Hitachi Data Systems Geek Day 0.9

http://www.hds.com/go/geekday/index.html

It is all about talking and talking about the greatness of Hitachi storage innovations.

the next Hitachi Data Systems' USP refresh

SearchStorage.com: So what can we expect HDS to focus on with the next big USP refresh?

Yoshida: Well, I'll give you a hint. We've announced ourdynamic provisioning capability. When we implemented dynamic provisioning, we created the concept of a page — you get a pool of RAID groups and divide them up into pages, and do thin provisioning based on the page dynamically because it's all pre-formatted. This page concept is game-changing for us. Now we will manage storage based upon pages rather than volumes and files. That, I think, will be a focus on future products.

http://searchstorage.techtarget.com/news/article/0,289142,sid5_gci1415232,00.html

Oracle and Informix I/O profiling - 1

This is an interesting subject in database and storage industry. Today number of I/O auto optimization techniques are evolving such as Oracle ASM at the database front and Dynamic/Thin provisioning at the storage front. Manual I/O designing and optimization for database are already a legacy subject?

Sometimes, we have to cope with the legacy. So I start listing the possible physical storage structure of Oracle and Informix.

Trick is the design of LUN (or volume)? Size, RAID level, Number of spindles … for different I/O profiles of database structures. Theory knowledge on database product and experience may come in handy here.

List of High-End/Enterprise storage systems

One of my customers asked me a list of enterprise storage systems. Interestingly I could list only five :-) This is the full list.

• 3PAR InServ T400
• 3PAR InServ T800
• EMC Symmetrix DMX-4
• EMC Symmetrix V-Max
• HDS USP VM
• HDS USP V
• HP StorageWorks XP20000
• HP StorageWorks XP24000
• IBM DS8100 Turbo
• IBM DS8300 Turbo
• IBM DS8700 Turbo
• NetApp (FAS Series) FAS6040
• NetApp (FAS Series) FAS6080
• Sun StorageTek 9985V
• Sun StorageTek 9990V

IT Infrastructure Solution Evaluation

Selecting an IT infrastructure solution is always a challenge. I was listing some of the possible criteria to differentiate solutions. I have seen many such solution selections, some of them are decided on personal/organizational relationships, political reasons :-). Whichever the key decision criteria, some technical backing is mandatory. May be this list (I will keep on adding/modifying it) would help someone in some way.

1. System Requirements and Compliance – Answers are binary type – (Yes/No)

Sample criteria:-

• Need a certified Unix OS?
• Need to comply with regulations for digital records (Sarbanes-Oxley Act, Securities and Exchange Commission rules 17a-3 and 17a-4 and the Health Insurance Portability and Accountability Act (HIPAA))?
• Need to comply with "EU Code of Conduct for Data Centers”?
• Need to comply with "TIA-942 Datacenter Infrastructure Standard”?

2. Solution Architecture Features – Answers are descriptive – (% marks can be given)

Sample criteria:-

• How far solution logically tired?
• How far solution physically tired?
• How far horizontally scalable?
• How far vertically scalable?
• How far inter-dependencies between building blocks (servers, VMs, storage arrays/trays ...etc) avoided?
• How easy to manage the solution?
• How much control and tracking enabled for the solution?

3. Solution Acceptance (How Well-known, proven)

Sample criteria:-

• How many professionals/customers approved/implemented?
• Presence of similar solution implementations?

Typical model of a Data Center - local mid/enterprise customer

Above diagram is a typical model of a data center - local mid/enterprise customer.

Local SMB customers developing data center for production and DR sites are indecisive with the network design. Once they decide routers, switches and cabling arrangements, it is hard to change the network architecture here and there.

When it comes to a new installation, customer needs to plan the correct place for the new infrastructure in his data center model.

Installation engineer has three repetitive network related questions for any customer.

- What is your Management network IPs for the new servers/storage?

- What is your User Access LAN IPs (hostnames, gateways, network ...etc) ?

- What is your Private Data (inter server/nodes) network IPs (cluster interconnects …etc)?

Essentially, any data center should have above three Ethernet networks with required expandability, isolation and access control (security).

This is one of the simple facts in data center architecting and there are many more such as edge/core design, end of the rack / top of the rack and list goes on.

I/O data gathering and I/O analysis

I had to do a data gathering for a Storage proposal done by a storage vendor. I was simply doing a passive work instructed by some other party. This makes me thinking how well we can gather I/O data and analyze them. So I start writing down a list of thing which came to my mind. From my old readings, I know I/O analysis and sizing is a tricky subject. There are two contradictory views on the subject as well. Old view is that I/O analysis and sizing is an art, it is interpreted as a mathematical work (not an art) as well.

I have done very few such and most of them are primitive analysis for study purposes, so I really don’t know the subject very well. The lecture notes in http://www.cs.berkeley.edu/~pattrsn/252S98/Lec13-queuing.pdf helps us understanding interesting principles of queuing theory in I/O.

So here are some tips to do an I/O data gathering. I am hoping to write about a practical I/O analysis case as well.

At which points can we track the I/O data? -> "Table 1 : Typical I/O Trip"

By which methods can we start gathering I/O data for a proper analysis ? -> "Table 2 : I/O data gathering"

Ideas on IT (Infrastructure) vendor selection / How to select an IT (Infrastructure) vendor

This post is about a business aspect of IT Infrastructure. IT vendor selection process is becoming more and more complicated with technology advances and stiff competition between vendors.
Here is sample process I used to pick and choose an IT infrastructure vendor while I was working for a Telecom client.

Architecture Suitability Criteria
 Architecture of the Main Solution
 Architecture of the Backup Solution
 Architecture of the DR Solution
 End of Life (EOL) and End of Support (EOS) of systems
 Platform and System Familiarity

Vendor Suitability Criteria
 Technical Support
 Human Resource pool (Technical)
 Reference Sites
 Past experience with us

Commercial Evaluation Criteria
 Hardware and Implementation
 Maintenance (5 years NPV)
 Implementation Time
 Support & Maintenance Mechanism
 Support for Version Upgrades and New Modules

SAP IT Infrastructure Landscape – Example Design (Cont.)

IT Landscape designing for SAP gains a good experience in IT Infrastructure designing. We can see how some basic designing concepts (like below) behave and work in practice.

• N (Eg:- 3) tier architecture


• N (Eg:- 3) systems landscape


• Application load balancing


• High Availability


• Scalability (Horizontal and Vertical)


• Seamless upgrades and expansions

SAP is a superior ERP product. It has a clear cut guides and best practices in “IT architecture designing, IT architecture sizing”. This is compared to the many other products with no sizing guides, architecture guides.

Nice thing about the SAP’s architecture is that, if you size correctly with future buffers, expanding a SAP IT landscape is seamless. Such an example scenario is given here in this post – “SAP IT Infrastructure Landscape – Example Design”

SAP IT Infrastructure Landscape – Example Design (Cont.) – SAP HW setup after a new acquisition

SAP IT Infrastructure Landscape – Example Design – SAP HW setup for initial phase

Sun (Solaris) Cluster Compliance

Following is the summary for checking Hardware (Server and Storage) compliance for Solaris Cluster.

Server (Hardware) Compliance:-

Sun Servers:-
Most of the Sun servers (even some models of Sun Blades) are supported -http://www.sun.com/software/solaris/cluster/specs.xml

Third party (non Sun) Servers:-
According to “Solaris Cluster Open Hardware Program” -http://www.sun.com/software/cluster/ohp/index.jsp
Only a set of “HP ProLiant x64” are supported - http://www.sun.com/software/cluster/ohp/hp_proliant_interop.jsp
In short “Solaris Cluster is currently not supported on IBM, DELL, Intel servers. Only some of the HP ProLiant x64 servers supported” by November 2009.

Storage (Hardware) Compliance:-

http://www.sun.com/software/solaris/cluster/specs.xml

http://www.sun.com/software/cluster/osp/

Sample set of parameters - Solaris I/O - UFS Tuning

Some of I/O related /etc/system parameters are listed below, This system is a 32GB Solrais x86 server, Always refer to the "SolarisTunable Parameters ReferenceManual" before applying to the systems.

• set maxusers=2048 <- Number of concurrent users on the system determining the amount of physical memory allocated to the kernel (lot of memory and relatively few running processes can save system memory when the default value of maxusers is reduced)
• set ncsize=12000000 <- Defines the number of entries in the directory name look-up cache(DNLC) , use the kstat -n dnlcstats command to determine when entries have been removed from theDNLC because it was too small
• set maxphys=8388608 <- the maximum size of physical I/O requests. If a driver encounters a request larger than this size, the driver breaks the request into maxphys sized chunks
• set pt_cnt=4096 <- When you want to explicitly control the number of users who can remotely log in to the system.
• set npty=176 <- npty limits the number of BSD ptys
• set sadcnt=8192 <- nautopush should be set to twice sadcnt
• set nautopush=4096 <- Set to twice the value of pt_cnt
• set rlim_fd_max=1048576 <- Hard limit on the open file descripters for a single process
• set autoup=600 <- Controls the frequency in seconds of entire physical memory to be scanned for dirty pages
• set tune_t_fsflushr=1 <-Reduces fsflush daemon overhead
• set ufs <-ufs_HW=67108864 <- Limit the amount of I/O that may be outstanding to a single file on a system-wide basis
• set ufs <-ufs_LW=4194304 <- When ufs_HW bytes are outstanding, I/O requests will sleep until less than ufs_LW bytes are outstanding
• set rlim_fd_cur=524288 <- the "soft" limit on file descriptors that a single process can have open
• set segspt_minfree=12500 <- Identifies pages of system memory that cannot be allocated for ISM shared memory
• set maxpgio=65536 <-maximum number of page I/O requests that can be queued by the paging system
• set fastscan=65536 <- maximum number of pages per second that the system looks at when memory pressure is highest
• set segmap_percent=32 <- Amount of memory used, to hold file system pages, % from the physical memory

Sun Communications Suite – Sample Implementation Diagram