Gregs Server and StorageIOblog - Latest Comments

Re: Speaking of speeding up business with SSD storage

greg schulz — Sat, 30 Nov 2013 10:29:55 -0000

Here are some additional related posts:

How many IOPS can a HDD, HHDD or SSD do?
http://storageioblog.com/io...

Part II: How many IOPS can a HDD, HHDD or SSD do with VMware?http://storageioblog.com/pa...

RAID and IOPS and IO observations
http://storageioblog.com/ra...

Can we get a side of context with them IOPS and other storage metrics?
http://storageioblog.com/si...

Re: Fall 2013 StorageIO Update Newsletter

greg schulz — Wed, 27 Nov 2013 14:30:24 -0000

Turns out that SNIA and Computerworld/IDG have now announced the end of the SNW event, read more here:

https://www.etouches.com/eh...

http://siliconvalleypr.com/...

Re: Seagate Kinetic Cloud and Object Storage I/O platform (and Ethernet HDD)

Silvercor Data — Mon, 28 Oct 2013 12:09:15 -0000

Thanks Greg for the posting! Check out Silvercor (www.silvercor.com). Silvercor offer virtual disk drives which have been out for a year, run its own hypervisor, also supports all industry standard hypervisors and management tools. With full multicore 64bit processor, 1Gb/10Gb ethernet interface and has an outstanding performance supporting both object and block storage.

Re: Enabling Bitlocker on Microsoft Windows 7 Professional 64 bit

Scott — Mon, 28 Oct 2013 11:29:26 -0000

A few things -

1) TrueCrypt and SSDs do not get along. I wouldn't recommend anyone using it with an SSD.
http://www.truecrypt.org/do...

2) You could use the hardware encryption on your SSD. I know the BIOS should support it. The Lenovo t520 I am currently using does.
http://en.wikipedia.org/wik...

http://www.samsung.com/glob...

Full Disk Encryption:
http://whatis.techtarget.co...

Re: Virtual, Cloud and IT Availability, its a shared responsibility and common sense

Cloud Computing — Tue, 22 Oct 2013 16:42:01 -0000

At the end of the day the blaming has to stop. Providers and users needs to to come up with a workable solution on how they can integrate and make use of cloud without undergoing through so much security issues.

Re: Seagate Kinetic Cloud and Object Storage I/O platform (and Ethernet HDD)

Kalen Kimm — Tue, 22 Oct 2013 13:58:18 -0000

Very interesting indeed! First I heard of it - thanks for the great info.

Re: Cloud conversations: Has Nirvanix shutdown caused cloud confidence concerns?

Greg Schulz — Fri, 18 Oct 2013 11:11:18 -0000

Thanks Greg,

Remember, don't be scared of clouds (or any tech for that matter), however do your homework, be prepared, make smart acquisition decisions including doing due diligence on providers, suppliers, partners from technology along with business perspectives...

Re: Cloud conversations: Has Nirvanix shutdown caused cloud confidence concerns?

Greg Knieriemen — Fri, 18 Oct 2013 10:57:39 -0000

Great post Greg!

Re: Enabling Bitlocker on Microsoft Windows 7 Professional 64 bit

Alex Gubin — Fri, 27 Sep 2013 23:51:05 -0000

BitLocker is available exclusively on Enterprise and Ultimate editions of Win 7. So no, you could not have saved money by using a workaround.

Re: Cisco buys Whiptail continuing the SSD storage I/O flash cash cache dash

Greg Schulz — Wed, 11 Sep 2013 16:31:10 -0000

Here are some additional comments and perspectives pertaining to Cisco/Whiptail:

http://searchsolidstatestor...

Re: VMworld 2013 Vmware, server, storage I/O and networking update (Day 1)

Server Management — Mon, 09 Sep 2013 16:52:02 -0000

There are a lot of things to look forward this year. 10 years of experience and expertise in the VMworld is really a big achievement. Server management companies should at least follow the tips given by this company to be successful in the business.

Re: EMC New VNX MCx doing more storage I/O work vs. just being more

Greg Schulz — Wed, 04 Sep 2013 18:32:51 -0000

Thanks Lee, what will be interesting to see is if more vendors start providing context in terms of what they can do, vs. simply touting more ports, processors, memory and thus assume more performance. The hybrid caching model ironicly was validated a decade or two ago, granted some today are not aware of that (present company excluded ;)...

Re: EMC New VNX MCx doing more storage I/O work vs. just being more

Lee Johns — Wed, 04 Sep 2013 18:27:09 -0000

Nice Summary Greg. EMC is validating the hybrid caching model and it will be interesting to see how this effects the overall market both from the big boys and the new vendors.

Re: Is more of something always better? Depends on what you are doing

greg schulz — Tue, 03 Sep 2013 15:31:05 -0000

Thanks for the comment

Re: Is more of something always better? Depends on what you are doing

Akhil Bhansali — Tue, 03 Sep 2013 15:23:36 -0000

Very well said!!!

Re: Putting some VMware ESX storage tips together (Part I)

greg schulz — Sun, 01 Sep 2013 11:15:06 -0000

Hello Paul, thanks for stopping by and commenting.

You are correct in that in this post there is no benchmark, it was focused on using a trick to make vsphere/ESXi think a non SSD device is a SSD device for functionality purposes. However you can find some benchmarks in other existing posts including this one ( http://storageioblog.com/pa... ) which while not testing or comparing a "tricked" SSD device, they leverage some of the tricks in this post as well as in the part II (found here: http://storageioblog.com/pu... ). Also watch for some future posts that will be doing more with these and other tricks, SSDs, SSHD/HHDDs, performance and benchmarking.

However in the meantime to answer your question of did the HHDD (Hybrid Hard Disk Drive) aka Solid State Hard Disk (SSHD) that I tricked VMware into thinking it is a SSD improve performance? Yes, however the HHDD helps on reads even without doing the VMware SSD trick which is more for functionalty, at least for now ;).

Re: Care to help Coraid with a Storage I/O Content Conversation?

Chris K — Tue, 27 Aug 2013 16:56:39 -0000

As a leading enterprise data management services organisation (UK), Oriium need to stand out from the crowd. Providing ground breaking solutions and highly scalable technology within our stack allows us to compete with our larger counterparts and provides a key differentiator.

Having discovered CORAID during a customer workshop we realised the potential the technology offers; High performance storage capable of direct attached connectivity (removing the requirement for expensive low-latency storage switching) with an infinitely scalable SAN architecture for larger deployments. CORAID is a "glove fit" for data management solutions such as CommVault and compliments application based data de-duplication, replication and indexing technologies.

The CORAID SRX Storage Arrays allow mixed disk formats (SAS, SATA & SSD), partial drive population and a choice of 10Gb/s interfaces (10GbE & SFP+) giving customers complete configuration choice. This provides a storage platform well suited to differing workloads including both Hyper-V and VMware virtualisation technologies.

Operating in the mid-market space is challenging. Customers demand performance, scalability and affordability, attributes that don’t often co-exist. In our experience CORAID delivers performance in spades at the right price point and the SRX 6000 series looks set to continue this ethos.

Oriium (www.oriium.co.uk) is a (channel-only) professional services organisation specialising in "big data" management, enterprise infrastructure services and data intelligence technologies.

Re: Cloud, Virtual, Server, Storage I/O and other technology tiering

greg schulz — Wed, 07 Aug 2013 13:45:29 -0000

Thanks Marc for the comments and good to hear from you.

Concur, there is a subtle difference too often overlooked, confused or simply not known about which is data when transformed becomes information. E.g. the classic computer science reading Algorithms + Data Structures = Programs = Transform = Information... Thus garbage in (or garbage algorithms) results in garbage data structures (whats stored in memory or on storage) = garbage out...

The better the algorithms and associated data structures/organizations the better the results, otherwise just simply pushing, moving, processing and storing, well, you said it...

Hope all is well.

Cheers
gs

Re: Cloud, Virtual, Server, Storage I/O and other technology tiering

Marc D. Johnson — Wed, 07 Aug 2013 13:30:30 -0000

Greg, nicely put. You touch on something that many do not truly understand. The difference between information and data is a subtle but important point. Too many organizations (and the people within) do not treat them differently; therefore, they are heading down a deep and treacherous path to big garbage instead of big data as mentioned in another one of your blogs. I have long believed that machines cannot truly vaulate data as information because there is little context to the value an organization places upon the information (data over time).

Re: Industry Trends and Perspectives: Tiered Hypervisors and Microsoft Hyper-V

greg schulz — Tue, 06 Aug 2013 13:56:00 -0000

Also check out this related post:

Cloud, Virtual, Server, Storage I/O and other technology tiering
http://storageioblog.com/cl...

Re: Can RAID extend the life of nand flash SSD?

greg schulz — Mon, 05 Aug 2013 17:38:47 -0000

Kirill good to hear that you at Starboard are among others that do write and endurance optimization. Whats interesting is that a decade or so ago those types of conversations or level of information was more common, now not discussed much or glossed over. The result is what some take for granted will differ from what various vendors/products can do or proliferate various myths.

Re: Can RAID extend the life of nand flash SSD?

Kirill Malkin — Mon, 05 Aug 2013 16:54:31 -0000

Greg, "Those 32K chunks would then be written to a drive and then processed based on the SSD or HDD for that matter firmware/FTL/etc handles the writes with its own buffering" << exactly how we are doing it! ;)

Re: Can RAID extend the life of nand flash SSD?

greg schulz — Mon, 05 Aug 2013 13:47:58 -0000

Kirill thanks for the comments and perspectives, concur on "it depends".

That is "it depends" on if RAID impacts positive, neutral or negative on nand flash SSD wear based on the implementation, configuration, IO profile/activity among others. You touch on something that many miss or do not realize which is based on implementation, RAID does not have to have the extra IOs (writes) that are often mentioned, or perceived. Otoh, a poor implementation can cause extra writes if less than full stripes are not managed.

Likewise the larger the IO size that can fill or use or leverage a cache line/stripe with the right sized chunk/shard can help, not to mention if the width, chunk and other configuration options are aligned to the IO (writes) arrival rate. If a high IO arrival rate (e.g. how many IOPS) with chunk sized to be a function of the average IO size with properly sized width, you could have a very efficient solution.

Something else that is missed, and this can be a bit of a spin granted is that if you only write to a single drive an entire 1GB file, thats 1GB of writes to that single drive (ok, overly simple, however trying to keep math simple for a Monday morning ;). On the other hand say there are nine drives in a 8+1 R5 scheme with chunk set at say 32K and an IO size of 256K (e.g. optimized for large writes) and RAID algorithm (in sw or hardware/firmware) that can group the writes without incurring a stale read. There would be in the neighborhood of around or just under 4,000 256K IOs arriving to the raid solution (hw or sw) with each of those 256K IOs being placed into the cache line/stripe of 32K size. Those 32K chunks would then be written to a drive and then processed based on the SSD or HDD for that matter firmware/FTL/etc handles the writes with its own buffering. Thus in the case of nand flash whose underlying media prefers larger IOs due to its larger low level page/block size, could benefit depending on implementation.

Granted there are lots of if's and assumptions thus it depends, on the other hand keep in mind that while more drives are used, each drive only handles 1/8 of the data (or parity) for a given stripe/cache line vs. a single drive handling all activity.

Thus it all depends...

Re: Can RAID extend the life of nand flash SSD?

greg schulz — Mon, 05 Aug 2013 13:28:55 -0000

As an FWIW/FYI, here are some links to comments and conversations related to this post/topic occuring over on Linkedin Groups:

In the Storage Group:

http://www.linkedin.com/gro...

In The Virtualization & Cloud Computing Group:

http://www.linkedin.com/gro...

Join in the conversations here or in those among other venues...

Re: Can RAID extend the life of nand flash SSD?

Kirill Malkin — Mon, 05 Aug 2013 12:13:47 -0000

Greg, good article! Completely agree that the RAID wear factor depends on implementation, and it could be positive or negative. For example, if you manage to (almost) always write full stripes in a wide-stripe design, then your single/dual parity overhead is actually better than a simple mirror. Instead of duplicating the data, you only need to write the extra parity block, which could be as low as only 10% extra. In hybrid designs, you can actually completely bypass flash for "wholesale" writes (such as HD video streams), and send them directly to HDD, that way reducing flash wear even further.