers. Right now, if a database server has to be rebooted, all clients on that machine are down for the length of the reboot. Even scarier, if anything on any of the machines dies, all the clients on that database server is down until we fix the server! Other times, a single server for a group of clients gets overloaded and we currently move their db to another server. We need something that can let the extra load spill over to another, less busy machine.

We need redundancy somehow. I want to be able to upgrade hardware on the machine, install MS updates, or know that if a cpu dies that our clients won't be left in the dark. What can help this? Clusters? Replication?

Will replication work? If so, how would you tell the scripts on the webservers to look at another database server if 1 went down?

I know clusters can be super expensive if you get a SAN, but is there another way that will work with less money? A machine with raid 0+1 for 500gb hd redundant space that does nothing but filesharing with gigabit ethernet to each machine?

I'm not limiting the possible solution to clusters or replication. If you have any other suggestions, please, throw them out.

Thanks.
asshopo is offline  
Old 10-13-2004, 08:07 AM   #2 (permalink)
Professional Loafer
 
bendsley's Avatar
 
Location: texas
Windows Server 2003 provides several methods of improving system- or server-level fault tolerance by using a few of the services included in the Enterprise and Datacenter platforms.

Also, if you build machines that are there strictly for replication, you should just be able to change your router to point to the new machine, assuming it's replicating data on the fly.

This link might help you some too.

I do have one question for you though, what RAID are you using? We run a quad xeon machine with MS SQL on it in RAID 1 over 6 harddrives in three arrays. We have found that for SQL, RAID 1 is actually faster than the RAID 5 alternative. Everything else we have uses the RAID 5.
__________________
"You hear the one about the fella who died, went to the pearly gates? St. Peter let him in. Sees a guy in a suit making a closing argument. Says, "Who's that?" St. Peter says, "Oh, that's God. Thinks he's Denny Crane."

Last edited by bendsley; 10-13-2004 at 08:10 AM..
bendsley is offline  
Old 10-13-2004, 08:24 AM   #3 (permalink)
Insane
 
Location: Michigan
Quote:
Originally Posted by bendsley
Windows Server 2003 provides several methods of improving system- or server-level fault tolerance by using a few of the services included in the Enterprise and Datacenter platforms.

Also, if you build machines that are there strictly for replication, you should just be able to change your router to point to the new machine, assuming it's replicating data on the fly.

This link might help you some too.

I do have one question for you though, what RAID are you using? We run a quad xeon machine with MS SQL on it in RAID 1 over 6 harddrives in three arrays. We have found that for SQL, RAID 1 is actually faster than the RAID 5 alternative. Everything else we have uses the RAID 5.
Awesome link, thanks! Some of our database servers don't even have raid, and I believe our others have raid 0+1. But also, as I said, sometimes a single database server overloaded and would be nice to let the excess load spill over to another, less active machine. Can this only be done thru clustering?

A buddy made a suggestion of having a load balancer split up the work between, say 4 machines, and have them all replicate. Should 1 machine go down, the load balancer would see this and not use that machine until it sees that it is up again. Would this work?
asshopo is offline  
Old 10-14-2004, 05:41 AM   #4 (permalink)
Professional Loafer
 
bendsley's Avatar
 
Location: texas
Yeah, thats really what load balancing is. You would have to spend some money, possibly setup some more drives and machines, but the load balancer would detect faults coming through and route them accordingly to the fastest machine. Once the machine that was the fastest has a lot more traffic coming to it, it's obviously not the fastest machine anymore and the traffic gets routed elsewhere.

I would also hope that the machines you have running Raid 0 have tape backups for them? That would be nice if they aren't in place.

What might need to happen is one night you tell your customers that you need to save their sql tables and bring down the servers. Cluster all of the machines and let them replicate on the fly. Enable load-balancing and go from there. I believe that MS SQL can be load balanced, in that it doesn't use shared memory on one machine (ie. they would use memory on the local machine).

Another suggesting might be that you actually just build a large NAS through Fibre Channel connections and store all of your data on there. Have all of the traffic balanced on several machines and have MS SQL point to the NAS. If you do this, your NAS will generate considerable heat considering the drives would be moving most all the time (if you're really getting hit a lot). At least this way you would gain redundancy and have speed through Fibre Channel.
__________________
"You hear the one about the fella who died, went to the pearly gates? St. Peter let him in. Sees a guy in a suit making a closing argument. Says, "Who's that?" St. Peter says, "Oh, that's God. Thinks he's Denny Crane."
bendsley is offline  
Old 10-14-2004, 08:15 AM   #5 (permalink)
Insane
 
Location: Michigan
Any machines we have that have raid 0 also have raid 1 (0+1) in that there is a mirror of the raid 0 array.

As for the on the fly replication.. what happens if the same record is updated on 2 machines at the same time. For example, we have a counter field to see how many times an image has been loaded. Is the lag of waitng for the replication going to affect the data? I know the replication is on the fly, but say you have 4 machines load balanced and all are at 80% useage.. its not going to be able to instantly replicate data between all the servers and there is a possiblity of something not getting updated properly.

Thanks for the help, there doesn't seem to be alot of help out there on this subject.
asshopo is offline  
Old 10-14-2004, 11:01 AM   #6 (permalink)
Professional Loafer
 
bendsley's Avatar
 
Location: texas
What would happen is that all of the servers would be replicated first. All would have the same data. They need to have the same amount of space on them for this to work effectively. Anyway, all servers would have the same data and when something is added, all of the servers have the same access to the same information. The load-balancing comes in to where the "balancer" tells the application which server to route to.

What I would suggest is putting all of the data in one central location, instead of on several different servers. Then, on each of your servers, you have SQL loaded and tell it where the data store is located (NAS or what-have you). The SQL on every server will always pull the same info from the same place, but the SQL server is being load-balanced for how to handle the information.

Data storage would be easy since all you're putting into it is text from SQL. What is power hungry is how what you do with the SQL info, or how MS SQL uses the info. MS SQL is quite a power hungry/memory hungry program.
__________________
"You hear the one about the fella who died, went to the pearly gates? St. Peter let him in. Sees a guy in a suit making a closing argument. Says, "Who's that?" St. Peter says, "Oh, that's God. Thinks he's Denny Crane."
bendsley is offline  
Old 10-14-2004, 12:09 PM   #7 (permalink)
Insane
 
Location: Michigan
Sounds like an idea. Only problem that might occur is file locking. Doesnt MS SQL lock the data and log files? If it does, how do the other machines connect to those data/log files? I did a bit of research and they say that multiple machines can not read/write to the same mdf file, however, reading would be fine. This wont work for us as we need to load balance primarily writing/updating of data.
asshopo is offline  
Old 10-14-2004, 12:17 PM   #8 (permalink)
Professional Loafer
 
bendsley's Avatar
 
Location: texas
You might end up just having to buy larger machines then to handle the load. Ours is running Quad Xeon 3.2Ghz with 2mb of cache on each processor and 8gigs of ram.

Are you machines comparable to this?
__________________
"You hear the one about the fella who died, went to the pearly gates? St. Peter let him in. Sees a guy in a suit making a closing argument. Says, "Who's that?" St. Peter says, "Oh, that's God. Thinks he's Denny Crane."
bendsley is offline  
Old 10-16-2004, 01:44 AM   #9 (permalink)
Upright
 
Location: Berkeley, CA
I'm not in the IT dept, but I believe the strategy my company uses is something like this:
1. cluster two or four servers all sharing the same SAN disk array and with load-balancer
2. 2PC replication to a standby cluster
3. log-shipping to a remote standby server cluster every five min
4. daily full-backups to remote disk
5. daily full/diff backups to tape

Sounds like overkill, but we haven't had any major problems yet.
littlebighead is offline  
 

Tags
performance or redundancy, server, sql

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On



All times are GMT -8. The time now is 09:29 PM.

Tilted Forum Project

Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2024, vBulletin Solutions, Inc.
Search Engine Optimization by vBSEO 3.6.0 PL2
© 2002-2012 Tilted Forum Project

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360