Tuesday, December 4, 2007

Enclosure Types







Enclosure Types


The above page diagrams the back-end structure of a Clariion. How the disks are laid out. Before we discuss the back-end bus structure, we should discuss the different types of enclosures that the Clariion contains.


1.DAE. The Disk Array Enclosure. Disk Array Enclosures exist in all Clariions. DAE’s are the enclosures that house the disks in the Clariion. Each DAE is holds fifteen (15) disks. The disks are in slots that are numbered 0 to 14.


2.DPE. The Disk Processor Enclosure. The Disk Processor Enclosure is in the Clariion Models CX300, CX400, CX500. The DPE is made up of two components. It contains the Storage Processors, and the first fifteen (15) disks of the Clariion.


3.SPE. The Storage Processor Enclosure. The Storage Processor Enclosure is in the Clariion Models CX700 and the CX-3 Series. The SPE is the enclosure that houses the Storage Processors.



The diagrams above lay out the DAE’s back-end bus structure. Data that leaves Cache and is written to disk, or data that is read from disk and placed into Cache travels along these back-end buses or loops. Some Clariions have one back-end bus/loop to get data from enclosure to enclosure. Others have two and four back-end buses/loops to push and pull data from the disks. The more buses/loops, the more expected throughput for data on the back-end of the Clariion.


The Clariion Model on the left is a diagram of a CX300/CX3-10 and CX3-20. These models have a single back-end bus/loop to connect all of the enclosures. The CX300 will have one back-end bus/loop running at a speed of 2 GB/sec, while the CX3-Series Clariions have the ability to run up to 4 GB/sec on the back-end.


The Clariion Model in the middle is a diagram of a CX500. The CX500 has two back-end buses/loops. This gives the CX500, twice the amount of potential throughput for I/Os than the CX300.


The Clariion Model on the right is a diagram of a CX700, CX3-40 and CX3-80. These Clariions contain four back-end buses/loops. The CX3-80 will contain the maximum back-end throughput with all four buses having the ability to run at a 4 GB/sec speed.


Each enclosure has a redundant connection for the bus that it is connected. This is in the event that the Clariion loses a Link Control Card (LCC) that allow the enclosures to move data, or the loss of a Storage Processor. You will see one bus cabled out of SP A and SP B, allowing both SP’s access to each enclosure.



Enclosure Addresses


To determine an address of an enclosure, we need to know two things, what bus it is on, and what number enclosure it is on that bus. On the Clariions in the left diagram, there is only one back-end bus/loop. Every enclosure on these Clariions will be on Bus 0. The enclosure numbers start at zero (0) for the first enclosure and work their way up. On these Clariions, the first enclosure of disks is labeled Bus 0_Enclosure 0 (0_0). The next enclosure of disks is going to be Bus 0_Enclosure 1 (0_1). The next enclosure of disks 0_2, and so on.


The CX500, with two back-end buses will alternate enclosures with the buses. The first enclosure of disks will be the same as the Clariions on the left of Bus 0_Enclosure 0 (0_0). The next enclosure of disks will utilize the other back-end bus/loop, Bus 1. This enclosure is Bus 1_Enclosure 0 (1_0). It is Enclosure 0, because it is the first enclosure of disks on Bus 1. The third enclosure of disks is going to be back on Bus 0, 0_1. The next one up is on Bus 1, 1_1. The enclosures will continue to alternate until the Clariion has all of the supported enclosures. You might ask why it is cabled this way, alternating buses. The reason being is that most companies don’t purchase Clariions fully populated. Most companies buy disks on an as needed basis. By alternating enclosures, you are using all of the back-end resources available for that Clariion.

The Clariions on the right show the four bus structure. The first enclosure of disks is going to be Bus 0_Enclosure 0 (0_0) as all other Clariions. The next enclosure of disks is Bus 1_Enclosure 0 (1_0). Again, using the next available back-end bus, and being the first enclosure of disks on that bus. The third DAE is going to be Bus 2_Enclosure 0 (2_0). The fourth DAE is on the fourth and last back-end bus. It is Bus 3_Enclosure 0 (3_0). From here, we are back to Bus 0 for the next enclosure of disks. Bus 0_Enclosure 1 (0_1). The next DAE is 1_1. The next would be 2_1 if we had one. 3_1, 0_2, and so on until the Clariions were fully populated.


Disk Address


The last topic for this page are the disks themselves. To find a specific disk’s address, we use the Enclosure Address and add the Slot number the disk is in. This gives us the address that is called the B_E_D. Bus_Enclosure_Disk. The Clariion on the left has a disk in slot number 13. The address of that disk would be 0_2_13. The Clariion in the middle has a disk in slot number 10 of Enclosure 1_1. This disk address would be 1_1_10. And the Clariion on the right has a disk in Bus 2_Enclosure 0. It’s address is 2_0_6. And the disk in Bus 1_Enclosure 1 is in slot 9. Address = 1_1_9.


Finally, each Clariion has a limit to the number of disks that it will support. The chart below the diagrams provides the number of how many disks each model can contain. The CX300 can have a maximum of 60 disks, whereas the CX3-80 can have up to 480 disks.


The importance of this page is to know where the disks live in the back of the Clariion in the event of disk failures, and more importantly how you are going to lay out the disks. Meaning, what applications on going to be on certain disks. In order to put that data onto disks, we have to create LUNs (will get to it), which are carved out of RAID Groups (again, getting there shortly). RAID Groups are a grouping of disks. To have a nice balance and to achieve as much performance and throughput on the Clariion, we have to know how the Clariion labels the disks and how the DAE’s are structured.

Cache WaterMarks






WaterMarks


WaterMarks is what controls writing data out of Cache to disk. It is used to manage how long data stays in Write Cache before it is written to disk.


This diagram is used to describe the types of “Flushing” data to disk, or writing data out of Cache to disk.



The first type of Write Cache Flushing is Idle Flushing.


Idle Flushing is when the Clariion has the ability to take the ‘writes” into cache, send the acknowledgement back to the host that the data is on “disk.” While this is happening, the Clariion can also write data out to disk. The Clariion will try to write to disk in a 64 KB “Chunk.” The cache is absorbing the writes, grouping them together, and writing them to disk. This will come into play later when we discuss how the Clariion formats the disks. This is the perfect case scenario. The Cache takes in the writes, the Clariion has the resources to write the blocks to disk.



The second type of Flushing is WaterMark Flushing.


This is maintained by percentages that you can configure in Cache. The goal with WaterMark Flushing is to keep the Write Cache level between these two percentages. We are using the default Low WaterMark Setting of 60%, and High WaterMark Setting of 80%. These can be changed, and we will discuss that later. With WaterMark Flushing, Cache is going to do it’s best to keep Write Cache between these two levels. As Write Cache hits the High WaterMark, the Clariion tries to flush down to the Low WaterMark. If the amount of Write Cache is constantly between these two levels, the Clariion is doing its job.



The last type of flushing is the “Forced Flush.”


A Forced Flush of Cache results in the Write Cache reaching capacity. The Clariion will no longer accept data into write cache, as there is no more room.


When a Forced Flush occurs, the following take place:


1. The Clariion disables Write Cache.
2. The Clariion begins to destage/flush the write data in Cache out to disk.
3. Now comes the performance issue. With the Clariion disabling Write Cache, any new writes that come in from a host will bypass cache and be written directly to disk. The host/application is now waiting for the acknowledgement to return after the data was written to disk.
4. The Clariion will keep Write Cache disabled until it flushes to the Low WaterMark.
5. Once Write Cache is flushed to the Low WaterMark level, Write Caching is automatically re-enabled.